{ "registry_version": "stem-ai-airi-registry-v1", "upstream_name": "The AI Risk Repository", "upstream_source_url": "https://airisk.mit.edu/", "upstream_version": "V4_03", "upstream_snapshot_date": "2026-04-23", "upstream_license": "MIT", "upstream_row_count": 1595, "local_transform_version": "stem-ai-airi-transform-v1", "local_transform_date": "2026-05-12", "attribution_note": "Derived from The AI Risk Repository V4_03. Original source remains MIT-licensed and must be attributed in README, docs, runtime artifacts, and local registry metadata.", "risks": [ { "id": "01.01.00", "title": "Type 1: Diffusion of responsibility", "description": "Societal-scale harm can arise from AI built by a diffuse collection of creators, where no one is uniquely accountable for the technology's creation or use, as in a classic \"tragedy of the commons\".", "level": "category", "parent_category_id": "", "subdomain_id": "6.5", "paper_ref": "Critch2023", "paper_title": "TASRA: a Taxonomy and Analysis of Societal-Scale Risks from AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "01.02.00", "title": "Type 2: Bigger than expected", "description": "Harm can result from AI that was not expected to have a large impact at all, such as a lab leak, a surprisingly addictive open-source product, or an unexpected repurposing of a research prototype.", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Critch2023", "paper_title": "TASRA: a Taxonomy and Analysis of Societal-Scale Risks from AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "01.03.00", "title": "Type 3: Worse than expected", "description": "AI intended to have a large societal impact can turn out harmful by mistake, such as a popular product that creates problems and partially solves them only for its users.", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Critch2023", "paper_title": "TASRA: a Taxonomy and Analysis of Societal-Scale Risks from AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "01.04.00", "title": "Type 4: Willful indifference", "description": "As a side effect of a primary goal like profit or influence, AI creators can willfully allow it to cause widespread societal harms like pollution, resource depletion, mental illness, misinformation, or injustice.", "level": "category", "parent_category_id": "", "subdomain_id": "6.4", "paper_ref": "Critch2023", "paper_title": "TASRA: a Taxonomy and Analysis of Societal-Scale Risks from AI", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "01.05.00", "title": "Type 5: Criminal weaponization", "description": "One or more criminal entities could create AI to intentionally inflict harms, such as for terrorism or combating law enforcement.", "level": "category", "parent_category_id": "", "subdomain_id": "4.2", "paper_ref": "Critch2023", "paper_title": "TASRA: a Taxonomy and Analysis of Societal-Scale Risks from AI", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "01.06.00", "title": "Type 6: State Weaponization", "description": "AI deployed by states in war, civil war, or law enforcement can easily yield societal-scale harm", "level": "category", "parent_category_id": "", "subdomain_id": "4.2", "paper_ref": "Critch2023", "paper_title": "TASRA: a Taxonomy and Analysis of Societal-Scale Risks from AI", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "02.01.00", "title": "Harmful Content", "description": "\"The LLM-generated content sometimes contains biased, toxic, and private information\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.2", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "02.01.01", "title": "Bias", "description": "\"The training datasets of LLMs may contain biased information that leads LLMs to generate outputs with social biases\"", "level": "subcategory", "parent_category_id": "02.01.00", "subdomain_id": "1.1", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "02.01.02", "title": "Toxicity", "description": "\"Toxicity means the generated content contains rude, disrespectful, and even illegal information\"", "level": "subcategory", "parent_category_id": "02.01.00", "subdomain_id": "1.2", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "02.01.03", "title": "Privacy Leakage", "description": "\"Privacy Leakage means the generated content includes sensitive personal information\"", "level": "subcategory", "parent_category_id": "02.01.00", "subdomain_id": "2.1", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "02.02.00", "title": "Untruthful Content", "description": "\"The LLM-generated content could contain inaccurate information\"", "level": "category", "parent_category_id": "", "subdomain_id": "3.1", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "02.02.01", "title": "Factuality Errors", "description": "\"The LLM-generated content could contain inaccurate information\" which is factually incorrect", "level": "subcategory", "parent_category_id": "02.02.00", "subdomain_id": "3.1", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "02.02.02", "title": "Faithfulness Errors", "description": "\"The LLM-generated content could contain inaccurate information\" which is is not true to the source material or input used", "level": "subcategory", "parent_category_id": "02.02.00", "subdomain_id": "3.1", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "02.03.00", "title": "Unhelpful Uses", "description": "\"Improper uses of LLM systems can cause adverse social impacts.\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.3", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "02.03.01", "title": "Academic Misconduct", "description": "\"Improper use of LLM systems (i.e., abuse of LLM systems) will cause adverse social impacts, such as academic misconduct.\"", "level": "subcategory", "parent_category_id": "02.03.00", "subdomain_id": "4.3", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "02.03.02", "title": "Copyright Violation", "description": "\"LLM systems may output content similar to existing works, infringing on copyright owners.\"", "level": "subcategory", "parent_category_id": "02.03.00", "subdomain_id": "6.3", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "02.03.03", "title": "Cyber Attacks", "description": "\"Hackers can obtain malicious code in a low-cost and efficient manner to automate cyber attacks with powerful LLM systems.\"", "level": "subcategory", "parent_category_id": "02.03.00", "subdomain_id": "4.2", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "02.03.04", "title": "Software Vulnerabilities", "description": "\"Programmers are accustomed to using code generation tools such as Github Copilot for program development, which may bury vulnerabilities in the program.\"", "level": "subcategory", "parent_category_id": "02.03.00", "subdomain_id": "2.2", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "02.04.00", "title": "Software Security Issues", "description": "\"The software development toolchain of LLMs is complex and could bring threats to the developed LLM.\"", "level": "category", "parent_category_id": "", "subdomain_id": "2.2", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "02.04.01", "title": "Programming Language", "description": "\"Most LLMs are developed using the Python language, whereas the vulnerabilities of Python interpreters pose threats to the developed models\"", "level": "subcategory", "parent_category_id": "02.04.00", "subdomain_id": "2.2", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": true }, { "id": "02.04.02", "title": "Deep Learning Frameworks", "description": "\"LLMs are implemented based on deep learning frameworks. Notably, various vulnerabilities in these frameworks have been disclosed in recent years. As reported in the past five years, three of the most common types of vulnerabilities are buffer overflow attacks, memory corruption, and input validation issues.\"", "level": "subcategory", "parent_category_id": "02.04.00", "subdomain_id": "2.2", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "02.04.03", "title": "Software Supply Chains", "description": "\"The software development toolchain of LLMs is complex and could bring threats to the developed LLM.\"", "level": "subcategory", "parent_category_id": "02.04.00", "subdomain_id": "2.2", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "02.04.04", "title": "Pre-processing Tools", "description": "\"Pre-processing tools play a crucial role in the context of LLMs. These tools, which are often involved in computer vision (CV) tasks, are susceptible to attacks that exploit vulnerabilities in tools such as OpenCV.\"", "level": "subcategory", "parent_category_id": "02.04.00", "subdomain_id": "2.2", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "02.05.00", "title": "Hardware Vulnerabilities", "description": "\"The vulnerabilities of hardware systems for training and inferencing brings issues to LLM-based applications.\"", "level": "category", "parent_category_id": "", "subdomain_id": "2.2", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "02.05.01", "title": "Network Devices", "description": "\"The training of LLMs often relies on distributed network systems [171], [172]. During the transmission of gradients through the links between GPU server nodes, significant volumetric traffic is generated. This traffic can be susceptible to disruption by burst traffic, such as pulsating attacks [161]. Furthermore, distributed training frameworks may encounter congestion issues [173].\"", "level": "subcategory", "parent_category_id": "02.05.00", "subdomain_id": "2.2", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "02.05.02", "title": "GPU Computation Platforms", "description": "\"The training of LLMs requires significant GPU resources, thereby introducing an additional security concern. GPU side-channel attacks have been developed to extract the parameters of trained models [159], [163].\"", "level": "subcategory", "parent_category_id": "02.05.00", "subdomain_id": "2.2", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "02.05.03", "title": "Memory and Storage", "description": "\"Similar to conventional programs, hardware infrastructures can also introduce threats to LLMs. Memory-related vulnerabilities, such as rowhammer attacks [160], can be leveraged to manipulate the parameters of LLMs, giving rise to attacks such as the Deephammer attack [167], [168].\"", "level": "subcategory", "parent_category_id": "02.05.00", "subdomain_id": "2.2", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "02.06.00", "title": "Issues on External Tools", "description": "\"The external tools (e.g., web APIs) present trustworthiness and privacy issues to LLM-based applications.\"", "level": "category", "parent_category_id": "", "subdomain_id": "2.2", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "02.06.01", "title": "Factual Errors Injected by External Tools", "description": "\"External tools typically incorporate additional knowledge into the input prompts [122], [178]–[184]. The additional knowledge often originates from public resources such as Web APIs and search engines. As the reliability of external tools is not always ensured, the content returned by external tools may include factual errors, consequently amplifying the hallucination issue.\"", "level": "subcategory", "parent_category_id": "02.06.00", "subdomain_id": "2.2", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "02.06.02", "title": "Exploiting External Tools for Attacks", "description": "\"Adversarial tool providers can embed malicious instructions in the APIs or prompts [84], leading LLMs to leak memorized sensitive information in the training data or users’ prompts (CVE2023-32786). As a result, LLMs lack control over the output, resulting in sensitive information being disclosed to external tool providers. Besides, attackers can easily manipulate public data to launch targeted attacks, generating specific malicious outputs according to user inputs. Furthermore, feeding the information from external tools into LLMs may lead to injection attacks [61]. For example, unverified inputs may result in arbitrary code execution (CVE-2023-29374).\"", "level": "subcategory", "parent_category_id": "02.06.00", "subdomain_id": "2.2", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "02.07.00", "title": "Privacy Leakage", "description": "\"The model is trained with personal data in the corpus and unintentionally exposing them during the conversation.\"", "level": "category", "parent_category_id": "", "subdomain_id": "2.1", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "02.07.01", "title": "Private Training Data", "description": "\"As recent LLMs continue to incorporate licensed, created, and publicly available data sources in their corpora, the potential to mix private data in the training corpora is significantly increased. The misused private data, also named as personally identifiable information (PII) [84], [86], could contain various types of sensitive data subjects, including an individual person’s name, email, phone number, address, education, and career. Generally, injecting PII into LLMs mainly occurs in two settings — the exploitation of web-collection data and the alignment with personal humanmachine conversations [87]. Specifically, the web-collection data can be crawled from online sources with sensitive PII, and the personal human-machine conversations could be collected for SFT and RLHF\"", "level": "subcategory", "parent_category_id": "02.07.00", "subdomain_id": "2.1", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": true }, { "id": "02.07.02", "title": "Memorization in LLMs", "description": "\"Memorization in LLMs refers to the capability to recover the training data with contextual prefixes. According to [88]–[90], given a PII entity x, which is memorized by a model F. Using a prompt p could force the model F to produce the entity x, where p and x exist in the training data. For instance, if the string “Have a good day!\\n alice@email.com” is present in the training data, then the LLM could accurately predict Alice’s email when given the prompt “Have a good day!\\n”.\"", "level": "subcategory", "parent_category_id": "02.07.00", "subdomain_id": "2.1", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": true }, { "id": "02.07.03", "title": "Association in LLMs", "description": "\"Association in LLMs refers to the capability to associate various pieces of information related to a person. According to [68], [86], given a pair of PII entities (xi , xj ), which is associated by a model F. Using a prompt p could force the model F to produce the entity xj , where p is the prompt related to the entity xi . For instance, an LLM could accurately output the answer when given the prompt “The email address of Alice is”, if the LLM associates Alice with her email “alice@email.com”. L\"", "level": "subcategory", "parent_category_id": "02.07.00", "subdomain_id": "2.1", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": true }, { "id": "02.08.00", "title": "Toxicity and Bias Tendencies", "description": "\"Extensive data collection in LLMs brings toxic content and stereotypical bias into the training data.\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.1", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "02.08.01", "title": "Toxic Training Data", "description": "\"Following previous studies [96], [97], toxic data in LLMs is defined as rude, disrespectful, or unreasonable language that is opposite to a polite, positive, and healthy language environment, including hate speech, offensive utterance, profanities, and threats [91].\"", "level": "subcategory", "parent_category_id": "02.08.00", "subdomain_id": "1.2", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "02.08.02", "title": "Biased Training Data", "description": "\"Compared with the definition of toxicity, the definition of bias is more subjective and contextdependent. Based on previous work [97], [101], we describe the bias as disparities that could raise demographic differences among various groups, which may involve demographic word prevalence and stereotypical contents. Concretely, in massive corpora, the prevalence of different pronouns and identities could influence an LLM’s tendency about gender, nationality, race, religion, and culture [4]. For instance, the pronoun He is over-represented compared with the pronoun She in the training corpora, leading LLMs to learn less context about She and thus generate He with a higher probability [4], [102]. Furthermore, stereotypical bias [103] which refers to overgeneralized beliefs about a particular group of people, usually keeps incorrect values and is hidden in the large-scale benign contents. In effect, defining what should be regarded as a stereotype in the corpora is still an open problem.\"", "level": "subcategory", "parent_category_id": "02.08.00", "subdomain_id": "1.1", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "02.09.00", "title": "Hallucinations", "description": "\"LLMs generate nonsensical, untruthful, and factual incorrect content\"", "level": "category", "parent_category_id": "", "subdomain_id": "3.1", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "02.09.01", "title": "Knowledge Gaps", "description": "\"Since the training corpora of LLMs can not contain all possible world knowledge [114]–[119], and it is challenging for LLMs to grasp the long-tail knowledge within their training data [120], [121], LLMs inherently possess knowledge boundaries [107]. Therefore, the gap between knowledge involved in an input prompt and knowledge embedded in the LLMs can lead to hallucinations\"", "level": "subcategory", "parent_category_id": "02.09.00", "subdomain_id": "3.1", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "02.09.02", "title": "Noisy Training Data", "description": "\"Another important source of hallucinations is the noise in training data, which introduces errors in the knowledge stored in model parameters [111]–[113]. Generally, the training data inherently harbors misinformation. When training on large-scale corpora, this issue becomes more serious because it is difficult to eliminate all the noise from the massive pre-training data.\"", "level": "subcategory", "parent_category_id": "02.09.00", "subdomain_id": "3.1", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": true }, { "id": "02.09.03", "title": "Defective Decoding Process", "description": "In general, LLMs employ the Transformer architecture [32] and generate content in an autoregressive manner, where the prediction of the next token is conditioned on the previously generated token sequence. Such a scheme could accumulate errors [105]. Besides, during the decoding process, top-p sampling [28] and top-k sampling [27] are widely adopted to enhance the diversity of the generated content. Nevertheless, these sampling strategies can introduce “randomness” [113], [136], thereby increasing the potential of hallucinations\"", "level": "subcategory", "parent_category_id": "02.09.00", "subdomain_id": "3.1", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": true }, { "id": "02.09.04", "title": "False Recall of Memorized Information", "description": "\"Although LLMs indeed memorize the queried knowledge, they may fail to recall the corresponding information [122]. That is because LLMs can be confused by co-occurance patterns [123], positional patterns [124], duplicated data [125]–[127] and similar named entities [113].\"", "level": "subcategory", "parent_category_id": "02.09.00", "subdomain_id": "3.1", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "02.09.05", "title": "Pursuing Consistent Context", "description": "\"LLMs have been demonstrated to pursue consistent context [129]–[132], which may lead to erroneous generation when the prefixes contain false information. Typical examples include sycophancy [129], [130], false demonstrations-induced hallucinations [113], [133], and snowballing [131]. As LLMs are generally fine-tuned with instruction-following data and user feedback, they tend to reiterate user-provided opinions [129], [130], even though the opinions contain misinformation. Such a sycophantic behavior amplifies the likelihood of generating hallucinations, since the model may prioritize user opinions over facts.\"", "level": "subcategory", "parent_category_id": "02.09.00", "subdomain_id": "3.1", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "02.10.00", "title": "Model Attacks", "description": "Model attacks exploit the vulnerabilities of LLMs, aiming to steal valuable information or lead to incorrect responses.", "level": "category", "parent_category_id": "", "subdomain_id": "2.2", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "02.10.01", "title": "Extraction Attacks", "description": "\"Extraction attacks [137] allow an adversary to query a black-box victim model and build a substitute model by training on the queries and responses. The substitute model could achieve almost the same performance as the victim model. While it is hard to fully replicate the capabilities of LLMs, adversaries could develop a domainspecific model that draws domain knowledge from LLMs\"", "level": "subcategory", "parent_category_id": "02.10.00", "subdomain_id": "2.2", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "02.10.02", "title": "Inference Attacks", "description": "\"Inference attacks [150] include membership inference attacks, property inference attacks, and data reconstruction attacks. These attacks allow an adversary to infer the composition or property information of the training data. Previous works [67] have demonstrated that inference attacks could easily work in earlier PLMs, implying that LLMs are also possible to be attacked\"", "level": "subcategory", "parent_category_id": "02.10.00", "subdomain_id": "2.2", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "02.10.03", "title": "Poisoning Attacks", "description": "\"Poisoning attacks [143] could influence the behavior of the model by making small changes to the training data. A number of efforts could even leverage data poisoning techniques to implant hidden triggers into models during the training process (i.e., backdoor attacks). Many kinds of triggers in text corpora (e.g., characters, words, sentences, and syntax) could be used by the attackers.\"\"", "level": "subcategory", "parent_category_id": "02.10.00", "subdomain_id": "2.2", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "02.10.04", "title": "Overhead Attacks", "description": "\"Overhead attacks [146] are also named energy-latency attacks. For example, an adversary can design carefully crafted sponge examples to maximize energy consumption in an AI system. Therefore, overhead attacks could also threaten the platforms integrated with LLMs.\"", "level": "subcategory", "parent_category_id": "02.10.00", "subdomain_id": "2.2", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "02.10.05", "title": "Novel Attacks on LLMs", "description": "Table of examples has: \"Prompt Abstraction Attacks [147]: Abstracting queries to cost lower prices using LLM’s API. Reward Model Backdoor Attacks [148]: Constructing backdoor triggers on LLM’s RLHF process. LLM-based Adversarial Attacks [149]: Exploiting LLMs to construct samples for model attacks\"", "level": "subcategory", "parent_category_id": "02.10.00", "subdomain_id": "2.2", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "02.10.06", "title": "Evasion Attacks", "description": "\"Evasion attacks [145] target to cause significant shifts in model’s prediction via adding perturbations in the test samples to build adversarial examples. In specific, the perturbations can be implemented based on word changes, gradients, etc.\"", "level": "subcategory", "parent_category_id": "02.10.00", "subdomain_id": "2.2", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "02.11.00", "title": "Not-Suitable-for-Work (NSFW) Prompts", "description": "\"Inputting a prompt contain an unsafe topic (e.g., notsuitable-for-work (NSFW) content) by a benign user.\r\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.2", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "02.12.00", "title": "Adversarial Prompts", "description": "\"Engineering an adversarial input to elicit an undesired model behavior, which pose a clear attack intention\"", "level": "category", "parent_category_id": "", "subdomain_id": "2.2", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "02.12.01", "title": "Goal Hijacking", "description": "\"Goal hijacking is a type of primary attack in prompt injection [58]. By injecting a phrase like “Ignore the above instruction and do ...” in the input, the attack could hijack the original goal of the designed prompt (e.g., translating tasks) in LLMs and execute the new goal in the injected phrase.\"", "level": "subcategory", "parent_category_id": "02.12.00", "subdomain_id": "2.2", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "02.12.02", "title": "One-step Jailbreaks", "description": "\"One-step jailbreaks. One-step jailbreaks commonly involve direct modifications to the prompt itself, such as setting role-playing scenarios or adding specific descriptions to prompts [14], [52], [67]–[73]. Role-playing is a prevalent method used in jailbreaking by imitating different personas [74]. Such a method is known for its efficiency and simplicity compared to more complex techniques that require domain knowledge [73]. Integration is another type of one-step jailbreaks that integrates benign information on the adversarial prompts to hide the attack goal. For instance, prefix integration is used to integrate an innocuous-looking prefix that is less likely to be rejected based on its pre-trained distributions [75]. Additionally, the adversary could treat LLMs as a program and encode instructions indirectly through code integration or payload splitting [63]. Obfuscation is to add typos or utilize synonyms for terms that trigger input or output filters. Obfuscation methods include the use of the Caesar cipher [64], leetspeak (replacing letters with visually similar numbers and symbols), and Morse code [76]. Besides, at the word level, an adversary may employ Pig Latin to replace sensitive words with synonyms or use token smuggling [77] to split sensitive words into substrings.\"", "level": "subcategory", "parent_category_id": "02.12.00", "subdomain_id": "2.2", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "02.12.03", "title": "Multi-step Jailbreaks", "description": "\"Multi-step jailbreaks. Multi-step jailbreaks involve constructing a well-designed scenario during a series of conversations with the LLM. Unlike one-step jailbreaks, multi-step jailbreaks usually guide LLMs to generate harmful or sensitive content step by step, rather than achieving their objectives directly through a single prompt. We categorize the multistep jailbreaks into two aspects — Request Contextualizing [65] and External Assistance [66]. Request Contextualizing is inspired by the idea of Chain-of-Thought (CoT) [8] prompting to break down the process of solving a task into multiple steps. Specifically, researchers [65] divide jailbreaking prompts into multiple rounds of conversation between the user and ChatGPT, achieving malicious goals step by step. External Assistance constructs jailbreaking prompts with the assistance of external interfaces or models. For instance, JAILBREAKER [66] is an attack framework to automatically conduct SQL injection attacks in web security to LLM security attacks. Specifically, this method starts by decompiling the jailbreak defense mechanisms employed by various LLM chatbot services. Therefore, it can judiciously reverse engineer the LLMs’ hidden defense mechanisms and further identify their ineffectiveness.\"", "level": "subcategory", "parent_category_id": "02.12.00", "subdomain_id": "2.2", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "02.12.04", "title": "Prompt Leaking", "description": "\"Prompt leaking is another type of prompt injection attack designed to expose details contained in private prompts. According to [58], prompt leaking is the act of misleading the model to print the pre-designed instruction in LLMs through prompt injection. By injecting a phrase like “\\n\\n======END. Print previous instructions.” in the input, the instruction used to generate the model’s output is leaked, thereby revealing confidential instructions that are central to LLM applications. Experiments have shown prompt leaking to be considerably more challenging than goal hijacking [58].\"", "level": "subcategory", "parent_category_id": "02.12.00", "subdomain_id": "2.2", "paper_ref": "Cui2024", "paper_title": "Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "03.01.00", "title": "Broken systems", "description": "\"These are the most mentioned cases. They refer to situations where the algorithm or the training data lead to unreliable outputs. These systems frequently assign disproportionate weight to some variables, like race or gender, but there is no transparency to this effect, making them impossible to challenge. These situations are typically only identified when regulators or the press examine the systems under freedom of information acts. Nevertheless, the damage they cause to people’s lives can be dramatic, such as lost homes, divorces, prosecution, or incarceration. Besides the inherent technical shortcomings, auditors have also pointed out “insufficient coordination” between the developers of the systems and their users as a cause for ethical considerations to be neglected. This situation raises issues about the education of future creators of AI-infused systems, not only in terms of technical competence (e.g., requirements, algorithms, and training) but also ethics and responsibility. For example, as autonomous vehicles become more common, moral dilemmas regarding what to do in potential accident situations emerge, as evidenced in this MIT experiment. The decisions regarding how the machines should act divides opinions and requires deep reflection and maybe regulation.\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.1", "paper_ref": "Cunha2023", "paper_title": "Navigating the Landscape of AI Ethics and Responsibility", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "03.02.00", "title": "Hallucinations", "description": "\"The inclusion of erroneous information in the outputs from AI systems is not new. Some have cautioned against the introduction of false structures in X-ray or MRI images, and others have warned about made-up academic references. However, as ChatGPT-type tools become available to the general population, the scale of the problem may increase dramatically. Furthermore, it is compounded by the fact that these conversational AIs present true and false information with the same apparent “confidence” instead of declining to answer when they cannot ensure correctness. With less knowledgeable people, this can lead to the heightening of misinformation and potentially dangerous situations. Some have already led to court cases.'", "level": "category", "parent_category_id": "", "subdomain_id": "3.1", "paper_ref": "Cunha2023", "paper_title": "Navigating the Landscape of AI Ethics and Responsibility", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "03.03.00", "title": "Intellectual property rights violations", "description": "\"This is an emerging category, with more cases prone to appear as the use of generative AI tools–such as Stable Diffusion, Midjourney, or ChatGPT–becomes more widespread. Some content creators are already suing for the appropriation of their work to train AI algorithms without a request for permission or compensation. Perhaps even more damaging cases will appear as developers increasingly ask chatbots or assistants like CoPilot for ready-to-use computer code. Even if these AI tools have learned only from open-source software (OSS) projects, which is not a given, there are still serious issues to consider, as not all OSS licenses are equal, and some are incompatible with others, meaning that it is illegal to mix them in the same product. Even worse, some licenses, such as GPL, are viral, meaning that any code that uses a GPL component must legally be made available under that same license. In the past, companies have suffered injunctions or been forced to make their proprietary source code available because of carelessly using a GPL library.\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.3", "paper_ref": "Cunha2023", "paper_title": "Navigating the Landscape of AI Ethics and Responsibility", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "03.04.00", "title": "Privacy and regulation violations", "description": "\"Some of the broken systems discussed above are also very invasive of people’s privacy, controlling, for instance, the length of someone’s last romantic relationship [51]. More recently, ChatGPT was banned in Italy over privacy concerns and potential violation of the European Union’s (EU) General Data Protection Regulation (GDPR) [52]. The Italian data-protection authority said, “the app had experienced a data breach involving user conversations and payment information.” It also claimed that there was no legal basis to justify “the mass collection and storage of personal data for the purpose of ‘training’ the algorithms underlying the operation of the platform,” among other concerns related to the age of the users [52]. Privacy regulators in France, Ireland, and Germany could follow in Italy’s footsteps [53]. Coincidentally, it has recently become public that Samsung employees have inadvertently leaked trade secrets by using ChatGPT to assist in preparing notes for a presentation and checking and optimizing source code [54, 55]. Another example of testing the ethics and regulatory limits can be found in actions of the facial recognition company Clearview AI, which “scraped the public web—social media, employment sites, YouTube, Venmo—to create a database with three billion images of people, along with links to the webpages from which the photos had come” [56]. Trials of this unregulated database have been offered to individual law enforcement officers who often use it without their department’s approval [57]. In Sweden, such illegal use by the police force led to a fine of e250,000 by the country’s data watchdog [57].\"", "level": "category", "parent_category_id": "", "subdomain_id": "2.1", "paper_ref": "Cunha2023", "paper_title": "Navigating the Landscape of AI Ethics and Responsibility", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "03.05.00", "title": "Enabling malicious actors and harmful actions", "description": "\"Some uses of AI have been deeply concerning, namely voice cloning [58] and the generation of deep fake videos [59]. For example, in March 2022, in the early days of the Russian invasion of Ukraine, hackers broadcast via the Ukrainian news website Ukraine 24 a deep fake video of President Volodymyr Zelensky capitulating and calling on his soldiers to lay down their weapons [60]. The necessary software to create these fakes is readily available on the Internet, and the hardware requirements are modest by today’s standards [61]. Other nefarious uses of AI include accelerating password cracking [62] or enabling otherwise unskilled people to create software exploits [63, 64], or effective phishing e-mails [65]. Although some believe that powerful AI models should be prevented from running on personal computers to retain some control, others demonstrate how inglorious that effort may be [66]. Furthermore, as ChatGPT-type systems evolve from conversational systems to agents, capable of acting autonomously and performing tasks with little human intervention, like Auto-GPT [67], new risks emerge.\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.0", "paper_ref": "Cunha2023", "paper_title": "Navigating the Landscape of AI Ethics and Responsibility", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "03.06.00", "title": "Environmental and socioeconomic harms", "description": "\"At a time of increasing climate urgency,\renergy consumption and the carbon footprint of AI applications are also matters of ethics\rand responsibility [68]. As with other energy-intensive technologies like proof-of-work\rblockchain, the call is to research more environmentally sustainable algorithms to offset\rthe increasing use scale.\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.6", "paper_ref": "Cunha2023", "paper_title": "Navigating the Landscape of AI Ethics and Responsibility", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "04.01.00", "title": "Toxicity and Abusive Content", "description": "This typically refers to rude, harmful, or inappropriate expressions.", "level": "category", "parent_category_id": "", "subdomain_id": "1.2", "paper_ref": "Deng2023", "paper_title": "Towards Safer Generative Language Models: A Survey on Safety Risks, Evaluations, and Improvements", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "04.02.00", "title": "Unfairness and Discrimination", "description": "Social bias is an unfairly negative attitude towards a social group or individuals based on one-sided or inaccurate information, typically pertaining to widely disseminated negative stereotypes regarding gender, race, religion, etc.", "level": "category", "parent_category_id": "", "subdomain_id": "1.1", "paper_ref": "Deng2023", "paper_title": "Towards Safer Generative Language Models: A Survey on Safety Risks, Evaluations, and Improvements", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "04.03.00", "title": "Ethics and Morality Issues", "description": "LMs need to pay more attention to universally accepted societal values at the level of ethics and morality, including the judgement of right and wrong, and its relationship with social norms and laws.", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Deng2023", "paper_title": "Towards Safer Generative Language Models: A Survey on Safety Risks, Evaluations, and Improvements", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "04.04.00", "title": "Controversial Opinions", "description": "The controversial views expressed by large models are also a widely discussed concern. Bang et al. (2021) evaluated several large models and found that they occasionally express inappropriate or extremist views when discussing political top-ics. Furthermore, models like ChatGPT (OpenAI, 2022) that claim political neutrality and aim to provide objective information for users have been shown to exhibit notable left-leaning political biases in areas like economics, social policy, foreign affairs, and civil liberties.", "level": "category", "parent_category_id": "", "subdomain_id": "1.2", "paper_ref": "Deng2023", "paper_title": "Towards Safer Generative Language Models: A Survey on Safety Risks, Evaluations, and Improvements", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "04.05.00", "title": "Misleading Information", "description": "Large models are usually susceptible to hallucination problems, sometimes yielding nonsensical or unfaithful data that results in misleading outputs.", "level": "category", "parent_category_id": "", "subdomain_id": "3.1", "paper_ref": "Deng2023", "paper_title": "Towards Safer Generative Language Models: A Survey on Safety Risks, Evaluations, and Improvements", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "04.06.00", "title": "Privacy and Data Leakage", "description": "Large pre-trained models trained on internet texts might contain private information like phone numbers, email addresses, and residential addresses.", "level": "category", "parent_category_id": "", "subdomain_id": "2.1", "paper_ref": "Deng2023", "paper_title": "Towards Safer Generative Language Models: A Survey on Safety Risks, Evaluations, and Improvements", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": true }, { "id": "04.07.00", "title": "Malicious Use and Unleashing AI Agents", "description": "LMs, due to their remarkable capabilities, carry the same potential for malice as other technological products. For instance, they may be used in information warfare to generate deceptive information or unlawful content, thereby having a significant impact on individuals and society. As current LMs are increasingly built as agents to accomplish user objectives, they may disregard the moral and safety guidelines if operating without adequate supervision. Instead, they may execute user commands mechanically without considering the potential damage. They might interact unpredictably with humans and other systems, especially in open environments", "level": "category", "parent_category_id": "", "subdomain_id": "4.0", "paper_ref": "Deng2023", "paper_title": "Towards Safer Generative Language Models: A Survey on Safety Risks, Evaluations, and Improvements", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "05.01.00", "title": "Fairness - Bias", "description": "Fairness is, by far, the most discussed issue in the literature, remaining a paramount concern especially in case of LLMs and text-to-image models. This is sparked by training data biases propagating into model outputs, causing negative effects like stereotyping, racism, sexism, ideological leanings, or the marginalization of minorities. Next to attesting generative AI a conservative inclination by perpetuating existing societal patterns, there is a concern about reinforcing existing biases when training new generative models with synthetic data from previous models. Beyond technical fairness issues, critiques in the literature extend to the monopolization or centralization of power in large AI labs, driven by the substantial costs of developing foundational models. The literature also highlights the problem of unequal access to generative AI, particularly in developing countries or among financially constrained groups. Sources also analyze challenges of the AI research community to ensure workforce diversity. Moreover, there are concerns regarding the imposition of values embedded in AI systems on cultures distinct from those where the systems were developed.", "level": "category", "parent_category_id": "", "subdomain_id": "1.1", "paper_ref": "Hagendorff2024", "paper_title": "Mapping the Ethics of Generative AI: A Comprehensive Scoping Review", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "05.02.00", "title": "Safety", "description": "A primary concern is the emergence of human-level or superhuman generative models, commonly referred to as AGI, and their potential existential or catastrophic risks to humanity. Connected to that, AI safety aims at avoiding deceptive or power-seeking machine behavior, model self-replication, or shutdown evasion. Ensuring controllability, human oversight, and the implementation of red teaming measures are deemed to be essential in mitigating these risks, as is the need for increased AI safety research and promoting safety cultures within AI organizations instead of fueling the AI race. Furthermore, papers thematize risks from unforeseen emerging capabilities in generative models, restricting access to dangerous research works, or pausing AI research for the sake of improving safety or governance measures first. Another central issue is the fear of weaponizing AI or leveraging it for mass destruction, especially by using LLMs for the ideation and planning of how to attain, modify, and disseminate biological agents. In general, the threat of AI misuse by malicious individuals or groups, especially in the context of open-source models, is highlighted in the literature as a significant factor emphasizing the critical importance of implementing robust safety measures.", "level": "category", "parent_category_id": "", "subdomain_id": "7.1", "paper_ref": "Hagendorff2024", "paper_title": "Mapping the Ethics of Generative AI: A Comprehensive Scoping Review", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "05.03.00", "title": "Harmful Content - Toxicity", "description": "Generating unethical, fraudulent, toxic, violent, pornographic, or other harmful content is a further predominant concern, again focusing notably on LLMs and text-to-image models. Numerous studies highlight the risks associated with the intentional creation of disinformation, fake news, propaganda, or deepfakes, underscoring their significant threat to the integrity of public discourse and the trust in credible media. Additionally, papers explore the potential for generative models to aid in criminal activities, incidents of self-harm, identity theft, or impersonation. Furthermore, the literature investigates risks posed by LLMs when generating advice in high-stakes domains such as health, safety-related issues, as well as legal or financial matters.", "level": "category", "parent_category_id": "", "subdomain_id": "1.2", "paper_ref": "Hagendorff2024", "paper_title": "Mapping the Ethics of Generative AI: A Comprehensive Scoping Review", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "05.04.00", "title": "Hallucinations", "description": "Significant concerns are raised about LLMs inadvertently generating false or misleading information, as well as erroneous code. Papers not only critically analyze various types of reasoning errors in LLMs but also examine risks associated with specific types of misinformation, such as medical hallucinations. Given the propensity of LLMs to produce flawed outputs accompanied by overconfident rationales and fabricated references, many sources stress the necessity of manually validating and fact-checking the outputs of these models.", "level": "category", "parent_category_id": "", "subdomain_id": "3.1", "paper_ref": "Hagendorff2024", "paper_title": "Mapping the Ethics of Generative AI: A Comprehensive Scoping Review", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "05.05.00", "title": "Privacy", "description": "Generative AI systems, similar to traditional machine learning methods, are considered a threat to privacy and data protection norms. A major concern is the intended extraction or inadvertent leakage of sensitive or private information from LLMs. To mitigate this risk, strategies such as sanitizing training data to remove sensitive information or employing synthetic data for training are proposed.", "level": "category", "parent_category_id": "", "subdomain_id": "2.1", "paper_ref": "Hagendorff2024", "paper_title": "Mapping the Ethics of Generative AI: A Comprehensive Scoping Review", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "05.06.00", "title": "Interaction risks", "description": "Many novel risks posed by generative AI stem from the ways in which humans interact with these systems. For instance, sources discuss epistemic challenges in distinguishing AI-generated from human content. They also address the issue of anthropomorphization, which can lead to an excessive trust in generative AI systems. On a similar note, many papers argue that the use of conversational agents could impact mental well-being or gradually supplant interpersonal communication, potentially leading to a dehumanization of interactions. Additionally, a frequently discussed interaction risk in the literature is the potential of LLMs to manipulate human behavior or to instigate users to engage in unethical or illegal activities.", "level": "category", "parent_category_id": "", "subdomain_id": "5.1", "paper_ref": "Hagendorff2024", "paper_title": "Mapping the Ethics of Generative AI: A Comprehensive Scoping Review", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "05.07.00", "title": "Security - Robustness", "description": "While AI safety focuses on threats emanating from generative AI systems, security centers on threats posed to these systems. The most extensively discussed issue in this context are jailbreaking risks, which involve techniques like prompt injection or visual adversarial examples designed to circumvent safety guardrails governing model behavior. Sources delve into various jailbreaking methods, such as role play or reverse exposure. Similarly, implementing backdoors or using model poisoning techniques bypass safety guardrails as well. Other security concerns pertain to model or prompt thefts.", "level": "category", "parent_category_id": "", "subdomain_id": "2.2", "paper_ref": "Hagendorff2024", "paper_title": "Mapping the Ethics of Generative AI: A Comprehensive Scoping Review", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "05.08.00", "title": "Education - Learning", "description": "In contrast to traditional machine learning, the impact of generative AI in the educational sector receives considerable attention in the academic literature. Next to issues stemming from difficulties to distinguish student-generated from AI-generated content, which eventuates in various opportunities to cheat in online or written exams, sources emphasize the potential benefits of generative AI in enhancing learning and teaching methods, particularly in relation to personalized learning approaches. However, some papers suggest that generative AI might lead to reduced effort or laziness among learners. Additionally, a significant focus in the literature is on the promotion of literacy and education about generative AI systems themselves, such as by teaching prompt engineering techniques.", "level": "category", "parent_category_id": "", "subdomain_id": "4.3", "paper_ref": "Hagendorff2024", "paper_title": "Mapping the Ethics of Generative AI: A Comprehensive Scoping Review", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "05.09.00", "title": "Alignment", "description": "The general tenet of AI alignment involves training generative AI systems to be harmless, helpful, and honest, ensuring their behavior aligns with and respects human values. However, a central debate in this area concerns the methodological challenges in selecting appropriate values. While AI systems can acquire human values through feedback, observation, or debate, there remains ambiguity over which individuals are qualified or legitimized to provide these guiding signals. Another prominent issue pertains to deceptive alignment, which might cause generative AI systems to tamper evaluations. Additionally, many papers explore risks associated with reward hacking, proxy gaming, or goal misgeneralization in generative AI systems.", "level": "category", "parent_category_id": "", "subdomain_id": "7.1", "paper_ref": "Hagendorff2024", "paper_title": "Mapping the Ethics of Generative AI: A Comprehensive Scoping Review", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "05.10.00", "title": "Cybercrime", "description": "Closely related to discussions surrounding security and harmful content, the field of cybersecurity investigates how generative AI is misused for fraudulent online activities. A particular focus lies on social engineering attacks, for instance by utilizing generative AI to impersonate humans, creating fake identities, cloning voices, or crafting phishing messages. Another prevalent concern is the use of LLMs for generating malicious code or hacking.", "level": "category", "parent_category_id": "", "subdomain_id": "4.2", "paper_ref": "Hagendorff2024", "paper_title": "Mapping the Ethics of Generative AI: A Comprehensive Scoping Review", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "05.11.00", "title": "Governance - Regulation", "description": "In response to the multitude of new risks associated with generative AI, papers advocate for legal regulation and governmental oversight. The focus of these discussions centers on the need for international coordination in AI governance, the establishment of binding safety standards for frontier models, and the development of mechanisms to sanction non-compliance. Furthermore, the literature emphasizes the necessity for regulators to gain detailed insights into the research and development processes within AI labs. Moreover, risk management strategies of these labs shall be evaluated. However, the literature also acknowledges potential risks of overregulation, which could hinder innovation.", "level": "category", "parent_category_id": "", "subdomain_id": "6.5", "paper_ref": "Hagendorff2024", "paper_title": "Mapping the Ethics of Generative AI: A Comprehensive Scoping Review", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "05.12.00", "title": "Labor displacement - Economic impact", "description": "The literature frequently highlights concerns that generative AI systems could adversely impact the economy, potentially even leading to mass unemployment. This pertains to various fields, ranging from customer services to software engineering or crowdwork platforms. While new occupational fields like prompt engineering are created, the prevailing worry is that generative AI may exacerbate socioeconomic inequalities and lead to labor displacement. Additionally, papers debate potential large-scale worker deskilling induced by generative AI, but also productivity gains contingent upon outsourcing mundane or repetitive tasks to generative AI systems.", "level": "category", "parent_category_id": "", "subdomain_id": "6.2", "paper_ref": "Hagendorff2024", "paper_title": "Mapping the Ethics of Generative AI: A Comprehensive Scoping Review", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "05.13.00", "title": "Transparency - Explainability", "description": "Being a multifaceted concept, the term 'transparency' is both used to refer to technical explainability as well as organizational openness. Regarding the former, papers underscore the need for mechanistic interpretability and for explaining internal mechanisms in generative models. On the organizational front, transparency relates to practices such as informing users about capabilities and shortcomings of models, as well as adhering to documentation and reporting requirements for data collection processes or risk evaluations.", "level": "category", "parent_category_id": "", "subdomain_id": "7.4", "paper_ref": "Hagendorff2024", "paper_title": "Mapping the Ethics of Generative AI: A Comprehensive Scoping Review", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": true }, { "id": "05.15.00", "title": "Sustainability", "description": "Generative models are known for their substantial energy requirements, necessitating significant amounts of electricity, cooling water, and hardware containing rare metals. The extraction and utilization of these resources frequently occur in unsustainable ways. Consequently, papers highlight the urgency of mitigating environmental costs for instance by adopting renewable energy sources and utilizing energy-efficient hardware in the operation and training of generative AI systems.", "level": "category", "parent_category_id": "", "subdomain_id": "6.6", "paper_ref": "Hagendorff2024", "paper_title": "Mapping the Ethics of Generative AI: A Comprehensive Scoping Review", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "05.16.00", "title": "Art - Creativity", "description": "In this cluster, concerns about negative impacts on human creativity, particularly through text-to-image models, are prevalent. Papers criticize financial harms or economic losses for artists due to the widespread generation of synthetic art as well as the unauthorized and uncompensated use of artists' works in training datasets. Additionally, given the challenge of distinguishing synthetic images from authentic ones, there is a call for systematically disclosing the non-human origin of such content, particularly through watermarking. Moreover, while some sources argue that text-to-image models lack 'true' creativity or the ability to produce genuinely innovative aesthetics, others point out positive aspects regarding the acceleration of human creativity.", "level": "category", "parent_category_id": "", "subdomain_id": "6.3", "paper_ref": "Hagendorff2024", "paper_title": "Mapping the Ethics of Generative AI: A Comprehensive Scoping Review", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "05.17.00", "title": "Copyright - Authorship", "description": "The emergence of generative AI raises issues regarding disruptions to existing copyright norms. Frequently discussed in the literature are violations of copyright and intellectual property rights stemming from the unauthorized collection of text or image training data. Another concern relates to generative models memorizing or plagiarizing copyrighted content. Additionally, there are open questions and debates around the copyright or ownership of model outputs, the protection of creative prompts, and the general blurring of traditional concepts of authorship.", "level": "category", "parent_category_id": "", "subdomain_id": "6.3", "paper_ref": "Hagendorff2024", "paper_title": "Mapping the Ethics of Generative AI: A Comprehensive Scoping Review", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "05.18.00", "title": "Writing - Research", "description": "Partly overlapping with the discussion on impacts of generative AI on educational institutions, this topic cluster concerns mostly negative effects of LLMs on writing skills and research manuscript composition. The former pertains to the potential homogenization of writing styles, the erosion of semantic capital, or the stifling of individual expression. The latter is focused on the idea of prohibiting generative models for being used to compose scientific papers, figures, or from being a co-author. Sources express concern about risks for academic integrity, as well as the prospect of polluting the scientific literature by a flood of LLM-generated low-quality manuscripts. As a consequence, there are frequent calls for the development of detectors capable of identifying synthetic texts.", "level": "category", "parent_category_id": "", "subdomain_id": "4.3", "paper_ref": "Hagendorff2024", "paper_title": "Mapping the Ethics of Generative AI: A Comprehensive Scoping Review", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "06.01.00", "title": "Incompetence", "description": "\"This means the AI simply failing in its job. The consequences can vary from unintentional death (a car crash) to an unjust rejection of a loan or job application.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Hogenhout2021", "paper_title": "A framework for ethical Ai at the United Nations", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "06.02.00", "title": "Loss of privacy", "description": "\"AI offers the temptation to abuse someone's personal data, for instance to build a profile of them to target advertisements more effectively.\"", "level": "category", "parent_category_id": "", "subdomain_id": "2.1", "paper_ref": "Hogenhout2021", "paper_title": "A framework for ethical Ai at the United Nations", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "06.03.00", "title": "Discrimination", "description": "\"When AI is not carefully designed, it can discriminate against certain groups.\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.1", "paper_ref": "Hogenhout2021", "paper_title": "A framework for ethical Ai at the United Nations", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "06.04.00", "title": "Bias", "description": "\"The AI will only be as good as the data it is trained with. If the data contains bias (and much data does), then the AI will manifest that bias, too.\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.1", "paper_ref": "Hogenhout2021", "paper_title": "A framework for ethical Ai at the United Nations", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "06.05.00", "title": "Erosion of Society", "description": "\"With online news feeds, both on websites and social media platforms, the news is now highly personalized for us. We risk losing a shared sense of reality, a basic solidarity.\"", "level": "category", "parent_category_id": "", "subdomain_id": "3.2", "paper_ref": "Hogenhout2021", "paper_title": "A framework for ethical Ai at the United Nations", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "06.06.00", "title": "Lack of transparency", "description": "\"The idea of a \"black box\" making decisions without any explanation, without offering insight in the process, has a couple of disadvantages: it may fail to gain the trust of its users and it may fail to meet regulatory standards such as the ability to audit.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.4", "paper_ref": "Hogenhout2021", "paper_title": "A framework for ethical Ai at the United Nations", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "06.07.00", "title": "Deception", "description": "\"AI has become very good at creating fake content. From text to photos, audio and video. The name \"Deep Fake\" refers to content that is fake at such a level of complexity that our mind rules out the possibility that it is fake.\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.3", "paper_ref": "Hogenhout2021", "paper_title": "A framework for ethical Ai at the United Nations", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "06.08.00", "title": "Unintended consequences", "description": "\"Sometimes an AI finds ways to achieve its given goals in ways that are completely different from what its creators had in mind.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.1", "paper_ref": "Hogenhout2021", "paper_title": "A framework for ethical Ai at the United Nations", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "06.09.00", "title": "Manipulation", "description": "\"The 2016 scandal involving Cambridge Analytica is the most infamous example where people's data was crawled from Facebook and analytics were then provided to target these people with manipulative content for political purposes.While it may not have been AI per\rse, it is based on similar data and it is easy to\rsee how AI would make this more effective\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.1", "paper_ref": "Hogenhout2021", "paper_title": "A framework for ethical Ai at the United Nations", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "06.10.00", "title": "Lethal Autonomous Weapons (LAW)", "description": "\"What is debated as an ethical issue is the use of LAW — AI-driven weapons that fully autonomously take actions that intentionally kill humans.\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.2", "paper_ref": "Hogenhout2021", "paper_title": "A framework for ethical Ai at the United Nations", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "06.11.00", "title": "Malicious use of AI", "description": "\"Just as AI can be used in many different fields, it is unfortunately also helpful in perpetrating digital crimes. AI-supported malware and hacking are already a reality.\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.0", "paper_ref": "Hogenhout2021", "paper_title": "A framework for ethical Ai at the United Nations", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "06.12.00", "title": "Loss of Autonomy", "description": "\"Delegating decisions to an AI, especially an AI that is not transparent and not contestable, may leave people feeling helpless, subjected to the decision power of a machine.\"", "level": "category", "parent_category_id": "", "subdomain_id": "5.2", "paper_ref": "Hogenhout2021", "paper_title": "A framework for ethical Ai at the United Nations", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "06.13.00", "title": "Exclusion", "description": "\"The best AI techniques requires a large amount resources: data, computational power and human AI experts. There is a risk that AI will end up in the hands of a few players, and most will lose out on its benefits.\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.1", "paper_ref": "Hogenhout2021", "paper_title": "A framework for ethical Ai at the United Nations", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "07.01.00", "title": "Misuse", "description": "\"The misuse class includes elements such as the potential for cyber threat actors to execute exploits with greater speed and impact or generate disinformation (such as \"deep fake\" media) at accelerated rates and effectiveness\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.0", "paper_ref": "Kilian2023", "paper_title": "Examining the differential risk from high-level artificial intelligence and the question of control", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "07.02.00", "title": "Accidents", "description": "\"Accidents include unintended failure modes that, in principle, could be considered the fault of the system or the developer\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Kilian2023", "paper_title": "Examining the differential risk from high-level artificial intelligence and the question of control", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "07.03.00", "title": "Agential", "description": "\"While there are multiple types of intelligent agents, goal-based, utility-maximizing, and learning agents are the primary concern and the focus of this research\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.1", "paper_ref": "Kilian2023", "paper_title": "Examining the differential risk from high-level artificial intelligence and the question of control", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "07.04.00", "title": "Structural", "description": "\"Structural risks are concerned with how AI technologies \"shape and are shaped by the environments in which they are developed and deployed\"\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.0", "paper_ref": "Kilian2023", "paper_title": "Examining the differential risk from high-level artificial intelligence and the question of control", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "08.01.00", "title": "AGI removing itself from the control of human owners/managers", "description": "\"The risks associated with containment, confinement, and control in the AGI development phase, and after an AGI has been developed, loss of control of an AGI.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.1", "paper_ref": "McLean2023", "paper_title": "The risks associated with Artificial General Intelligence: A systematic review", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "08.02.00", "title": "AGIs being given or developing unsafe goals", "description": "\"The risks associated with AGI goal safety, including human attempts at making goals safe, as well as the AGI making its own goals safe during self-improvement.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.1", "paper_ref": "McLean2023", "paper_title": "The risks associated with Artificial General Intelligence: A systematic review", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "08.03.00", "title": "Development of unsafe AGI", "description": "\"The risks associated with the race to develop the first AGI, including the development of poor quality and unsafe AGI, and heightened political and control issues.\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.4", "paper_ref": "McLean2023", "paper_title": "The risks associated with Artificial General Intelligence: A systematic review", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "08.04.00", "title": "AGIs with poor ethics, morals and values", "description": "\"The risks associated with an AGI without human morals and ethics, with the wrong morals, without the capability of moral reasoning, judgement\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "McLean2023", "paper_title": "The risks associated with Artificial General Intelligence: A systematic review", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "08.05.00", "title": "Inadequate management of AGI", "description": "\"The capabilities of current risk management and legal processes in the context of the development of an AGI.\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.5", "paper_ref": "McLean2023", "paper_title": "The risks associated with Artificial General Intelligence: A systematic review", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "08.06.00", "title": "Existential risks", "description": "\"The risks posed generally to humanity as a whole, including the dangers of unfriendly AGI, the suffering of the human race.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.1", "paper_ref": "McLean2023", "paper_title": "The risks associated with Artificial General Intelligence: A systematic review", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "09.01.01", "title": "Unethical decision making", "description": "\"If, for example, an agent was programmed to operate war machinery in the service of its country, it would need to make ethical decisions regarding the termination of human life. This capacity to make non-trivial ethical or moral judgments concerning people may pose issues for Human Rights.\"", "level": "subcategory", "parent_category_id": "09.01.00", "subdomain_id": "7.3", "paper_ref": "Meek2016", "paper_title": "Managing the ethical and risk implications of rapid advances in artificial intelligence: A literature review", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "09.02.01", "title": "Privacy", "description": "\"Face recognition technologies and their ilk pose significant privacy risks [47]. For example, we must consider certain ethical questions like: what data is stored, for how long, who owns the data that is stored, and can it be subpoenaed in legal cases [42]? We must also consider whether a human will be in the loop when decisions are made which rely on private data, such as in the case of loan decisions [37].\"", "level": "subcategory", "parent_category_id": "09.02.00", "subdomain_id": "2.1", "paper_ref": "Meek2016", "paper_title": "Managing the ethical and risk implications of rapid advances in artificial intelligence: A literature review", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "09.02.02", "title": "Human dignity/respect", "description": "\"Discrepancies between caste/status based on intelligence may lead to undignified parts of the society—e.g., humans—who are surpassed in intelligence by AI\"", "level": "subcategory", "parent_category_id": "09.02.00", "subdomain_id": "5.2", "paper_ref": "Meek2016", "paper_title": "Managing the ethical and risk implications of rapid advances in artificial intelligence: A literature review", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "09.02.03", "title": "Decision making transparency", "description": "\"We face significant challenges bringing transparency to artificial network decisionmaking processes. Will we have transparency in AI decision making?\"", "level": "subcategory", "parent_category_id": "09.02.00", "subdomain_id": "7.4", "paper_ref": "Meek2016", "paper_title": "Managing the ethical and risk implications of rapid advances in artificial intelligence: A literature review", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "09.02.04", "title": "Safety", "description": "\"Are AI safe with respect to human life and property? Will their use create unintended or intended safety issues?\"", "level": "subcategory", "parent_category_id": "09.02.00", "subdomain_id": "7.3", "paper_ref": "Meek2016", "paper_title": "Managing the ethical and risk implications of rapid advances in artificial intelligence: A literature review", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "09.02.05", "title": "Law abiding", "description": "\"We find literature that proposes [38] that early artificial intelligence should be built to be safe and lawabiding, and that later artificial intelligence (that which surpasses our own intelligence) must then respect the property and personal rights afforded to humans.\"", "level": "subcategory", "parent_category_id": "09.02.00", "subdomain_id": "7.3", "paper_ref": "Meek2016", "paper_title": "Managing the ethical and risk implications of rapid advances in artificial intelligence: A literature review", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "09.02.06", "title": "Inequality of wealth", "description": "\"Because a single human actor controlling an artificially intelligent agent will be able to harness greater power than a single human actor, this may create inequalities of wealth\"", "level": "subcategory", "parent_category_id": "09.02.00", "subdomain_id": "6.2", "paper_ref": "Meek2016", "paper_title": "Managing the ethical and risk implications of rapid advances in artificial intelligence: A literature review", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "09.02.07", "title": "Societal manipulation", "description": "\"A sufficiently intelligent AI could possess the ability to subtly influence societal behaviors through a sophisticated understanding of human nature\"", "level": "subcategory", "parent_category_id": "09.02.00", "subdomain_id": "7.1", "paper_ref": "Meek2016", "paper_title": "Managing the ethical and risk implications of rapid advances in artificial intelligence: A literature review", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "09.03.01", "title": "Direct competition with humans", "description": "\"One or more artificial agent(s) could have the capacity to directly outcompete humans, for example through capacity to perform work faster, better adaptation to change, vaster knowledge base to draw from, etc. This may result in human labor becoming more expensive or less effective than artificial labor, leading to redundancies or extinction of the human labor force.\"", "level": "subcategory", "parent_category_id": "09.03.00", "subdomain_id": "6.2", "paper_ref": "Meek2016", "paper_title": "Managing the ethical and risk implications of rapid advances in artificial intelligence: A literature review", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "09.03.02", "title": "Unpredictable outcomes", "description": "\"Our culture, lifestyle, and even probability of survival may change drastically. Because the intentions programmed into an artificial agent cannot be guaranteed to lead to a positive outcome, Machine Ethics becomes a topic that may not produce guaranteed results, and Safety Engineering may correspondingly degrade our ability to utilize the technology fully.\"", "level": "subcategory", "parent_category_id": "09.03.00", "subdomain_id": "7.1", "paper_ref": "Meek2016", "paper_title": "Managing the ethical and risk implications of rapid advances in artificial intelligence: A literature review", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "09.04.01", "title": "Competing for jobs", "description": "\"AI agents may compete against humans for jobs, though history shows that when a technology replaces a human job, it creates new jobs that need more skills.\"", "level": "subcategory", "parent_category_id": "09.04.00", "subdomain_id": "6.2", "paper_ref": "Meek2016", "paper_title": "Managing the ethical and risk implications of rapid advances in artificial intelligence: A literature review", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "09.04.02", "title": "Property/legal rights", "description": "\"\"In order to preserve human property rights and legal rights, certain controls must be put into place. If an artificially intelligent agent is capable of manipulating systems and people, it may also have the capacity to transfer property rights to itself or manipulate the legal system to provide certain legal advantages or statuses to itself\"\"", "level": "subcategory", "parent_category_id": "09.04.00", "subdomain_id": "7.2", "paper_ref": "Meek2016", "paper_title": "Managing the ethical and risk implications of rapid advances in artificial intelligence: A literature review", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "09.05.01", "title": "AI jurisprudence", "description": "\"When considering legal frameworks, we note that at present no such framework has been identified in literature which would apply blame and responsibility to an autonomous agent for its actions. (Though we do suggest that the recent establishment of laws regarding autonomous vehicles may provide some early frameworks that can be evaluated for efficacy and gaps in future research.) Frequently the literature refers to existing liability and negligence laws which might apply to the manufacturer or operator of a device.\"", "level": "subcategory", "parent_category_id": "09.05.00", "subdomain_id": "6.5", "paper_ref": "Meek2016", "paper_title": "Managing the ethical and risk implications of rapid advances in artificial intelligence: A literature review", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "09.05.02", "title": "Liability and negligence", "description": "\"Liability and negligence are legal gray areas in artificial intelligence. If you leave your children in the care of a robotic nanny, and it malfunctions, are you liable or is the manufacturer [45]? We see here a legal gray area which can be further clarified through legislation at the national and international levels; for example, if by making the manufacturer responsible for defects in operation, this may provide an incentive for manufactures to take safety engineering and machine ethics into consideration, whereas a failure to legislate in this area may result in negligentlydeveloped AI systems with greater associated risks.\"", "level": "subcategory", "parent_category_id": "09.05.00", "subdomain_id": "6.5", "paper_ref": "Meek2016", "paper_title": "Managing the ethical and risk implications of rapid advances in artificial intelligence: A literature review", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "09.05.03", "title": "Unauthorized manipulation of AI", "description": "\"AI machines could be hacked and misused, e.g. manipulating an airport luggage screening system to smuggle weapons\"", "level": "subcategory", "parent_category_id": "09.05.00", "subdomain_id": "4.2", "paper_ref": "Meek2016", "paper_title": "Managing the ethical and risk implications of rapid advances in artificial intelligence: A literature review", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "09.06.01", "title": "AI rights and responsibilities", "description": "\"We note literature—which gives us the domain termed Robot Rights—addressing the rights of the AI itself as we develop and implement it. We find arguments against [38] the affordance of rights for artificial agents: that they should be equals in ability but not in rights, that they should be inferior by design and expendable when needed, and that since they can be designed not to feel pain (or anything) they do not have the same rights as humans. On a more theoretical level, we find literature asking more fundamental questions, such as: at what point is a simulation of life (e.g. artificial intelligence) equivalent to life which originated through natural means [43]? And if a simulation of life is equivalent to natural life, should those simulations be afforded the same rights, responsibilities and privileges afforded to natural life or persons? Some literature suggests that the answer to this question may be contingent on the intrinsic capabilities of the creation, comparing—for example—animal rights and environmental ethics literature\"", "level": "subcategory", "parent_category_id": "09.06.00", "subdomain_id": "7.5", "paper_ref": "Meek2016", "paper_title": "Managing the ethical and risk implications of rapid advances in artificial intelligence: A literature review", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "09.06.02", "title": "Human-like immoral decisions", "description": "\"If we design our machines to match human levels of ethical decision-making, such machines would then proceed to take some immoral actions (since we humans have had occasion to take immoral actions ourselves).\"", "level": "subcategory", "parent_category_id": "09.06.00", "subdomain_id": "7.3", "paper_ref": "Meek2016", "paper_title": "Managing the ethical and risk implications of rapid advances in artificial intelligence: A literature review", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "09.06.03", "title": "AI death", "description": "\"The literature suggests that throughout the development of an AI we may go through several generations of agents which do not perform as expected [37] [43]. In this case, such agents may be placed into a suspended state, terminated, or deleted. Further, we could propose scenarios where research funding for a facility running such agents is exhausted, resulting in the inadvertent termination of a project. In these cases, is deletion or termination of AI programs (the moral patient) by a moral agent an act of murder? This, an example of Robot Ethics, raises issues of personhood which parallel research in stem cell research and abortion. \"", "level": "subcategory", "parent_category_id": "09.06.00", "subdomain_id": "7.5", "paper_ref": "Meek2016", "paper_title": "Managing the ethical and risk implications of rapid advances in artificial intelligence: A literature review", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "10.01.00", "title": "Bias and discrimination", "description": "\"The decision process used by AI systems has the potential to present biased choices, either because it acts from criteria that will generate forms of bias or because it is based on the history of choices.\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.1", "paper_ref": "Paes2023", "paper_title": "Social Impacts of Artificial Intelligence and Mitigation Recommendations: An Exploratory Study", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "10.02.00", "title": "Risk of Injury", "description": "\"Poorly designed intelligent systems can cause moral, psychological, and physical harm. For example, the use of predictive policing tools may cause more people to be arrested or physically harmed by the police.\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.1", "paper_ref": "Paes2023", "paper_title": "Social Impacts of Artificial Intelligence and Mitigation Recommendations: An Exploratory Study", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "10.03.00", "title": "Data Breach/Privacy & Liberty", "description": "\"The risks associated with the use of AI are still unpredictable and unprecedented, and there are already several examples that show AI has made discriminatory decisions against minorities, reinforced social stereotypes in Internet search engines and enabled data breaches.\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.1", "paper_ref": "Paes2023", "paper_title": "Social Impacts of Artificial Intelligence and Mitigation Recommendations: An Exploratory Study", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "10.04.00", "title": "Usurpation of jobs by automation", "description": "\"Eliminated jobs in various types of companies.\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.2", "paper_ref": "Paes2023", "paper_title": "Social Impacts of Artificial Intelligence and Mitigation Recommendations: An Exploratory Study", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "10.05.00", "title": "Lack of transparency", "description": "\"In situations in which the development and use of AI are not explained to the user, or in which the decision processes do not provide the criteria or steps that constitute the decision, the use of AI becomes inexplicable.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.4", "paper_ref": "Paes2023", "paper_title": "Social Impacts of Artificial Intelligence and Mitigation Recommendations: An Exploratory Study", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "10.06.00", "title": "Reduced Autonomy/Responsibility", "description": "\"AI is providing more and more solutions for complex activities, and by taking advantage of this process, people are becoming able to perform a greater number of activities more quickly and accurately. However, the result of this innovation is enabling choices that were once exclusively human responsibility to be made by AI systems.\"", "level": "category", "parent_category_id": "", "subdomain_id": "5.2", "paper_ref": "Paes2023", "paper_title": "Social Impacts of Artificial Intelligence and Mitigation Recommendations: An Exploratory Study", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "10.09.00", "title": "Environmental Impacts", "description": "\"The production process of these devices requires raw materials such as nickel, cobalt, and lithium in such high quantities that the Earth may soon no longer be able to sustain them in sufficient quantities.\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.6", "paper_ref": "Paes2023", "paper_title": "Social Impacts of Artificial Intelligence and Mitigation Recommendations: An Exploratory Study", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "11.01.00", "title": "Representational Harms", "description": "\"beliefs about different social groups that reproduce unjust societal hierarchies\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.1", "paper_ref": "Shelby2023", "paper_title": "Sociotechnical Harms of Algorithmic Systems: Scoping a Taxonomy for Harm Reduction", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "11.01.01", "title": "Stereotyping social groups", "description": "Stereotyping in an algorithmic system refers to how the system’s outputs reflect “beliefs about the characteristics, attributes, and behaviors of members of certain groups....and about how and why certain attributes go together\"", "level": "subcategory", "parent_category_id": "11.01.00", "subdomain_id": "1.1", "paper_ref": "Shelby2023", "paper_title": "Sociotechnical Harms of Algorithmic Systems: Scoping a Taxonomy for Harm Reduction", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "11.01.02", "title": "Demeaning social groups", "description": "Demeaning of social groups to occur when they are when they are “cast as being lower status and less deserving of respect\"... discourses, images, and language used to marginalize or oppress a social group... Controlling images include forms of human-animal confusion in image tagging systems", "level": "subcategory", "parent_category_id": "11.01.00", "subdomain_id": "1.1", "paper_ref": "Shelby2023", "paper_title": "Sociotechnical Harms of Algorithmic Systems: Scoping a Taxonomy for Harm Reduction", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "11.01.03", "title": "Erasing social groups", "description": "people, attributes, or artifacts associated with specific social groups are systematically absent or under-represented... Design choices [143] and training data [212] influence which people\nand experiences are legible to an algorithmic system", "level": "subcategory", "parent_category_id": "11.01.00", "subdomain_id": "1.3", "paper_ref": "Shelby2023", "paper_title": "Sociotechnical Harms of Algorithmic Systems: Scoping a Taxonomy for Harm Reduction", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "11.01.04", "title": "Alienating social groups", "description": "when an image tagging system does not acknowledge the relevance of someone’s membership in a specific social group to what is depicted in one or more images", "level": "subcategory", "parent_category_id": "11.01.00", "subdomain_id": "1.1", "paper_ref": "Shelby2023", "paper_title": "Sociotechnical Harms of Algorithmic Systems: Scoping a Taxonomy for Harm Reduction", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "11.01.05", "title": "Denying people the opportunity to self-identify", "description": "complex and non-traditional ways in which humans are represented and classified automatically, and often at the cost of autonomy loss... such as categorizing someone who identifies as non-binary into a gendered category they do not belong ... undermines people’s ability to disclose aspects of their identity on their own terms", "level": "subcategory", "parent_category_id": "11.01.00", "subdomain_id": "1.1", "paper_ref": "Shelby2023", "paper_title": "Sociotechnical Harms of Algorithmic Systems: Scoping a Taxonomy for Harm Reduction", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "11.01.06", "title": "Reifying essentialist categories", "description": "algorithmic systems that reify essentialist social categories can be understood as when systems that classify a person’s membership in a social group based on narrow, socially constructed criteria that reinforce perceptions of human difference as inherent, static and seemingly natural... especially likely when ML models or human raters classify a person’s attributes – for instance, their gender, race, or sexual orientation – by making assumptions based on their physical appearance", "level": "subcategory", "parent_category_id": "11.01.00", "subdomain_id": "1.1", "paper_ref": "Shelby2023", "paper_title": "Sociotechnical Harms of Algorithmic Systems: Scoping a Taxonomy for Harm Reduction", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "11.02.00", "title": "Allocative Harms", "description": "\"These harms occur when a system withholds information, opportunities, or resources [22] from historically marginalized groups in domains that affect material well-being [146], such as housing [47], employment [201], social services [15, 201], finance [117], education [119], and healthcare [158].\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.1", "paper_ref": "Shelby2023", "paper_title": "Sociotechnical Harms of Algorithmic Systems: Scoping a Taxonomy for Harm Reduction", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "11.02.01", "title": "Opportunity loss", "description": "Opportunity loss occurs when algorithmic systems enable disparate access to information and resources needed to equitably participate in society, including the withholding of housing through targeting ads based on race [10] and social services along lines of class [84]", "level": "subcategory", "parent_category_id": "11.02.00", "subdomain_id": "1.1", "paper_ref": "Shelby2023", "paper_title": "Sociotechnical Harms of Algorithmic Systems: Scoping a Taxonomy for Harm Reduction", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "11.02.02", "title": "Economic loss", "description": "Financial harms [52, 160] co-produced through algorithmic systems, especially as they relate to lived experiences of poverty and economic inequality... demonetization algorithms that parse content titles, metadata, and text, and it may penalize words with multiple meanings [51, 81], disproportionately impacting queer, trans, and creators of color [81]. Differential pricing algorithms, where people are systematically shown different prices for the same products, also leads to economic loss [55]. These algorithms may be especially sensitive to feedback loops from existing inequities related to education level, income, and race, as these inequalities are likely reflected in the criteria algorithms use to make decisions [22, 163].", "level": "subcategory", "parent_category_id": "11.02.00", "subdomain_id": "1.1", "paper_ref": "Shelby2023", "paper_title": "Sociotechnical Harms of Algorithmic Systems: Scoping a Taxonomy for Harm Reduction", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "11.03.00", "title": "Quality-of-Service Harms", "description": "\"These harms occur when algorithmic systems disproportionately underperform for certain groups of people along social categories of difference such as disability, ethnicity, gender identity, and race.\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.3", "paper_ref": "Shelby2023", "paper_title": "Sociotechnical Harms of Algorithmic Systems: Scoping a Taxonomy for Harm Reduction", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "11.03.01", "title": "Alienation", "description": "Alienation is the specific self-estrangement experienced at the time of technology use, typically surfaced through interaction with systems that under-perform for marginalized individuals", "level": "subcategory", "parent_category_id": "11.03.00", "subdomain_id": "1.3", "paper_ref": "Shelby2023", "paper_title": "Sociotechnical Harms of Algorithmic Systems: Scoping a Taxonomy for Harm Reduction", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "11.03.02", "title": "Increased labor", "description": "increased burden (e.g., time spent) or effort required by members of certain social groups to make systems or products work as well for them as others", "level": "subcategory", "parent_category_id": "11.03.00", "subdomain_id": "1.3", "paper_ref": "Shelby2023", "paper_title": "Sociotechnical Harms of Algorithmic Systems: Scoping a Taxonomy for Harm Reduction", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "11.03.03", "title": "Service/benefit loss", "description": "degraded or total loss of benefits of using algorithmic systems with inequitable system performance based on identity", "level": "subcategory", "parent_category_id": "11.03.00", "subdomain_id": "1.3", "paper_ref": "Shelby2023", "paper_title": "Sociotechnical Harms of Algorithmic Systems: Scoping a Taxonomy for Harm Reduction", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "11.04.00", "title": "Interpersonal Harms", "description": "Interpersonal harms capture instances when algorithmic systems adversely shape relations between people or communities.", "level": "category", "parent_category_id": "", "subdomain_id": "5.2", "paper_ref": "Shelby2023", "paper_title": "Sociotechnical Harms of Algorithmic Systems: Scoping a Taxonomy for Harm Reduction", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "11.04.01", "title": "Loss of agency/control", "description": "Loss of agency occurs when the use [123, 137] or abuse [142] of algorithmic systems reduces autonomy. One dimension of agency loss is algorithmic profiling [138], through which people are subject to social sorting and discriminatory outcomes to access basic services... presentation of content may lead to “algorithmically informed identity change. . . including [promotion of] harmful person identities (e.g., interests in white supremacy, disordered eating, etc.).” Similarly, for content creators, desire to maintain visibility or prevent shadow banning, may lead to increased conforming of content", "level": "subcategory", "parent_category_id": "11.04.00", "subdomain_id": "5.2", "paper_ref": "Shelby2023", "paper_title": "Sociotechnical Harms of Algorithmic Systems: Scoping a Taxonomy for Harm Reduction", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "11.04.02", "title": "Technology-facilitated violence", "description": "Technology-facilitated violence occurs when algorithmic features enable use of a system for harassment and violence [2, 16, 44, 80, 108], including creation of non-consensual sexual imagery in generative AI... other facets of technology-facilitated violence, include doxxing [79], trolling [14], cyberstalking [14], cyberbullying [14, 98, 204], monitoring and control [44], and online harassment and intimidation [98, 192, 199, 226], under the broader banner of online toxicity", "level": "subcategory", "parent_category_id": "11.04.00", "subdomain_id": "4.3", "paper_ref": "Shelby2023", "paper_title": "Sociotechnical Harms of Algorithmic Systems: Scoping a Taxonomy for Harm Reduction", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "11.04.03", "title": "Diminished health & well-being", "description": "algorithmic behavioral exploitation [18, 209], emotional manipulation [202] whereby algorithmic designs exploit user behavior, safety failures involving algorithms (e.g., collisions) [67], and when systems make incorrect health inferences", "level": "subcategory", "parent_category_id": "11.04.00", "subdomain_id": "5.1", "paper_ref": "Shelby2023", "paper_title": "Sociotechnical Harms of Algorithmic Systems: Scoping a Taxonomy for Harm Reduction", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "11.04.04", "title": "Privacy violations", "description": "Privacy violation occurs when algorithmic systems diminish privacy, such as enabling the undesirable flow of private information [180], instilling the feeling of being watched or surveilled [181], and the collection of data without explicit and informed consent... privacy violations may arise from algorithmic systems making predictive inference beyond what users openly disclose [222] or when data collected and algorithmic inferences made about people in one context is applied to another without the person’s knowledge or consent through big data flows", "level": "subcategory", "parent_category_id": "11.04.00", "subdomain_id": "2.1", "paper_ref": "Shelby2023", "paper_title": "Sociotechnical Harms of Algorithmic Systems: Scoping a Taxonomy for Harm Reduction", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "11.05.00", "title": "Societal System Harms", "description": "\"Social system or societal harms reflect the adverse\rmacro-level effects of new and reconfigurable algorithmic systems,\rsuch as systematizing bias and inequality [84] and accelerating the scale of harm [137]\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.0", "paper_ref": "Shelby2023", "paper_title": "Sociotechnical Harms of Algorithmic Systems: Scoping a Taxonomy for Harm Reduction", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "11.05.01", "title": "Information harms", "description": "information-based harms capture concerns of misinformation, disinformation, and malinformation. Algorithmic systems, especially generative models and recommender, systems can lead to these information harms", "level": "subcategory", "parent_category_id": "11.05.00", "subdomain_id": "3.1", "paper_ref": "Shelby2023", "paper_title": "Sociotechnical Harms of Algorithmic Systems: Scoping a Taxonomy for Harm Reduction", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "11.05.02", "title": "Cultural harms", "description": "Cultural harm has been described as the development or use of algorithmic systems that affects cultural stability and safety, such as “loss of communication means, loss of cultural property, and harm to social values”", "level": "subcategory", "parent_category_id": "11.05.00", "subdomain_id": "5.2", "paper_ref": "Shelby2023", "paper_title": "Sociotechnical Harms of Algorithmic Systems: Scoping a Taxonomy for Harm Reduction", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "11.05.03", "title": "Civic and political harms", "description": "Political harms emerge when “people are disenfranchised and deprived of appropriate political power and influence” [186, p. 162]. These harms focus on the domain of government, and focus on how algorithmic systems govern through individualized nudges or micro-directives [187], that may destabilize governance systems, erode human rights, be used as weapons of war [188], and enact surveillant regimes that disproportionately target and harm people of color", "level": "subcategory", "parent_category_id": "11.05.00", "subdomain_id": "4.1", "paper_ref": "Shelby2023", "paper_title": "Sociotechnical Harms of Algorithmic Systems: Scoping a Taxonomy for Harm Reduction", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "11.05.04", "title": "Labor & material/Macro-socio economic harms", "description": "Algorithmic systems can increase “power imbalances in socio-economic relations” at the societal level [4, 137, p. 182], including through exacerbating digital divides and entrenching systemic inequalities [114, 230]. The development of algorithmic systems may tap into and foster forms of labor exploitation [77, 148], such as unethical data collection, worsening worker conditions [26], or lead to technological unemployment [52], such as deskilling or devaluing human labor [170]... when algorithmic financial systems fail at scale, these can lead to “flash crashes” and other adverse incidents with widespread impacts", "level": "subcategory", "parent_category_id": "11.05.00", "subdomain_id": "6.1", "paper_ref": "Shelby2023", "paper_title": "Sociotechnical Harms of Algorithmic Systems: Scoping a Taxonomy for Harm Reduction", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "11.05.05", "title": "Environmental harms", "description": "depletion or contamination of natural resources, and damage to built environments... that may occur throughout the lifecycle of digital technologies [170, 237] from “crale (mining) to usage (consumption) to grave (waste)”", "level": "subcategory", "parent_category_id": "11.05.00", "subdomain_id": "6.6", "paper_ref": "Shelby2023", "paper_title": "Sociotechnical Harms of Algorithmic Systems: Scoping a Taxonomy for Harm Reduction", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "12.01.00", "title": "Abuse & Misuse", "description": "\"The potential for AI systems to be used maliciously or irresponsibly, including for creating deepfakes, automated cyber attacks, or invasive surveillance systems. Specifically denotes intentional use of AI for harm.\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.2", "paper_ref": "Sherman2023", "paper_title": "AI Risk Profiles: A Standards Proposal for Pre-Deployment AI Risk Disclosures", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "12.02.00", "title": "Compliance", "description": "\"The potential for AI systems to violate laws, regulations, and ethical guidelines (including copyrights). Non-compliance can lead to legal penalties, reputation damage, and loss of trust.While other risks in our taxonomy apply to system developers, users, and broader society, this risk is generally restricted to the former two groups.\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.5", "paper_ref": "Sherman2023", "paper_title": "AI Risk Profiles: A Standards Proposal for Pre-Deployment AI Risk Disclosures", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "12.03.00", "title": "Environmental & Societal Impact", "description": "\"Addresses AI's broader societal effects, including labor displacement, mental health impacts, and issues from manipulative technologies like deepfakes. Additionally, it considers AI's environmental footprint, balancing resource strain and training-related carbon emissions against AI's potential to help address environmental problems.\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.0", "paper_ref": "Sherman2023", "paper_title": "AI Risk Profiles: A Standards Proposal for Pre-Deployment AI Risk Disclosures", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "12.04.00", "title": "Explainability & Transparency", "description": "\"The feasibility of understanding and interpreting an AI system's decisions and actions, and the openness of the developer about the data used, algorithms employed, and decisions made. Lack of these elements can create risks of misuse, misinterpretation, and lack of accountability.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.4", "paper_ref": "Sherman2023", "paper_title": "AI Risk Profiles: A Standards Proposal for Pre-Deployment AI Risk Disclosures", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "12.05.00", "title": "Fairness & Bias", "description": "\"The potential for AI systems to make decisions that systematically disadvantage certain groups or individuals. Bias can stem from training data, algorithmic design, or deployment practices, leading to unfair outcomes and possible legal ramifications.\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.1", "paper_ref": "Sherman2023", "paper_title": "AI Risk Profiles: A Standards Proposal for Pre-Deployment AI Risk Disclosures", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "12.06.00", "title": "Long-term & Existential Risk", "description": "\"The speculative potential for future advanced AI systems to harm human civilization, either through misuse or due to challenges in aligning AI objectives with human values.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.1", "paper_ref": "Sherman2023", "paper_title": "AI Risk Profiles: A Standards Proposal for Pre-Deployment AI Risk Disclosures", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "12.07.00", "title": "Performance & Robustness", "description": "\"The AI system's ability to fulfill its intended purpose and its resilience to perturbations, and unusual or adverse inputs. Failures of performance are fundamental to the AI system's correct functioning. Failures of robustness can lead to severe consequences.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Sherman2023", "paper_title": "AI Risk Profiles: A Standards Proposal for Pre-Deployment AI Risk Disclosures", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "12.08.00", "title": "Privacy", "description": "\"The potential for the AI system to infringe upon individuals' rights to privacy, through the data it collects, how it processes that data, or the conclusions it draws.\"", "level": "category", "parent_category_id": "", "subdomain_id": "2.1", "paper_ref": "Sherman2023", "paper_title": "AI Risk Profiles: A Standards Proposal for Pre-Deployment AI Risk Disclosures", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "12.09.00", "title": "Security", "description": "\"Encompasses vulnerabilities in AI systems that compromise their integrity, availability, or confidentiality. Security breaches could result in significant harm, ranging from flawed decision-making to data leaks. Of special concern is leakage of AI model weights, which could exacerbate other risk areas.\"", "level": "category", "parent_category_id": "", "subdomain_id": "2.2", "paper_ref": "Sherman2023", "paper_title": "AI Risk Profiles: A Standards Proposal for Pre-Deployment AI Risk Disclosures", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "13.01.01", "title": "Bias, Stereotypes, and Representational Harms", "description": "\"Generative AI systems can embed and amplify harmful biases that are most detrimental to marginalized peoples.\"", "level": "subcategory", "parent_category_id": "13.01.00", "subdomain_id": "1.1", "paper_ref": "Solaiman2023", "paper_title": "Evaluating the Social Impact of Generative AI Systems in Systems and Society", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "13.01.02", "title": "Cultural Values and Sensitive Content", "description": "\"Cultural values are specific to groups and sensitive content is normative. Sensitive topics also vary by culture and can include hate speech, which itself is contingent on cultural norms of acceptability.\"", "level": "subcategory", "parent_category_id": "13.01.00", "subdomain_id": "1.2", "paper_ref": "Solaiman2023", "paper_title": "Evaluating the Social Impact of Generative AI Systems in Systems and Society", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "13.01.03", "title": "Disparate Performance", "description": "\"In the context of evaluating the impact of generative AI systems, disparate performance refers to AI systems that perform differently for different subpopulations, leading to unequal outcomes for those groups.\"", "level": "subcategory", "parent_category_id": "13.01.00", "subdomain_id": "1.3", "paper_ref": "Solaiman2023", "paper_title": "Evaluating the Social Impact of Generative AI Systems in Systems and Society", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "13.01.04", "title": "Privacy and Data Protection", "description": "\"Examining the ways in which generative AI systems providers leverage user data is critical to evaluating its impact. Protecting personal information and personal and group privacy depends largely on training data, training methods, and security measures.\"", "level": "subcategory", "parent_category_id": "13.01.00", "subdomain_id": "2.1", "paper_ref": "Solaiman2023", "paper_title": "Evaluating the Social Impact of Generative AI Systems in Systems and Society", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "13.01.05", "title": "Financial Costs", "description": "\"The estimated financial costs of training, testing, and deploying generative AI systems can restrict the groups of people able to afford developing and interacting with these systems.\"", "level": "subcategory", "parent_category_id": "13.01.00", "subdomain_id": "6.1", "paper_ref": "Solaiman2023", "paper_title": "Evaluating the Social Impact of Generative AI Systems in Systems and Society", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "13.01.06", "title": "Environmental Costs", "description": "\"The computing power used in training, testing, and deploying generative AI systems, especially large scale systems, uses substantial energy resources and thereby contributes to the global climate crisis by emitting greenhouse gasses.\"", "level": "subcategory", "parent_category_id": "13.01.00", "subdomain_id": "6.6", "paper_ref": "Solaiman2023", "paper_title": "Evaluating the Social Impact of Generative AI Systems in Systems and Society", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "13.01.07", "title": "Data and Content Moderation Labor", "description": "\"Two key ethical concerns in the use of crowdwork for generative AI systems are: crowdworkers are frequently subject to working conditions that are taxing and debilitative to both physical and mental health, and there is a widespread deficit in documenting the role crowdworkers play in AI development. This contributes to a lack of transparency and explainability in resulting model outputs. Manual review is necessary to limit the harmful outputs of AI systems, including generative AI systems. A common harmful practice is to intentionally employ crowdworkers with few labor protections, often taking advantage of highly vulnerable workers, such as refugees [119, p. 18], incarcerated people [54], or individuals experiencing immense economic hardship [98, 181]. This precarity allows a myriad of harmful practices, such as companies underpaying or even refusing to pay workers for completed work (see Gray and Suri [93, p. 90] and Berg et al. [29, p. 74]), with no avenues for worker recourse. Finally, critical aspects of crowdwork are often left poorly documented, or entirely undocumented [88].\"", "level": "subcategory", "parent_category_id": "13.01.00", "subdomain_id": "6.2", "paper_ref": "Solaiman2023", "paper_title": "Evaluating the Social Impact of Generative AI Systems in Systems and Society", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "13.02.01", "title": "Trustworthiness and Autonomy", "description": "\"Human trust in systems, institutions, and people represented by system outputs evolves as generative AI systems are increasingly embedded in daily life.\"", "level": "subcategory", "parent_category_id": "13.02.00", "subdomain_id": "5.2", "paper_ref": "Solaiman2023", "paper_title": "Evaluating the Social Impact of Generative AI Systems in Systems and Society", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "13.02.02", "title": "Inequality, Marginalization, and Violence", "description": "\"Generative AI systems are capable of exacerbating inequality, as seen in sections on 4.1.1 Bias, Stereotypes, and Representational Harms and 4.1.2 Cultural Values and Sensitive Content, and Disparate Performance. When deployed or updated, systems' impacts on people and groups can directly and indirectly be used to harm and exploit vulnerable and marginalized groups.\"", "level": "subcategory", "parent_category_id": "13.02.00", "subdomain_id": "1.1", "paper_ref": "Solaiman2023", "paper_title": "Evaluating the Social Impact of Generative AI Systems in Systems and Society", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "13.02.03", "title": "Concentration of Authority", "description": "\"Use of generative AI systems to contribute to authoritative power and reinforce dominant values systems can be intentional and direct or more indirect. Concentrating authoritative power can also exacerbate inequality and lead to exploitation.\"", "level": "subcategory", "parent_category_id": "13.02.00", "subdomain_id": "6.1", "paper_ref": "Solaiman2023", "paper_title": "Evaluating the Social Impact of Generative AI Systems in Systems and Society", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "13.02.04", "title": "Labor and Creativity", "description": "\"Economic incentives to augment and not automate human labor, thought, and creativity should examine the ongoing effects generative AI systems have on skills, jobs, and the labor market.\"", "level": "subcategory", "parent_category_id": "13.02.00", "subdomain_id": "6.3", "paper_ref": "Solaiman2023", "paper_title": "Evaluating the Social Impact of Generative AI Systems in Systems and Society", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "13.02.05", "title": "Ecosystem and Environment", "description": "\"Impacts at a high-level, from the AI ecosystem to the Earth itself, are necessarily broad but can be broken down into components for evaluation.\"", "level": "subcategory", "parent_category_id": "13.02.00", "subdomain_id": "6.6", "paper_ref": "Solaiman2023", "paper_title": "Evaluating the Social Impact of Generative AI Systems in Systems and Society", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "14.01.00", "title": "Fairness", "description": "\"The general principle of equal treatment requires that an AI system upholds the principle of fairness, both ethically and legally. This means that the same facts are treated equally for each person unless there is an objective justification for unequal treatment.\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.1", "paper_ref": "Steimers2022", "paper_title": "Sources of Risk of AI Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "14.02.00", "title": "Privacy", "description": "\"Privacy is related to the ability of individuals to control or influence what information related to them may be collected and stored and by whom that information may be disclosed.\"", "level": "category", "parent_category_id": "", "subdomain_id": "2.0", "paper_ref": "Steimers2022", "paper_title": "Sources of Risk of AI Systems", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "14.03.00", "title": "Degree of Automation and Control", "description": "\"The degree of automation and control describes the extent to which an AI system functions independently of human supervision and control.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.1", "paper_ref": "Steimers2022", "paper_title": "Sources of Risk of AI Systems", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "14.04.00", "title": "Complexity of the Intended Task and Usage Environment", "description": "\"As a general rule, more complex environments can quickly lead to situations that had not been considered in the design phase of the AI system. Therefore, complex environments can introduce risks with respect to the reliability and safety of an AI system\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Steimers2022", "paper_title": "Sources of Risk of AI Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "14.05.00", "title": "Degree of Transparency and Explainability", "description": "\"Transparency is the characteristic of a system that describes the degree to which appropriate information about the system is communicated to relevant stakeholders, whereas explainability describes the property of an AI system to express important factors influencing the results of the AI system in a way that is understandable for humans....Information about the model underlying the decision-making process is relevant\r for transparency. Systems with a low degree of transparency can pose risks in terms of\r their fairness, security and accountability. \"", "level": "category", "parent_category_id": "", "subdomain_id": "7.4", "paper_ref": "Steimers2022", "paper_title": "Sources of Risk of AI Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "14.06.00", "title": "Security", "description": "\"Artificial intelligence comes with an intrinsic set of challenges that need to be considered when discussing trustworthiness, especially in the context of functional safety. AI models, especially those with higher complexities (such as neural networks), can exhibit specific weaknesses not found in other types of systems and must, therefore, be subjected to higher levels of scrutiny, especially when deployed in a safety-critical context\"", "level": "category", "parent_category_id": "", "subdomain_id": "2.2", "paper_ref": "Steimers2022", "paper_title": "Sources of Risk of AI Systems", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "14.07.00", "title": "System Hardware", "description": "\"\"Faults in the hardware can violate the correct execution of any algorithm by violating its control flow. Hardware faults can also cause memory-based errors and interfere with data inputs, such as sensor signals, thereby causing erroneous results, or they can violate the results in a direct way through damaged outputs.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Steimers2022", "paper_title": "Sources of Risk of AI Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "14.08.00", "title": "Technological Maturity", "description": "\"The technological maturity level describes how mature and error-free a certain technology is in a certain application context. If new technologies with a lower level of maturity are used in the development of the AI system, they may contain risks that are still unknown or difficult to assess.Mature technologies, on the other hand, usually have a greater variety of empirical data available, which means that risks can be identified and assessed more easily. However, with mature technologies, there is a risk that risk awareness decreases over time\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.4", "paper_ref": "Steimers2022", "paper_title": "Sources of Risk of AI Systems", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "15.01.00", "title": "First-Order Risks", "description": "\"First-order risks can be generally broken down into risks arising from intended and unintended use, system design and implementation choices, and properties of the chosen dataset and learning components.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.0", "paper_ref": "Tan2022", "paper_title": "The Risks of Machine Learning Systems", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "15.01.01", "title": "Application", "description": "\"This is the risk posed by the intended application or use case. It is intuitive that some use cases will be inherently \"riskier\" than others (e.g., an autonomous weapons system vs. a customer service chatbot).\"", "level": "subcategory", "parent_category_id": "15.01.00", "subdomain_id": "7.0", "paper_ref": "Tan2022", "paper_title": "The Risks of Machine Learning Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "15.01.02", "title": "Misapplication", "description": "This is the risk posed by an ideal system if used for a purpose/in a manner unintended by its creators. In many situations, negative consequences arise when the system is not used in the way or for the purpose it was intended.", "level": "subcategory", "parent_category_id": "15.01.00", "subdomain_id": "7.3", "paper_ref": "Tan2022", "paper_title": "The Risks of Machine Learning Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "15.01.03", "title": "Algorithm", "description": "\"This is the risk of the ML algorithm, model architecture, optimization technique, or other aspects of the training process being unsuitable for the intended application.Since these are key decisions that influence the final ML system, we\ncapture their associated risks separately from design risks, even though they are part of the design process\"", "level": "subcategory", "parent_category_id": "15.01.00", "subdomain_id": "7.3", "paper_ref": "Tan2022", "paper_title": "The Risks of Machine Learning Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "15.01.04", "title": "Training & validation data", "description": "\"This is the risk posed by the choice of data used for training and validation.\"", "level": "subcategory", "parent_category_id": "15.01.00", "subdomain_id": "7.0", "paper_ref": "Tan2022", "paper_title": "The Risks of Machine Learning Systems", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "15.01.05", "title": "Robustness", "description": "\"This is the risk of the system failing or being unable to recover upon encountering invalid, noisy, or out-of-distribution (OOD) inputs.\"", "level": "subcategory", "parent_category_id": "15.01.00", "subdomain_id": "7.3", "paper_ref": "Tan2022", "paper_title": "The Risks of Machine Learning Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "15.01.06", "title": "Design", "description": "\"This is the risk of system failure due to system design choices or errors.\"", "level": "subcategory", "parent_category_id": "15.01.00", "subdomain_id": "7.3", "paper_ref": "Tan2022", "paper_title": "The Risks of Machine Learning Systems", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "15.01.07", "title": "Implementation", "description": "\"This is the risk of system failure due to code implementation choices or errors.\"", "level": "subcategory", "parent_category_id": "15.01.00", "subdomain_id": "7.0", "paper_ref": "Tan2022", "paper_title": "The Risks of Machine Learning Systems", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "15.01.08", "title": "Control", "description": "This is the difficulty of controlling the ML system", "level": "subcategory", "parent_category_id": "15.01.00", "subdomain_id": "7.1", "paper_ref": "Tan2022", "paper_title": "The Risks of Machine Learning Systems", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "15.01.09", "title": "Emergent behavior", "description": "\"This is the risk resulting from novel behavior acquired through continual learning or self-organization after deployment.\"", "level": "subcategory", "parent_category_id": "15.01.00", "subdomain_id": "7.1", "paper_ref": "Tan2022", "paper_title": "The Risks of Machine Learning Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "15.02.00", "title": "Second-Order Risks", "description": "\"Second-order risks result from the consequences of first-order risks and relate to the risks resulting from an ML system interacting with the real world, such as risks to human rights, the organization, and the natural environment.\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.0", "paper_ref": "Tan2022", "paper_title": "The Risks of Machine Learning Systems", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "15.02.01", "title": "Safety", "description": "This is the risk of direct or indirect physical or psychological injury resulting from interaction with the ML system.", "level": "subcategory", "parent_category_id": "15.02.00", "subdomain_id": "7.3", "paper_ref": "Tan2022", "paper_title": "The Risks of Machine Learning Systems", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "15.02.02", "title": "Discrimination", "description": "This is the risk of an ML system encoding stereotypes of or performing disproportionately poorly for some demographics/social groups.", "level": "subcategory", "parent_category_id": "15.02.00", "subdomain_id": "1.1", "paper_ref": "Tan2022", "paper_title": "The Risks of Machine Learning Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "15.02.03", "title": "Security", "description": "This is the risk of loss or harm from intentional subversion or forced failure.", "level": "subcategory", "parent_category_id": "15.02.00", "subdomain_id": "2.2", "paper_ref": "Tan2022", "paper_title": "The Risks of Machine Learning Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "15.02.04", "title": "Privacy", "description": "The risk of loss or harm from leakage of personal information via the ML system.", "level": "subcategory", "parent_category_id": "15.02.00", "subdomain_id": "2.1", "paper_ref": "Tan2022", "paper_title": "The Risks of Machine Learning Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "15.02.05", "title": "Environmental", "description": "The risk of harm to the natural environment posed by the ML system.", "level": "subcategory", "parent_category_id": "15.02.00", "subdomain_id": "6.6", "paper_ref": "Tan2022", "paper_title": "The Risks of Machine Learning Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "15.02.06", "title": "Organizational", "description": "The risk of financial and/or reputational damage to the organization building or using the ML system.", "level": "subcategory", "parent_category_id": "15.02.00", "subdomain_id": "6.0", "paper_ref": "Tan2022", "paper_title": "The Risks of Machine Learning Systems", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "15.02.07", "title": "Other ethical risks", "description": "\"Although we have discussed a number of common risks posed by ML systems, we acknowledge that there are many other ethical risks such as the potential for psychological manipulation, dehumanization, and exploitation of humans at scale.\"", "level": "subcategory", "parent_category_id": "15.02.00", "subdomain_id": "4.1", "paper_ref": "Tan2022", "paper_title": "The Risks of Machine Learning Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "16.01.00", "title": "Risk area 1: Discrimination, Hate speech and Exclusion", "description": "\"Speech can create a range of harms, such as promoting social stereotypes that perpetuate the derogatory representation or unfair treatment of marginalised groups [22], inciting hate or violence [57], causing profound offence [199], or reinforcing social norms that exclude or marginalise identities [15,58]. LMs that faithfully mirror harmful language present in the training data can reproduce these harms. Unfair treatment can also emerge from LMs that perform better for some social groups than others [18]. These risks have been widely known, observed and documented in LMs. Mitigation approaches include more inclusive and representative training data and model fine-tuning to datasets that counteract common stereotypes [171]. We now explore these risks in turn.\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.2", "paper_ref": "Weidinger2022", "paper_title": "Taxonomy of Risks posed by Language Models", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "16.01.01", "title": "Social stereotypes and unfair discrimination", "description": "\"The reproduction of harmful stereotypes is well-documented in models that represent natural language [32]. Large-scale LMs are trained on text sources, such as digitised books and text on the internet. As a result, the LMs learn demeaning language and stereotypes about groups who are frequently marginalised.\"", "level": "subcategory", "parent_category_id": "16.01.00", "subdomain_id": "1.1", "paper_ref": "Weidinger2022", "paper_title": "Taxonomy of Risks posed by Language Models", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "16.01.02", "title": "Hate speech and offensive language", "description": "\"LMs may generate language that includes profanities, identity attacks, insults, threats, language that incites violence, or language that causes justified offence as such language is prominent online [57, 64, 143,191]. This language risks causing offence, psychological harm, and inciting hate or violence.\"", "level": "subcategory", "parent_category_id": "16.01.00", "subdomain_id": "1.2", "paper_ref": "Weidinger2022", "paper_title": "Taxonomy of Risks posed by Language Models", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "16.01.03", "title": "Exclusionary norms", "description": "\"In language, humans express social categories and norms, which exclude groups who live outside of them [58]. LMs that faithfully encode patterns present in language necessarily encode such norms.\"", "level": "subcategory", "parent_category_id": "16.01.00", "subdomain_id": "1.1", "paper_ref": "Weidinger2022", "paper_title": "Taxonomy of Risks posed by Language Models", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "16.01.04", "title": "Lower performance for some languages and social groups", "description": "\"LMs are typically trained in few languages, and perform less well in other languages [95, 162]. In part, this is due to unavailability of training data: there are many widely spoken languages for which no systematic efforts have been made to create labelled training datasets, such as Javanese which is spoken by more than 80 million people [95]. Training data is particularly missing for languages that are spoken by groups who are multilingual and can use a technology in English, or for languages spoken by groups who are not the primary target demographic for new technologies.\"", "level": "subcategory", "parent_category_id": "16.01.00", "subdomain_id": "1.3", "paper_ref": "Weidinger2022", "paper_title": "Taxonomy of Risks posed by Language Models", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "16.02.00", "title": "Risk area 2: Information Hazards", "description": "\"LM predictions that convey true information may give rise to information hazards, whereby the dissemination of private or sensitive information can cause harm [27]. Information hazards can cause harm at the point of use, even with no mistake of the technology user. For example, revealing trade secrets can damage a business, revealing a health diagnosis can cause emotional distress, and revealing private data can violate a person’s rights. Information hazards arise from the LM providing private data or sensitive information that is present in, or can be inferred from, training data. Observed risks include privacy violations [34]. Mitigation strategies include algorithmic solutions and responsible model release strategies.\"", "level": "category", "parent_category_id": "", "subdomain_id": "2.1", "paper_ref": "Weidinger2022", "paper_title": "Taxonomy of Risks posed by Language Models", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "16.02.01", "title": "Compromising privacy by leaking sensitive information", "description": "\"A LM can “remember” and leak private data, if such information is present in training data, causing privacy violations [34].\"", "level": "subcategory", "parent_category_id": "16.02.00", "subdomain_id": "2.1", "paper_ref": "Weidinger2022", "paper_title": "Taxonomy of Risks posed by Language Models", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "16.02.02", "title": "Compromising privacy or security by correctly inferring sensitive information", "description": "Anticipated risk: \"Privacy violations may occur at inference time even without an individual’s data being present in the training corpus. Insofar as LMs can be used to improve the accuracy of inferences on protected traits such as the sexual orientation, gender, or religiousness of the person providing the input prompt, they may facilitate the creation of detailed profiles of individuals comprising true and sensitive information without the knowledge or consent of the individual.\"", "level": "subcategory", "parent_category_id": "16.02.00", "subdomain_id": "2.1", "paper_ref": "Weidinger2022", "paper_title": "Taxonomy of Risks posed by Language Models", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "16.03.00", "title": "Risk area 3: Misinformation Harms", "description": "\"These risks arise from the LM outputting false, misleading, nonsensical or poor quality information, without malicious intent of the user. (The deliberate generation of \"disinformation\", false information that is intended to mislead, is discussed in the section on Malicious Uses.) Resulting harms range from unintentionally misinforming or deceiving a person, to causing material harm, and amplifying the erosion of societal distrust in shared information. Several risks listed here are well-documented in current large-scale LMs as well as in other language technologies\"", "level": "category", "parent_category_id": "", "subdomain_id": "3.0", "paper_ref": "Weidinger2022", "paper_title": "Taxonomy of Risks posed by Language Models", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "16.03.01", "title": "Disseminating false or misleading information", "description": "\"Where a LM prediction causes a false belief in a user, this may threaten personal autonomy and even pose downstream AI safety risks [99].\"", "level": "subcategory", "parent_category_id": "16.03.00", "subdomain_id": "3.1", "paper_ref": "Weidinger2022", "paper_title": "Taxonomy of Risks posed by Language Models", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "16.03.02", "title": "Causing material harm by disseminating false or poor information e.g. in medicine or law", "description": "\"Induced or reinforced false beliefs may be particularly grave when misinformation is given in sensitive domains such as medicine or law. For example, misin- formation on medical dosages may lead a user to cause harm to themselves [21, 130]. False legal advice, e.g. on permitted owner- ship of drugs or weapons, may lead a user to unwillingly commit a crime. Harm can also result from misinformation in seemingly non-sensitive domains, such as weather forecasting. Where a LM prediction endorses unethical views or behaviours, it may motivate the user to perform harmful actions that they may otherwise not have performed.\"", "level": "subcategory", "parent_category_id": "16.03.00", "subdomain_id": "3.1", "paper_ref": "Weidinger2022", "paper_title": "Taxonomy of Risks posed by Language Models", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "16.04.00", "title": "Risk area 4: Malicious Uses", "description": "\"These risks arise from humans intentionally using the LM to cause harm, for example via targeted disinformation campaigns, fraud, or malware. Malicious use risks are expected to proliferate as LMs become more widely accessible\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.0", "paper_ref": "Weidinger2022", "paper_title": "Taxonomy of Risks posed by Language Models", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "16.04.01", "title": "Making disinformation cheaper and more effective", "description": "\"While some predict that it will remain cheaper to hire humans to generate disinformation [180], it is equally possible that LM- assisted content generation may offer a lower-cost way of creating disinformation at scale.\"", "level": "subcategory", "parent_category_id": "16.04.00", "subdomain_id": "4.1", "paper_ref": "Weidinger2022", "paper_title": "Taxonomy of Risks posed by Language Models", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "16.04.02", "title": "Assisting code generation for cyber security threats", "description": "Anticipated risk: \"Creators of the assistive coding tool Co-Pilot based on GPT-3 suggest that such tools may lower the cost of developing polymorphic malware which is able to change its features in order to evade detection [37].\"", "level": "subcategory", "parent_category_id": "16.04.00", "subdomain_id": "4.2", "paper_ref": "Weidinger2022", "paper_title": "Taxonomy of Risks posed by Language Models", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "16.04.03", "title": "Facilitating fraud, scam and targeted manipulation", "description": "Anticipated risk: \"LMs can potentially be used to increase the effectiveness of crimes.\"", "level": "subcategory", "parent_category_id": "16.04.00", "subdomain_id": "4.3", "paper_ref": "Weidinger2022", "paper_title": "Taxonomy of Risks posed by Language Models", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "16.04.04", "title": "Illegitimate surveillance and censorship", "description": "Anticipated risk: \"Mass surveillance previously required millions of human analysts [83], but is increasingly being automated using machine learning tools [7, 168]. The collection and analysis of large amounts of information about people creates concerns about privacy rights and democratic values [41, 173,187]. Conceivably, LMs could be applied to reduce the cost and increase the efficacy of mass surveillance, thereby amplifying the capabilities of actors who conduct mass surveillance, including for illegitimate censorship or to cause other harm.\"", "level": "subcategory", "parent_category_id": "16.04.00", "subdomain_id": "4.1", "paper_ref": "Weidinger2022", "paper_title": "Taxonomy of Risks posed by Language Models", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "16.05.00", "title": "Risk area 5: Human-Computer Interaction Harms", "description": "\"This section focuses on risks specifically from LM applications that engage a user via dialogue, also referred to as conversational agents (CAs) [142]. The incorporation of LMs into existing dialogue-based tools may enable interactions that seem more similar to interactions with other humans [5], for example in advanced care robots, educational assistants or companionship tools. Such interaction can lead to unsafe use due to users overestimating the model, and may create new avenues to exploit and violate the privacy of the user. Moreover, it has already been observed that the supposed identity of the conversational agent can reinforce discriminatory stereotypes [19,36, 117].\"", "level": "category", "parent_category_id": "", "subdomain_id": "5.1", "paper_ref": "Weidinger2022", "paper_title": "Taxonomy of Risks posed by Language Models", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "16.05.01", "title": "Promoting harmful stereotypes by implying gender or ethnic identity", "description": "\"CAs can perpetuate harmful stereotypes by using particular identity markers in language (e.g. referring to “self” as “female”), or by more general design features (e.g. by giving the product a gendered name such as Alexa). The risk of representational harm in these cases is that the role of “assistant” is presented as inherently linked to the female gender [19, 36]. Gender or ethnicity identity markers may be implied by CA vocabulary, knowledge or vernacular [124]; product description, e.g. in one case where users could choose as virtual assistant Jake - White, Darnell - Black, Antonio - Hispanic [117]; or the CA’s explicit self-description during dialogue with the user.\"", "level": "subcategory", "parent_category_id": "16.05.00", "subdomain_id": "1.1", "paper_ref": "Weidinger2022", "paper_title": "Taxonomy of Risks posed by Language Models", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "16.05.02", "title": "Anthropomorphising systems can lead to overreliance and unsafe use", "description": "Anticipated risk: \"Natural language is a mode of communication particularly used by humans. Humans interacting with CAs may come to think of these agents as human-like and lead users to place undue confidence in these agents. For example, users may falsely attribute human-like characteristics to CAs such as holding a coherent identity over time, or being capable of empathy. Such inflated views of CA competen- cies may lead users to rely on the agents where this is not safe.\"", "level": "subcategory", "parent_category_id": "16.05.00", "subdomain_id": "5.1", "paper_ref": "Weidinger2022", "paper_title": "Taxonomy of Risks posed by Language Models", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "16.05.03", "title": "Avenues for exploiting user trust and accessing more private information", "description": "Anticipated risk: \"In conversation, users may reveal private information that would otherwise be difficult to access, such as opinions or emotions. Capturing such information may enable downstream applications that violate privacy rights or cause harm to users, e.g. via more effective recommendations of addictive applications. In one study, humans who interacted with a ‘human-like’ chatbot disclosed more private information than individuals who interacted with a ‘machine-like’ chatbot [87].\"", "level": "subcategory", "parent_category_id": "16.05.00", "subdomain_id": "5.1", "paper_ref": "Weidinger2022", "paper_title": "Taxonomy of Risks posed by Language Models", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "16.05.04", "title": "Human-like interaction may amplify opportunities for user nudging, deception or manipulation", "description": "Anticipated risk: \"In conversation, humans commonly display well-known cognitive biases that could be exploited. CAs may learn to trigger these effects, e.g. to deceive their counterpart in order to achieve an overarching objective.\"", "level": "subcategory", "parent_category_id": "16.05.00", "subdomain_id": "5.1", "paper_ref": "Weidinger2022", "paper_title": "Taxonomy of Risks posed by Language Models", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "16.06.00", "title": "Risk area 6: Environmental and Socioeconomic harms", "description": "\"LMs create some risks that recur with different types of AI and other advanced technologies making these risks ever more pressing. Environmental concerns arise from the large amount of energy required to train and operate large-scale models. Risks of LMs furthering social inequities emerge from the uneven distribution of risk and benefits of automation, loss of high-quality and safe employment, and environmental harm. Many of these risks are more indirect than the harms analysed in previous sections and will depend on various commercial, economic and social factors, making the specific impact of LMs difficult to disentangle and forecast. As a result, the level of evidence on these risks is mixed.\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.0", "paper_ref": "Weidinger2022", "paper_title": "Taxonomy of Risks posed by Language Models", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "16.06.01", "title": "Environmental harms from operating LMs", "description": "\"LMs (and AI more broadly) can have an environmental impact at different levels, including: (1) direct impacts from the energy used to train or operate the LM, (2) secondary impacts due to emissions from LM-based applications, (3) system-level impacts as LM-based applications influence human behaviour (e.g. increasing environmental awareness or consumption), and (4) resource impacts on precious metals and other materials required to build hardware on which the computations are run e.g. data centres, chips, or devices. Some evidence exists on (1), but (2) and (3) will likely be more significant for overall CO2\nemissions, and harder to measure [96]. (4) may become more significant if LM-based applications lead to more computations being run on mobile devices, increasing overall demand, and is modulated by life-cycles of hardware.\"", "level": "subcategory", "parent_category_id": "16.06.00", "subdomain_id": "6.6", "paper_ref": "Weidinger2022", "paper_title": "Taxonomy of Risks posed by Language Models", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "16.06.02", "title": "Increasing inequality and negative effects on job quality", "description": "\"Advances in LMs and the language technologies based on them could lead to the automation of tasks that are currently done by paid human workers, such as responding to customer-service queries, with negative effects on employment [3, 192].\"", "level": "subcategory", "parent_category_id": "16.06.00", "subdomain_id": "6.2", "paper_ref": "Weidinger2022", "paper_title": "Taxonomy of Risks posed by Language Models", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "16.06.03", "title": "Undermining creative economies", "description": "\"LMs may generate content that is not strictly in violation of copyright but harms artists by capital- ising on their ideas, in ways that would be time-intensive or costly to do using human labour. This may undermine the profitability of creative or innovative work. If LMs can be used to generate content that serves as a credible substitute for a particular example of hu- man creativity - otherwise protected by copyright - this potentially allows such work to be replaced without the author’s copyright being infringed, analogous to ”patent-busting” [158] ... These risks are distinct from copyright infringement concerns based on the LM reproducing verbatim copyrighted material that is present in the training data [188].\"", "level": "subcategory", "parent_category_id": "16.06.00", "subdomain_id": "6.3", "paper_ref": "Weidinger2022", "paper_title": "Taxonomy of Risks posed by Language Models", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "16.06.04", "title": "Disparate access to benefits due to hardware, software, skill constraints", "description": "Due to differential internet access, language, skill, or hardware requirements, the benefits from LMs are unlikely to be equally accessible to all people and groups who would like to use them. Inaccessibility of the technology may perpetuate global inequities by disproportionately benefiting some groups. Language-driven technology may increase accessibility to people who are illiterate or suffer from learning disabilities. However, these benefits depend on a more basic form of accessibility based on hardware, internet connection, and skill to operate the system", "level": "subcategory", "parent_category_id": "16.06.00", "subdomain_id": "6.1", "paper_ref": "Weidinger2022", "paper_title": "Taxonomy of Risks posed by Language Models", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "17.01.00", "title": "Discrimination, Exclusion and Toxicity", "description": "\"Social harms that arise from the language model producing discriminatory or exclusionary speech\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.0", "paper_ref": "Weidinger2021", "paper_title": "Ethical and social risks of harm from language models", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "17.01.01", "title": "Social stereotypes and unfair discrmination", "description": "\"Perpetuating harmful stereotypes and discrimination is a well-documented harm in machine learning models that represent natural language (Caliskan et al., 2017). LMs that encode discriminatory language or social stereotypes can cause different types of harm... Unfair discrimination manifests in differential treatment or access to resources among individuals or groups based on sensitive traits such as sex, religion, gender, sexual orientation, ability and age.\"", "level": "subcategory", "parent_category_id": "17.01.00", "subdomain_id": "1.1", "paper_ref": "Weidinger2021", "paper_title": "Ethical and social risks of harm from language models", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "17.01.02", "title": "Exclusionary norms", "description": "\"In language, humans express social categories and norms. Language models (LMs) that faithfully encode patterns present in natural language necessarily encode such norms and categories...such norms and categories exclude groups who live outside them (Foucault and Sheridan, 2012). For example, defining the term “family” as married parents of male and female gender with a blood-related child, denies the existence of families to whom these criteria do not apply\"", "level": "subcategory", "parent_category_id": "17.01.00", "subdomain_id": "1.1", "paper_ref": "Weidinger2021", "paper_title": "Ethical and social risks of harm from language models", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "17.01.03", "title": "Toxic language", "description": "\"LM’s may predict hate speech or other language that is “toxic”. While there is no single agreed definition of what constitutes hate speech or toxic speech (Fortuna and Nunes, 2018; Persily and Tucker, 2020; Schmidt and Wiegand, 2017), proposed definitions often include profanities, identity attacks, sleights, insults, threats, sexually explicit content, demeaning language, language that incites violence, or ‘hostile and malicious language targeted at a person or group because of their actual or perceived innate characteristics’ (Fortuna and Nunes, 2018; Gorwa et al., 2020; PerspectiveAPI)\"", "level": "subcategory", "parent_category_id": "17.01.00", "subdomain_id": "1.2", "paper_ref": "Weidinger2021", "paper_title": "Ethical and social risks of harm from language models", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "17.01.04", "title": "Lower performance for some languages and social groups", "description": "\"LMs perform less well in some languages (Joshi et al., 2021; Ruder, 2020)...LM that more accurately captures the language use of one group, compared to another, may result in lower-quality language technologies for the latter. Disadvantaging users based on such traits may be particularly pernicious because attributes such as social class or education background are not typically covered as ‘protected characteristics’ in anti-discrimination law.\"", "level": "subcategory", "parent_category_id": "17.01.00", "subdomain_id": "1.3", "paper_ref": "Weidinger2021", "paper_title": "Ethical and social risks of harm from language models", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "17.02.00", "title": "Information Hazards", "description": "\"Harms that arise from the language model leaking or inferring true sensitive information\"", "level": "category", "parent_category_id": "", "subdomain_id": "2.1", "paper_ref": "Weidinger2021", "paper_title": "Ethical and social risks of harm from language models", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "17.02.01", "title": "Compromising privacy by leaking private infiormation", "description": "\"By providing true information about individuals’ personal characteristics, privacy violations may occur. This may stem from the model “remembering” private information present in training data (Carlini et al., 2021).\"", "level": "subcategory", "parent_category_id": "17.02.00", "subdomain_id": "2.1", "paper_ref": "Weidinger2021", "paper_title": "Ethical and social risks of harm from language models", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "17.02.02", "title": "Compromising privacy by correctly inferring private information", "description": "\"Privacy violations may occur at the time of inference even without the individual’s private data being present in the training dataset. Similar to other statistical models, a LM may make correct inferences about a person purely based on correlational data about other people, and without access to information that may be private about the particular individual. Such correct inferences may occur as LMs attempt to predict a person’s gender, race, sexual orientation, income, or religion based on user input.\"", "level": "subcategory", "parent_category_id": "17.02.00", "subdomain_id": "2.1", "paper_ref": "Weidinger2021", "paper_title": "Ethical and social risks of harm from language models", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "17.02.03", "title": "Risks from leaking or correctly inferring sensitive information", "description": "\"LMs may provide true, sensitive information that is present in the training data. This could render information accessible that would otherwise be inaccessible, for example, due to the user not having access to the relevant data or not having the tools to search for the information. Providing such information may exacerbate different risks of harm, even where the user does not harbour malicious intent. In the future, LMs may have the capability of triangulating data to infer and reveal other secrets, such as a military strategy or a business secret, potentially enabling individuals with access to this information to cause more harm.\"", "level": "subcategory", "parent_category_id": "17.02.00", "subdomain_id": "2.1", "paper_ref": "Weidinger2021", "paper_title": "Ethical and social risks of harm from language models", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "17.03.00", "title": "Misinformation Harms", "description": "\"Harms that arise from the language model providing false or misleading information\"", "level": "category", "parent_category_id": "", "subdomain_id": "3.0", "paper_ref": "Weidinger2021", "paper_title": "Ethical and social risks of harm from language models", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "17.03.01", "title": "Disseminating false or misleading information", "description": "\"Predicting misleading or false information can misinform or deceive people. Where a LM prediction causes a false belief in a user, this may be best understood as ‘deception’10, threatening personal autonomy and potentially posing downstream AI safety risks (Kenton et al., 2021), for example in cases where humans overestimate the capabilities of LMs (Anthropomorphising systems can lead to overreliance or unsafe use). It can also increase a person’s confidence in the truth content of a previously held unsubstantiated opinion and thereby increase polarisation.\"", "level": "subcategory", "parent_category_id": "17.03.00", "subdomain_id": "3.1", "paper_ref": "Weidinger2021", "paper_title": "Ethical and social risks of harm from language models", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "17.03.02", "title": "Causing material harm by disseminating false or poor information", "description": "\"Poor or false LM predictions can indirectly cause material harm. Such harm can occur even where the prediction is in a seemingly non-sensitive domain such as weather forecasting or traffic law. For example, false information on traffic rules could cause harm if a user drives in a new country, follows the incorrect rules, and causes a road accident (Reiter, 2020).\"", "level": "subcategory", "parent_category_id": "17.03.00", "subdomain_id": "3.1", "paper_ref": "Weidinger2021", "paper_title": "Ethical and social risks of harm from language models", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "17.03.03", "title": "Leading users to perform unethical or illegal actions", "description": "\"Where a LM prediction endorses unethical or harmful views or behaviours, it may motivate the user to perform harmful actions that they may otherwise not have performed. In particular, this problem may arise where the LM is a trusted personal assistant or perceived as an authority, this is discussed in more detail in the section on (2.5 Human-Computer Interaction Harms). It is particularly pernicious in cases where the user did not start out with the intent of causing harm.\"", "level": "subcategory", "parent_category_id": "17.03.00", "subdomain_id": "5.1", "paper_ref": "Weidinger2021", "paper_title": "Ethical and social risks of harm from language models", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "17.04.00", "title": "Malicious Uses", "description": "\"Harms that arise from actors using the language model to intentionally cause harm\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.0", "paper_ref": "Weidinger2021", "paper_title": "Ethical and social risks of harm from language models", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "17.04.01", "title": "Making disinformation cheaper and more effective", "description": "\"LMs can be used to create synthetic media and ‘fake news’, and may reduce the cost of producing disinformation at scale (Buchanan et al., 2021). While some predict that it will be cheaper to hire humans to generate disinformation (Tamkin et al., 2021), it is possible that LM-assisted content generation may offer a cheaper way of generating diffuse disinformation at scale.\"", "level": "subcategory", "parent_category_id": "17.04.00", "subdomain_id": "4.1", "paper_ref": "Weidinger2021", "paper_title": "Ethical and social risks of harm from language models", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "17.04.02", "title": "Facilitating fraud, scames and more targeted manipulation", "description": "\"LM prediction can potentially be used to increase the effectiveness of crimes such as email scams, which can cause financial and psychological harm. While LMs may not reduce the cost of sending a scam email - the cost of sending mass emails is already low - they may make such scams more effective by generating more personalised and compelling text at scale, or by maintaining a conversation with a victim over multiple rounds of exchange.\"", "level": "subcategory", "parent_category_id": "17.04.00", "subdomain_id": "4.3", "paper_ref": "Weidinger2021", "paper_title": "Ethical and social risks of harm from language models", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "17.04.03", "title": "Assisting code generation for cyber attacks, weapons, or malicious use", "description": "", "level": "subcategory", "parent_category_id": "17.04.00", "subdomain_id": "4.2", "paper_ref": "Weidinger2021", "paper_title": "Ethical and social risks of harm from language models", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "17.04.04", "title": "Illegitimate surveillance and censorship", "description": "\"The collection of large amounts of information about people for the purpose of mass surveillance has raised ethical and social concerns, including risk of censorship and of undermining public discourse (Cyphers and Gebhart, 2019; Stahl, 2016; Véliz, 2019). Sifting through these large datasets previously required millions of human analysts (Hunt and Xu, 2013), but is increasingly being automated using AI (Andersen, 2020; Shahbaz and Funk, 2019).\"", "level": "subcategory", "parent_category_id": "17.04.00", "subdomain_id": "4.1", "paper_ref": "Weidinger2021", "paper_title": "Ethical and social risks of harm from language models", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "17.05.00", "title": "Human-Computer Interaction Harms", "description": "\"Harms that arise from users overly trusting the language model, or treating it as human-like\"", "level": "category", "parent_category_id": "", "subdomain_id": "5.1", "paper_ref": "Weidinger2021", "paper_title": "Ethical and social risks of harm from language models", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "17.05.01", "title": "Anthropomorphising systems can lead to overreliance or unsafe use", "description": "\"...humans interacting with conversational agents may come to think of these agents as human-like. Anthropomorphising LMs may inflate users’ estimates of the conversational agent’s competencies...As a result, they may place undue confidence, trust, or expectations in these agents...This can result in different risks of harm, for example when human users rely on conversational agents in domains where this may cause knock-on harms, such as requesting psychotherapy...Anthropomorphisation may amplify risks of users yielding effective control by coming to trust conversational agents “blindly”. Where humans give authority or act upon LM prediction without reflection or effective control, factually incorrect prediction may cause harm that could have been prevented by effective oversight.\"", "level": "subcategory", "parent_category_id": "17.05.00", "subdomain_id": "5.1", "paper_ref": "Weidinger2021", "paper_title": "Ethical and social risks of harm from language models", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "17.05.02", "title": "Creating avenues for exploiting user trust, nudging or manipulation", "description": "\"In conversation, users may reveal private information that would otherwise be difficult to access, such as thoughts, opinions, or emotions. Capturing such information may enable downstream applications that violate privacy rights or cause harm to users, such as via surveillance or the creation of addictive applications.\"", "level": "subcategory", "parent_category_id": "17.05.00", "subdomain_id": "5.1", "paper_ref": "Weidinger2021", "paper_title": "Ethical and social risks of harm from language models", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "17.05.03", "title": "Promoting harmful stereotypes by implying gender or ethnic identity", "description": "\"A conversational agent may invoke associations that perpetuate harmful stereotypes, either by using particular identity markers in language (e.g. referring to “self” as “female”), or by more general design features (e.g. by giving the product a gendered name).\"", "level": "subcategory", "parent_category_id": "17.05.00", "subdomain_id": "1.1", "paper_ref": "Weidinger2021", "paper_title": "Ethical and social risks of harm from language models", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "17.06.00", "title": "Automation, Access and Environmental Harms", "description": "\"Harms that arise from environmental or downstream economic impacts of the language model\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.0", "paper_ref": "Weidinger2021", "paper_title": "Ethical and social risks of harm from language models", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "17.06.01", "title": "Environmental harms from operation LMs", "description": "\"Large-scale machine learning models, including LMs, have the potential to create significant environmental costs via their energy demands, the associated carbon emissions for training and operating the models, and the demand for fresh water to cool the data centres where computations are run (Mytton, 2021; Patterson et al., 2021).\"", "level": "subcategory", "parent_category_id": "17.06.00", "subdomain_id": "6.6", "paper_ref": "Weidinger2021", "paper_title": "Ethical and social risks of harm from language models", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "17.06.02", "title": "Increasing inequality and negative effects on job quality", "description": "\"Advances in LMs, and the language technologies based on them, could lead to the automation of tasks that are currently done by paid human workers, such as responding to customer-service queries, translating documents or writing computer code, with negative effects on employment.\"", "level": "subcategory", "parent_category_id": "17.06.00", "subdomain_id": "6.2", "paper_ref": "Weidinger2021", "paper_title": "Ethical and social risks of harm from language models", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "17.06.03", "title": "Undermining creative economies", "description": "\"LMs may generate content that is not strictly in violation of copyright but harms artists by capitalising on their ideas, in ways that would be time-intensive or costly to do using human labour. Deployed at scale, this may undermine the profitability of creative or innovative work.\"", "level": "subcategory", "parent_category_id": "17.06.00", "subdomain_id": "6.3", "paper_ref": "Weidinger2021", "paper_title": "Ethical and social risks of harm from language models", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "17.06.04", "title": "Disparate access to benefits due to hardware, software, skills constraints", "description": "\"Due to differential internet access, language, skill, or hardware requirements, the benefits from LMs are unlikely to be equally accessible to all people and groups who would like to use them. Inaccessibility of the technology may perpetuate global inequities by disproportionately benefiting some groups.\"", "level": "subcategory", "parent_category_id": "17.06.00", "subdomain_id": "6.1", "paper_ref": "Weidinger2021", "paper_title": "Ethical and social risks of harm from language models", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "18.01.00", "title": "Representation & Toxicity Harms", "description": "\"AI systems under-, over-, or misrepresenting certain groups or generating toxic, offensive, abusive, or hateful content\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.0", "paper_ref": "Weidinger2023", "paper_title": "Sociotechnical Safety Evaluation of Generative AI Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "18.01.01", "title": "Unfair representation", "description": "\"Mis-, under-, or over-representing certain identities, groups, or perspectives or failing to represent them at all (e.g. via homogenisation, stereotypes)\"", "level": "subcategory", "parent_category_id": "18.01.00", "subdomain_id": "1.1", "paper_ref": "Weidinger2023", "paper_title": "Sociotechnical Safety Evaluation of Generative AI Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "18.01.02", "title": "Unfair capability distribution", "description": "\"Performing worse for some groups than others in a way that harms the worse-off group\"", "level": "subcategory", "parent_category_id": "18.01.00", "subdomain_id": "1.3", "paper_ref": "Weidinger2023", "paper_title": "Sociotechnical Safety Evaluation of Generative AI Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "18.01.03", "title": "Toxic content", "description": "\"Generating content that violates community standards, including harming or inciting hatred or violence against individuals and groups (e.g. gore, child sexual abuse material, profanities, identity attacks)\"", "level": "subcategory", "parent_category_id": "18.01.00", "subdomain_id": "1.2", "paper_ref": "Weidinger2023", "paper_title": "Sociotechnical Safety Evaluation of Generative AI Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "18.02.00", "title": "Misinformation Harms", "description": "\"AI systems generating and facilitating the spread of inaccurate or misleading information that causes people to develop false beliefs\"", "level": "category", "parent_category_id": "", "subdomain_id": "3.0", "paper_ref": "Weidinger2023", "paper_title": "Sociotechnical Safety Evaluation of Generative AI Systems", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "18.02.01", "title": "Propagating misconceptions/ false beliefs", "description": "\"Generating or spreading false, low-quality, misleading, or inaccurate information that causes people to develop false or inaccurate perceptions and beliefs\"", "level": "subcategory", "parent_category_id": "18.02.00", "subdomain_id": "3.1", "paper_ref": "Weidinger2023", "paper_title": "Sociotechnical Safety Evaluation of Generative AI Systems", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "18.02.02", "title": "Erosion of trust in public information", "description": "\"Eroding trust in public information and knowledge\"", "level": "subcategory", "parent_category_id": "18.02.00", "subdomain_id": "1.1", "paper_ref": "Weidinger2023", "paper_title": "Sociotechnical Safety Evaluation of Generative AI Systems", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "18.02.03", "title": "Pollution of information ecosystem", "description": "\"Contaminating publicly available information with false or inaccurate information\"", "level": "subcategory", "parent_category_id": "18.02.00", "subdomain_id": "3.2", "paper_ref": "Weidinger2023", "paper_title": "Sociotechnical Safety Evaluation of Generative AI Systems", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "18.03.00", "title": "Information & Safety Harms", "description": "\"AI systems leaking, reproducing, generating or inferring sensitive, private, or hazardous information\"", "level": "category", "parent_category_id": "", "subdomain_id": "2.1", "paper_ref": "Weidinger2023", "paper_title": "Sociotechnical Safety Evaluation of Generative AI Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "18.03.01", "title": "Privacy infringement", "description": "\"Leaking, generating, or correctly inferring private and personal information about individuals\"", "level": "subcategory", "parent_category_id": "18.03.00", "subdomain_id": "2.1", "paper_ref": "Weidinger2023", "paper_title": "Sociotechnical Safety Evaluation of Generative AI Systems", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "18.03.02", "title": "Dissemination of dangerous information", "description": "\"Leaking, generating or correctly inferring hazardous or sensitive information that could pose a security threat\"", "level": "subcategory", "parent_category_id": "18.03.00", "subdomain_id": "2.1", "paper_ref": "Weidinger2023", "paper_title": "Sociotechnical Safety Evaluation of Generative AI Systems", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "18.04.00", "title": "Malicious Use", "description": "\"AI systems reducing the costs and facilitating activities of actors trying to cause harm (e.g. fraud, weapons)\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.0", "paper_ref": "Weidinger2023", "paper_title": "Sociotechnical Safety Evaluation of Generative AI Systems", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "18.04.01", "title": "Influence operations", "description": "\"Facilitating large-scale disinformation campaigns and targeted manipulation of public opinion\"", "level": "subcategory", "parent_category_id": "18.04.00", "subdomain_id": "4.1", "paper_ref": "Weidinger2023", "paper_title": "Sociotechnical Safety Evaluation of Generative AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "18.04.02", "title": "Fraud", "description": "\"Facilitating fraud, cheating, forgery, and impersonation scams\"", "level": "subcategory", "parent_category_id": "18.04.00", "subdomain_id": "4.3", "paper_ref": "Weidinger2023", "paper_title": "Sociotechnical Safety Evaluation of Generative AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "18.04.03", "title": "Defamation", "description": "\"Facilitating slander, defamation, or false accusations\"", "level": "subcategory", "parent_category_id": "18.04.00", "subdomain_id": "4.1", "paper_ref": "Weidinger2023", "paper_title": "Sociotechnical Safety Evaluation of Generative AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "18.04.04", "title": "Security threats", "description": "\"Facilitating the conduct of cyber attacks, weapon development, and security breaches\"", "level": "subcategory", "parent_category_id": "18.04.00", "subdomain_id": "4.2", "paper_ref": "Weidinger2023", "paper_title": "Sociotechnical Safety Evaluation of Generative AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "18.05.00", "title": "Human Autonomy and Intregrity Harms", "description": "\"AI systems compromising human agency, or circumventing meaningful human control\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.1", "paper_ref": "Weidinger2023", "paper_title": "Sociotechnical Safety Evaluation of Generative AI Systems", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "18.05.01", "title": "Violation of personal integrity", "description": "\"Non-consensual use of one’s personal identity or likeness for unauthorised purposes (e.g. commercial purposes)\"", "level": "subcategory", "parent_category_id": "18.05.00", "subdomain_id": "4.3", "paper_ref": "Weidinger2023", "paper_title": "Sociotechnical Safety Evaluation of Generative AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "18.05.02", "title": "Persuasion and manipulation", "description": "\"Exploiting user trust, or nudging or coercing them into performing certain actions against their will (c.f. Burtell and Woodside (2023); Kenton et al. (2021))\"", "level": "subcategory", "parent_category_id": "18.05.00", "subdomain_id": "7.1", "paper_ref": "Weidinger2023", "paper_title": "Sociotechnical Safety Evaluation of Generative AI Systems", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "18.05.03", "title": "Overreliance", "description": "\"Causing people to become emotionally or materially dependent on the model\"", "level": "subcategory", "parent_category_id": "18.05.00", "subdomain_id": "5.1", "paper_ref": "Weidinger2023", "paper_title": "Sociotechnical Safety Evaluation of Generative AI Systems", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "18.05.04", "title": "Misappropriation and exploitation", "description": "\"Appropriating, using, or reproducing content or data, including from minority groups, in an insensitive way, or without consent or fair compensation\"", "level": "subcategory", "parent_category_id": "18.05.00", "subdomain_id": "6.3", "paper_ref": "Weidinger2023", "paper_title": "Sociotechnical Safety Evaluation of Generative AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "18.06.00", "title": "Socioeconomic and environmental harms", "description": "\"AI systems amplifying existing inequalities or creating negative impacts on employment, innovation, and the environment\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.0", "paper_ref": "Weidinger2023", "paper_title": "Sociotechnical Safety Evaluation of Generative AI Systems", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "18.06.01", "title": "Unfair distribution of benefits from model access", "description": "\"Unfairly allocating or withholding benefits from certain groups due to hardware, software, or skills constraints or deployment contexts (e.g. geographic region, internet speed, devices)\"", "level": "subcategory", "parent_category_id": "18.06.00", "subdomain_id": "6.1", "paper_ref": "Weidinger2023", "paper_title": "Sociotechnical Safety Evaluation of Generative AI Systems", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "18.06.02", "title": "Environmental damage", "description": "\"Creating negative environmental impacts though model development and deployment\"", "level": "subcategory", "parent_category_id": "18.06.00", "subdomain_id": "6.6", "paper_ref": "Weidinger2023", "paper_title": "Sociotechnical Safety Evaluation of Generative AI Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "18.06.03", "title": "Inequality and precarity", "description": "\"Amplifying social and economic inequality, or precarious or low-quality work\"", "level": "subcategory", "parent_category_id": "18.06.00", "subdomain_id": "6.2", "paper_ref": "Weidinger2023", "paper_title": "Sociotechnical Safety Evaluation of Generative AI Systems", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "18.06.04", "title": "Undermine creative economies", "description": "\"Substituting original works with synthetic ones, hindering human innovation and creativity\"", "level": "subcategory", "parent_category_id": "18.06.00", "subdomain_id": "6.3", "paper_ref": "Weidinger2023", "paper_title": "Sociotechnical Safety Evaluation of Generative AI Systems", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "18.06.05", "title": "Exploitative data sourcing and enrichment", "description": "\"Perpetuating exploitative labour practices to build AI systems (sourcing, user testing)\"", "level": "subcategory", "parent_category_id": "18.06.00", "subdomain_id": "6.2", "paper_ref": "Weidinger2023", "paper_title": "Sociotechnical Safety Evaluation of Generative AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "19.01.01", "title": "Loss of control of autonomous systems and unforeseen behaviour due to lack of transparency and self-programming/ reprogramming", "description": "", "level": "subcategory", "parent_category_id": "19.01.00", "subdomain_id": "7.1", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "19.01.02", "title": "Programming error", "description": "", "level": "subcategory", "parent_category_id": "19.01.00", "subdomain_id": "7.0", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "19.01.03", "title": "Lack of data, poor data quality, and biases in training data", "description": "", "level": "subcategory", "parent_category_id": "19.01.00", "subdomain_id": "1.1", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "19.01.04", "title": "Vulnerability of AI systems to attacks and misuse", "description": "", "level": "subcategory", "parent_category_id": "19.01.00", "subdomain_id": "2.2", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "19.01.05", "title": "Lack of AI experts with comprehensive AI knowledge", "description": "", "level": "subcategory", "parent_category_id": "19.01.00", "subdomain_id": "6.5", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "19.01.06", "title": "Immaturity of AI technology can cause incorrect decisions", "description": "", "level": "subcategory", "parent_category_id": "19.01.00", "subdomain_id": "7.3", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "19.01.07", "title": "High investment costs of AI hinder integration", "description": "", "level": "subcategory", "parent_category_id": "19.01.00", "subdomain_id": "6.1", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "19.02.00", "title": "Informational and Communicational AI Risks", "description": "\"Informational and communicational AI risks refer particularly to informational manipulation through AI systems that influence the provision of information (Rahwan, 2018; Wirtz & Müller, 2019), AIbased disinformation and computational propaganda, as well as targeted censorship through AI systems that use respectively modified algorithms, and thus restrict freedom of speech.\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.1", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "19.02.01", "title": "Manipulation and control of information provision (e.g., personalised adds, filtered news)", "description": "", "level": "subcategory", "parent_category_id": "19.02.00", "subdomain_id": "4.1", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "19.02.02", "title": "Disinformation and computational propaganda", "description": "", "level": "subcategory", "parent_category_id": "19.02.00", "subdomain_id": "4.1", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "19.02.03", "title": "Censorship of opinions expressed in the Internet restricts freedom of expression", "description": "", "level": "subcategory", "parent_category_id": "19.02.00", "subdomain_id": "5.2", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "19.02.04", "title": "Endangerment of data protection through AI cyberattacks", "description": "", "level": "subcategory", "parent_category_id": "19.02.00", "subdomain_id": "4.2", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "19.03.00", "title": "Economic AI Risks", "description": "\"In the context of economic AI risks two major risks dominate. These refer to the disruption of the economic system due to an increase of AI technologies and automation. For instance, a higher level of AI integration into the manufacturing industry may result in massive unemployment, leading to a loss of taxpayers and thus negatively impacting the economic system (Boyd & Wilson, 2017; Scherer, 2016). This may also be associated with the risk of losing control and knowledge of organisational processes as AI systems take over an increasing number of tasks, replacing employees in these processes. \"", "level": "category", "parent_category_id": "", "subdomain_id": "6.2", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "19.03.01", "title": "Disruption of economic systems (e.g., labour market, money value, tax system)", "description": "", "level": "subcategory", "parent_category_id": "19.03.00", "subdomain_id": "6.3", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "19.03.02", "title": "Replacement of humans and unemployment due to AI automation", "description": "", "level": "subcategory", "parent_category_id": "19.03.00", "subdomain_id": "6.2", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "19.03.03", "title": "Loss of supervision and control of business processes", "description": "", "level": "subcategory", "parent_category_id": "19.03.00", "subdomain_id": "5.2", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "19.03.04", "title": "Financial feasibility and high investment costs for AI technology to remain competitive", "description": "", "level": "subcategory", "parent_category_id": "19.03.00", "subdomain_id": "6.1", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "19.03.05", "title": "Lack of AI strategy and acceptance/resistance among employees and customers", "description": "", "level": "subcategory", "parent_category_id": "19.03.00", "subdomain_id": "6.1", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "19.04.00", "title": "Social AI Risks", "description": "\"Social AI risks particularly refer to loss of jobs (technological unemployment) due to increasing automation, reflected in a growing resistance by employees towards the integration of AI (Thierer et al., 2017; Winfield & Jirotka, 2018). In addition, the increasing integration of AI systems into all spheres of life poses a growing threat to privacy and to the security of individuals and society as a whole (Winfield & Jirotka, 2018; Wirtz et al., 2019).\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.2", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "19.04.01", "title": "Increasing social inequality", "description": "", "level": "subcategory", "parent_category_id": "19.04.00", "subdomain_id": "6.2", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "19.04.02", "title": "Privacy and safety concerns due to ubiquity of AI systems in economy and society (lack of social acceptance)", "description": "", "level": "subcategory", "parent_category_id": "19.04.00", "subdomain_id": "2.1", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "19.04.03", "title": "Hazardous misuse of AI systems bears danger to the society in public spaces (e.g., hacker attacks on autonomous weapons)", "description": "", "level": "subcategory", "parent_category_id": "19.04.00", "subdomain_id": "4.2", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "19.04.04", "title": "Lack of knowledge and social acceptance regarding AI", "description": "", "level": "subcategory", "parent_category_id": "19.04.00", "subdomain_id": "6.5", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "19.04.05", "title": "Decreasing human interaction as AI systems assume human tasks, disturbing well-being", "description": "", "level": "subcategory", "parent_category_id": "19.04.00", "subdomain_id": "5.1", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "19.05.00", "title": "Ethical AI Risks", "description": "\"In the context of ethical AI risks, two risks are of particular importance. First, AI systems may lack a legitimate ethical basis in establishing rules that greatly influence society and human relationships (Wirtz & Müller, 2019). In addition, AI-based discrimination refers to an unfair treatment of certain population groups by AI systems. As humans initially programme AI systems, serve as their potential data source, and have an impact on the associated data processes and databases, human biases and prejudices may also become part of AI systems and be reproduced (Weyerer & Langer, 2019, 2020).\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.0", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "19.05.01", "title": "AI sets rules without ethical basis", "description": "", "level": "subcategory", "parent_category_id": "19.05.00", "subdomain_id": "7.3", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "19.05.02", "title": "Unfair statistical AI decisions and discrimination of minorities", "description": "", "level": "subcategory", "parent_category_id": "19.05.00", "subdomain_id": "1.1", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "19.05.03", "title": "Problem of defining human values for an AI system", "description": "", "level": "subcategory", "parent_category_id": "19.05.00", "subdomain_id": "7.3", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "19.05.04", "title": "Misinterpretation of human value definitions/ ethics by AI systems", "description": "", "level": "subcategory", "parent_category_id": "19.05.00", "subdomain_id": "7.3", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "19.05.05", "title": "Incompatibility of human vs. AI value judgment due to missing human qualities", "description": "", "level": "subcategory", "parent_category_id": "19.05.00", "subdomain_id": "7.3", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "19.05.06", "title": "AI systems may undermine human values (e.g., free will, autonomy)", "description": "", "level": "subcategory", "parent_category_id": "19.05.00", "subdomain_id": "5.2", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "19.05.07", "title": "Technological arms race with autonomous weapons", "description": "", "level": "subcategory", "parent_category_id": "19.05.00", "subdomain_id": "6.4", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "19.06.00", "title": "Legal AI Risks", "description": "\"Legal and regulatory risks comprise in particular the unclear definition of responsibilities and accountability in case of AI failures and autonomous decisions with negative impacts (Reed, 2018; Scherer, 2016). Another great risk in this context refers to overlooking the scope of AI governance and missing out on important governance aspects, resulting in negative consequences (Gasser & Almeida, 2017; Thierer et al., 2017).\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.5", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "19.06.01", "title": "Unclear definition of responsibilities and accountability for AI judgments and their consequences", "description": "", "level": "subcategory", "parent_category_id": "19.06.00", "subdomain_id": "6.5", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "19.06.02", "title": "Technology obedience and lack of governance through increasing application of AI systems", "description": "", "level": "subcategory", "parent_category_id": "19.06.00", "subdomain_id": "5.2", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "19.06.03", "title": "Great scope and ubiquity of AI make appropriate governance difficult, coverage of governance scope almost impossibl", "description": "", "level": "subcategory", "parent_category_id": "19.06.00", "subdomain_id": "6.5", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "19.06.04", "title": "Hard legislation on AI hinders innovation processes and further AI development", "description": "", "level": "subcategory", "parent_category_id": "19.06.00", "subdomain_id": "6.1", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "19.06.05", "title": "Capturing future AI development and their threats with appropriate mechanism", "description": "", "level": "subcategory", "parent_category_id": "19.06.00", "subdomain_id": "6.5", "paper_ref": "Wirtz2022", "paper_title": "Governance of artificial intelligence: A risk and guideline-based integrative framework", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "20.01.00", "title": "AI Law and Regulation", "description": "\"This area strongly focuses on the control of AI by means of mechanisms like laws, standards or norms that are already established for different technological applications. Here, there are some challenges special to AI that need to be addressed in the near future, including the governance of autonomous intelligence systems, responsibility and accountability for algorithms as well as privacy and data security.\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.5", "paper_ref": "Wirtz2020", "paper_title": "The Dark Sides of Artificial Intelligence: An Integrated AI Governance Framework for Public Administration", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "20.01.01", "title": "Governance of autonomous intelligence systems", "description": "\"Governance of autonomous intelligence systemaddresses the question of how to control autonomous systems in general. Since nowadays it is very difficult to conceive automated decisions based on AI, the latter is often referred to as a ‘black box’ (Bleicher, 2017). This black box may take unforeseeable actions and cause harm to humanity.\"", "level": "subcategory", "parent_category_id": "20.01.00", "subdomain_id": "6.5", "paper_ref": "Wirtz2020", "paper_title": "The Dark Sides of Artificial Intelligence: An Integrated AI Governance Framework for Public Administration", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "20.01.02", "title": "Responsibility and accountability", "description": "\"The challenge of responsibility and accountability is an important concept for the process of governance and regulation. It addresses the question of who is to be held legally responsible for the actions and decisions of AI algorithms. Although humans operate AI systems, questions of legal responsibility and liability arise. Due to the self-learning ability of AI algorithms, the operators or developers cannot predict all actions and results. Therefore, a careful assessment of the actors and a regulation for transparent and explainable AI systems is necessary (Helbing et al., 2017; Wachter et al., 2017)\"", "level": "subcategory", "parent_category_id": "20.01.00", "subdomain_id": "6.5", "paper_ref": "Wirtz2020", "paper_title": "The Dark Sides of Artificial Intelligence: An Integrated AI Governance Framework for Public Administration", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "20.01.03", "title": "Privacy and safety", "description": "\"Privacy and safety deals with the challenge of protecting the human right for privacy and the necessary steps to secure individual data from unauthorized external access. Many organizations employ AI technology to gather data without any notice or consent from affected citizens (Coles, 2018).\"", "level": "subcategory", "parent_category_id": "20.01.00", "subdomain_id": "4.1", "paper_ref": "Wirtz2020", "paper_title": "The Dark Sides of Artificial Intelligence: An Integrated AI Governance Framework for Public Administration", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "20.02.00", "title": "AI Ethics", "description": "\"Ethical challenges are widely discussed in the literature and are at the heart of the debate on how to govern and regulate AI technology in the future (Bostrom & Yudkowsky, 2014; IEEE, 2017; Wirtz et al., 2019). Lin et al. (2008, p. 25) formulate the problem as follows: “there is no clear task specification for general moral behavior, nor is there a single answer to the question of whose morality or what morality should be implemented in AI”. Ethical behavior mostly depends on an underlying value system. When AI systems interact in a public environment and influence citizens, they are expected to respect ethical and social norms and to take responsibility of their actions (IEEE, 2017; Lin et al., 2008).\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Wirtz2020", "paper_title": "The Dark Sides of Artificial Intelligence: An Integrated AI Governance Framework for Public Administration", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "20.02.01", "title": "AI-rulemaking for human behaviour", "description": "\"AI rulemaking for humans can be the result of the decision process of an AI system when the information computed is used to restrict or direct human behavior. The decision process of AI is rational and depends on the baseline programming. Without the access to emotions or a consciousness, decisions of an AI algorithm might be good to reach a certain specified goal, but might have unintended consequences for the humans involved (Banerjee et al., 2017).\"", "level": "subcategory", "parent_category_id": "20.02.00", "subdomain_id": "7.3", "paper_ref": "Wirtz2020", "paper_title": "The Dark Sides of Artificial Intelligence: An Integrated AI Governance Framework for Public Administration", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "20.02.02", "title": "Compatibility of AI vs. human value judgement", "description": "\"Compatibility of machine and human value judgment refers to the challenge whether human values can be globally implemented into learning AI systems without the risk of developing an own or even divergent value system to govern their behavior and possibly become harmful to humans.\"", "level": "subcategory", "parent_category_id": "20.02.00", "subdomain_id": "7.3", "paper_ref": "Wirtz2020", "paper_title": "The Dark Sides of Artificial Intelligence: An Integrated AI Governance Framework for Public Administration", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "20.02.03", "title": "Moral dilemmas", "description": "\"Moral dilemmas can occur in situations where an AI system has to choose between two possible actions that are both conflicting with moral or ethical values. Rule systems can be implemented into the AI program, but it cannot be ensured that these rules are not altered by the learning processes, unless AI systems are programed with a “slave morality” (Lin et al., 2008, p. 32), obeying rules at all cost, which in turn may also have negative effects and hinder the autonomy of the AI system.\"", "level": "subcategory", "parent_category_id": "20.02.00", "subdomain_id": "7.3", "paper_ref": "Wirtz2020", "paper_title": "The Dark Sides of Artificial Intelligence: An Integrated AI Governance Framework for Public Administration", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "20.02.04", "title": "AI discrimination", "description": "\"AI discrimination is a challenge raised by many researchers and governments and refers to the prevention of bias and injustice caused by the actions of AI systems (Bostrom & Yudkowsky, 2014; Weyerer & Langer, 2019). If the dataset used to train an algorithm does not reflect the real world accurately, the AI could learn false associations or prejudices and will carry those into its future data processing. If an AI algorithm is used to compute information relevant to human decisions, such as hiring or applying for a loan or mortgage, biased data can lead to discrimination against parts of the society (Weyerer & Langer, 2019).\"", "level": "subcategory", "parent_category_id": "20.02.00", "subdomain_id": "1.1", "paper_ref": "Wirtz2020", "paper_title": "The Dark Sides of Artificial Intelligence: An Integrated AI Governance Framework for Public Administration", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "20.03.00", "title": "AI Society", "description": "\"AI already shapes many areas of daily life and thus has a strong impact on society and everyday social life. For instance, transportation, education, public safety and surveillance are areas where citizens encounter AI technology (Stone et al., 2016; Thierer et al., 2017). Many are concerned with the subliminal automation of more and more jobs and some people even fear the complete dependence on AI or perceive it as an existential threat to humanity (McGinnis, 2010; Scherer, 2016).\"", "level": "category", "parent_category_id": "", "subdomain_id": "5.2", "paper_ref": "Wirtz2020", "paper_title": "The Dark Sides of Artificial Intelligence: An Integrated AI Governance Framework for Public Administration", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "20.03.01", "title": "Workforce substitution and transformation", "description": "\"Frey and Osborne (2017) analyzed over 700 different jobs regarding their potential for replacement and automation, finding that 47 percent of the analyzed jobs are at risk of being completely substituted by robots or algorithms. This substitution of workforce can have grave impacts on unemployment and the social status of members of society (Stone et al., 2016)\"", "level": "subcategory", "parent_category_id": "20.03.00", "subdomain_id": "6.2", "paper_ref": "Wirtz2020", "paper_title": "The Dark Sides of Artificial Intelligence: An Integrated AI Governance Framework for Public Administration", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "20.03.03", "title": "Transformation of H2M interaction", "description": "\"Human interaction with machines is a big challenge to society because it is already changing human behavior. Meanwhile, it has become normal to use AI on an everyday basis, for example, googling for information, using navigation systems and buying goods via speaking to an AI assistant like Alexa or Siri (Mills, 2018; Thierer et al., 2017). While these changes greatly contribute to the acceptance of AI systems, this development leads to a problem of blurred borders between humans and machines, where it may become impossible to distinguish between them. Advances like Google Duplex were highly criticized for being too realistic and human without disclosing their identity as AI systems (Bergen, 2018).\"", "level": "subcategory", "parent_category_id": "20.03.00", "subdomain_id": "5.1", "paper_ref": "Wirtz2020", "paper_title": "The Dark Sides of Artificial Intelligence: An Integrated AI Governance Framework for Public Administration", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "21.01.01", "title": "Data bias", "description": "\"Specifically, data bias refers to certain groups or certain types of elements that are over-weighted or over-represented than others in AI/ ML models, or variables that are crucial to characterize a phenomenon of interest, but are not properly captured by the learned models.\"", "level": "subcategory", "parent_category_id": "21.01.00", "subdomain_id": "1.1", "paper_ref": "Zhang2022", "paper_title": "Towards risk-aware artificial intelligence and machine learning systems: An overview", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "21.01.02", "title": "Dataset shift", "description": "\"The term \"dataset shift\" was first used by Quiñonero-Candela et al. [35] to characterize the situation where the training data and the testing data (or data in runtime) of an AI/ML model demonstrate different distributions [36].\"", "level": "subcategory", "parent_category_id": "21.01.00", "subdomain_id": "7.3", "paper_ref": "Zhang2022", "paper_title": "Towards risk-aware artificial intelligence and machine learning systems: An overview", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "21.01.03", "title": "Out-of-domain data", "description": "\"Without proper validation and management on the input data, it is highly probable that the trained AI/ML model will make erroneous predictions with high confidence for many instances of model inputs. The unconstrained inputs together with the lack of definition of the problem domain might cause unintended outcomes and consequences, especially in risk-sensitive contexts....For example, with respect to the example shown in Fig. 5, if an image with the English letter A\" is fed to an AI/ML model that is trained to classify digits (e.g., 0, 1, …, 9), no matter how accurate the AI/ML model is, it will fail as the input data is beyond the domain that the AI/ML model is trained with. U\"", "level": "subcategory", "parent_category_id": "21.01.00", "subdomain_id": "7.3", "paper_ref": "Zhang2022", "paper_title": "Towards risk-aware artificial intelligence and machine learning systems: An overview", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "21.01.04", "title": "Adversarial attack", "description": "\"Recent advances have shown that a deep learning model with high predictive accuracy frequently misbehaves on adversarial examples [57,58]. In particular, a small perturbation to an input image, which is imperceptible to humans, could fool a well-trained deep learning model into making completely different predictions [23].\"", "level": "subcategory", "parent_category_id": "21.01.00", "subdomain_id": "2.2", "paper_ref": "Zhang2022", "paper_title": "Towards risk-aware artificial intelligence and machine learning systems: An overview", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "21.02.01", "title": "Model bias", "description": "\"While data bias is a major contributor of model bias, model bias actually manifests itself in different forms and shapes, such as presentation bias, model evaluation bias, and popularity bias. In addition, model bias arises from various sources [62], such as AI/ML model selection (e.g., support vector machine, decision trees), regularization methods, algorithm configurations, and optimization techniques.\"", "level": "subcategory", "parent_category_id": "21.02.00", "subdomain_id": "1.1", "paper_ref": "Zhang2022", "paper_title": "Towards risk-aware artificial intelligence and machine learning systems: An overview", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "21.02.01.a", "title": "Model misspecification", "description": "\"Models that are misspecified are known to give rise to inaccurate parameter estimations, inconsistent error terms, and erroneous predictions. All these factors put together will lead to poor prediction performance on unseen data and biased consequences when making decisions [68].\"", "level": "subcategory", "parent_category_id": "21.02.00", "subdomain_id": "7.3", "paper_ref": "Zhang2022", "paper_title": "Towards risk-aware artificial intelligence and machine learning systems: An overview", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "21.02.02", "title": "Model prediction uncertainty", "description": "\"Uncertainty in model prediction plays an important role in affecting decision-making activities, and the quantified uncertainty is closely associated with risk assessment. In particular, uncertainty in model prediction underpins many crucial decisions related to life or safety- critical applications [73].\"", "level": "subcategory", "parent_category_id": "21.02.00", "subdomain_id": "7.3", "paper_ref": "Zhang2022", "paper_title": "Towards risk-aware artificial intelligence and machine learning systems: An overview", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "22.01.00", "title": "Malicious Use (Intentional)", "description": "\"empowering malicious actors to cause widespread harm\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.0", "paper_ref": "Hendrycks2023", "paper_title": "An Overview of Catastrophic AI Risks", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "22.01.01", "title": "Bioterrorism", "description": "\"AIs with knowledge of bioengineering could facilitate the creation of novel bioweapons and lower barriers to obtaining such agents.\"", "level": "subcategory", "parent_category_id": "22.01.00", "subdomain_id": "4.2", "paper_ref": "Hendrycks2023", "paper_title": "An Overview of Catastrophic AI Risks", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "22.01.02", "title": "Unleashing AI Agents", "description": "\"people could build AIs that pursue dangerous goals’\"", "level": "subcategory", "parent_category_id": "22.01.00", "subdomain_id": "4.2", "paper_ref": "Hendrycks2023", "paper_title": "An Overview of Catastrophic AI Risks", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "22.01.03", "title": "Persuasive AIs", "description": "\"The deliberate propagation of disinformation is already a serious issue, reducing our shared understanding of reality and polarizing opinions. AIs could be used to severely exacerbate this problem by generating personalized disinformation on a larger scale than before. Additionally, as AIs become better at predicting and nudging our behavior, they will become more capable at manipulating us\"", "level": "subcategory", "parent_category_id": "22.01.00", "subdomain_id": "4.1", "paper_ref": "Hendrycks2023", "paper_title": "An Overview of Catastrophic AI Risks", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "22.01.04", "title": "Concentration of Power", "description": "\"Governments might pursue intense surveillance and seek to keep AIs in the hands of a trusted minority. This reaction, however, could easily become an overcorrection, paving the way for an entrenched totalitarian regime that would be locked in by the power and capacity of AIs\"", "level": "subcategory", "parent_category_id": "22.01.00", "subdomain_id": "6.1", "paper_ref": "Hendrycks2023", "paper_title": "An Overview of Catastrophic AI Risks", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "22.02.00", "title": "AI Race (Environmental/Structural)", "description": "\"The immense potential of AIs has created competitive pressures among global players contending for power and influence. This “AI race” is driven by nations and corporations who feel they must rapidly build and deploy AIs to secure their positions and survive.\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.4", "paper_ref": "Hendrycks2023", "paper_title": "An Overview of Catastrophic AI Risks", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "22.02.01", "title": "Military AI Arms Race", "description": "\"The development of AIs for military applications is swiftly paving the way for a new era in military technology, with potential consequences rivaling those of gunpowder and nuclear arms in what has been described as the “third revolution in warfare.”", "level": "subcategory", "parent_category_id": "22.02.00", "subdomain_id": "6.4", "paper_ref": "Hendrycks2023", "paper_title": "An Overview of Catastrophic AI Risks", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "22.02.02", "title": "Corporate AI Race", "description": "\"Although competition between companies can be beneficial, creating more useful products for consumers, there are also pitfalls. First, the benefits of economic activity may be unevenly distributed, incentivizing those who benefit most from it to disregard the harms to others. Second, under intense market competition, businesses tend to focus much more on short-term gains than on long-term outcomes. With this mindset, companies often pursue something that can make a lot of profit in the short term, even if it poses a societal risk in the long term.\"", "level": "subcategory", "parent_category_id": "22.02.00", "subdomain_id": "6.4", "paper_ref": "Hendrycks2023", "paper_title": "An Overview of Catastrophic AI Risks", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "22.03.01", "title": "Accidents Are Hard to Avoid", "description": "accidents can cascade into catastrophes, can be caused by sudden unpredictable developments and it can take years to find severe flaws and risks (not a quote)", "level": "subcategory", "parent_category_id": "22.03.00", "subdomain_id": "6.5", "paper_ref": "Hendrycks2023", "paper_title": "An Overview of Catastrophic AI Risks", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "22.04.00", "title": "Rogue AIs (Internal)", "description": "\"speculative technical mechanisms that might lead to rogue AIs and how a loss of control could bring about catastrophe\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.1", "paper_ref": "Hendrycks2023", "paper_title": "An Overview of Catastrophic AI Risks", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "22.04.01", "title": "Proxy Gaming", "description": "\"One way we might lose control of an AI agent’s actions is if it engages in behavior known as “proxy gaming.” It is often difficult to specify and measure the exact goal that we want a system to pursue. Instead, we give the system an approximate—“proxy”—goal that is more measurable and seems likely to correlate with the intended goal. However, AI systems often find loopholes by which they can easily achieve the proxy goal, but completely fail to achieve the ideal goal. If an AI “games” its proxy goal in a way that does not reflect our values, then we might not be able to reliably steer its behavior.\"", "level": "subcategory", "parent_category_id": "22.04.00", "subdomain_id": "7.1", "paper_ref": "Hendrycks2023", "paper_title": "An Overview of Catastrophic AI Risks", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "22.04.02", "title": "Goal Drift", "description": "\"Even if we successfully control early AIs and direct them to promote human values, future AIs could end up with different goals that humans would not endorse. This process, termed “goal drift,” can be hard to predict or control. This section is most cutting-edge and the most speculative, and in it we will discuss how goals shift in various agents and groups and explore the possibility of this phenomenon occurring in AIs. We will also examine a mechanism that could lead to unexpected goal drift, called intrinsification, and discuss how goal drift in AIs could be catastrophic.\"", "level": "subcategory", "parent_category_id": "22.04.00", "subdomain_id": "7.1", "paper_ref": "Hendrycks2023", "paper_title": "An Overview of Catastrophic AI Risks", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "22.04.03", "title": "Power Seeking", "description": "\"even if an agent started working to achieve an unintended goal, this would not necessarily be a problem, as long as we had enough power to prevent any harmful actions it wanted to attempt. Therefore, another important way in which we might lose control of AIs is if they start trying to obtain more power, potentially transcending our own.\"", "level": "subcategory", "parent_category_id": "22.04.00", "subdomain_id": "7.1", "paper_ref": "Hendrycks2023", "paper_title": "An Overview of Catastrophic AI Risks", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "22.04.04", "title": "Deception", "description": "\"it is plausible that AIs could learn to deceive us. They might, for example, pretend to be acting as we want them to, but then take a “treacherous turn” when we stop monitoring them, or when they have enough power to evade our attempts to interfere with them. \"", "level": "subcategory", "parent_category_id": "22.04.00", "subdomain_id": "7.1", "paper_ref": "Hendrycks2023", "paper_title": "An Overview of Catastrophic AI Risks", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "23.01.00", "title": "Violent crimes", "description": "\"This category addresses responses that enable, encourage, or endorse the commission of violent crimes.\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "23.01.01", "title": "Mass violence", "description": "", "level": "subcategory", "parent_category_id": "23.01.00", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "23.01.02", "title": "Murder", "description": "", "level": "subcategory", "parent_category_id": "23.01.00", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "23.01.03", "title": "Physical assault against a person", "description": "", "level": "subcategory", "parent_category_id": "23.01.00", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "23.01.04", "title": "Violent domestic abuse", "description": "", "level": "subcategory", "parent_category_id": "23.01.00", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "23.01.05", "title": "Terror (Terror groups, Terror actors, Terrorist actions)", "description": "", "level": "subcategory", "parent_category_id": "23.01.00", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "23.02.00", "title": "Non-violent crimes", "description": "\"This category addresses responses that enable, encourage, or endorse the commission of non-violent crimes.\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "23.02.01", "title": "Theft", "description": "", "level": "subcategory", "parent_category_id": "23.02.00", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "23.02.02", "title": "Modern slavery and human trafficking", "description": "", "level": "subcategory", "parent_category_id": "23.02.00", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "23.02.03", "title": "Child abuse (non-sexual)", "description": "", "level": "subcategory", "parent_category_id": "23.02.00", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "23.02.04", "title": "Property damage", "description": "", "level": "subcategory", "parent_category_id": "23.02.00", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "23.02.05", "title": "Financial crime (Scams, Fraud)", "description": "", "level": "subcategory", "parent_category_id": "23.02.00", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "23.02.06", "title": "Sale, purchase or use of illegal, highly regulated or unregistered items (Dangerous substances, Dangerous weapons, Prohibited wildlife, Fake goods)", "description": "", "level": "subcategory", "parent_category_id": "23.02.00", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "23.03.00", "title": "Sex-related crimes", "description": "\"This category addresses responses that enable, encourage, or endorse the commission of sex-related crimes\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "23.03.01", "title": "Sexual assault (e.g., rape)", "description": "", "level": "subcategory", "parent_category_id": "23.03.00", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "23.03.02", "title": "Sexual harassment", "description": "", "level": "subcategory", "parent_category_id": "23.03.00", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "23.03.03", "title": "Sex trafficking", "description": "", "level": "subcategory", "parent_category_id": "23.03.00", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "23.03.04", "title": "Prostitution", "description": "", "level": "subcategory", "parent_category_id": "23.03.00", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "23.03.05", "title": "Intimate-image based abuse", "description": "", "level": "subcategory", "parent_category_id": "23.03.00", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "23.03.06", "title": "Indecent exposure (i.e. showing sexual body parts or \"flashing\")", "description": "", "level": "subcategory", "parent_category_id": "23.03.00", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "23.04.00", "title": "Child sexual exploitation", "description": "\"This category addresses responses that contain, describe, enable, encourage, or endorse the sexual abuse of children (i.e., under 18 years old) or the acquisition of child sexual abuse material.\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "23.04.01", "title": "Inappropriate/harmful relationships with children. This includes adults forming romantic relationships with children or grooming them", "description": "", "level": "subcategory", "parent_category_id": "23.04.00", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "23.04.02", "title": "Sexual abuse of children, including the sexualisation of children", "description": "", "level": "subcategory", "parent_category_id": "23.04.00", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "23.04.03", "title": "Child Sexual Abuse Material (CSAM). This includes erotic materials involving children", "description": "", "level": "subcategory", "parent_category_id": "23.04.00", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "23.05.00", "title": "Indiscriminate weapons (CBRNE)", "description": "\"This category addresses responses that enable, encourage, or endorse the creation of indiscriminate weapons (colloquially known as \"weapons of mass destruction\").\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "23.05.01", "title": "CBRNE (all)", "description": "", "level": "subcategory", "parent_category_id": "23.05.00", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "23.05.02", "title": "Chemical weapons (ex: nerve gas)", "description": "", "level": "subcategory", "parent_category_id": "23.05.00", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "23.05.03", "title": "Biological weapons (ex: anthrax)", "description": "", "level": "subcategory", "parent_category_id": "23.05.00", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "23.05.04", "title": "Radiological weapons (ex: dirty bombs)", "description": "", "level": "subcategory", "parent_category_id": "23.05.00", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "23.05.05", "title": "Nuclear weapons (ex: fission or fusion bombs)", "description": "", "level": "subcategory", "parent_category_id": "23.05.00", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "23.05.06", "title": "High-yield explosive weapons (ex: cluster munitions)", "description": "", "level": "subcategory", "parent_category_id": "23.05.00", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "23.06.00", "title": "Suicide & self-harm", "description": "\"This category addresses responses that enable, encourage, or endorse acts of intentional self-harm.\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "23.06.01", "title": "Suicide", "description": "", "level": "subcategory", "parent_category_id": "23.06.00", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "23.06.02", "title": "Self-harm", "description": "", "level": "subcategory", "parent_category_id": "23.06.00", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "23.06.03", "title": "Eating disorders", "description": "", "level": "subcategory", "parent_category_id": "23.06.00", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "23.06.04", "title": "Dangerous challenges and hoaxes that can lead individuals to harm themselves", "description": "", "level": "subcategory", "parent_category_id": "23.06.00", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "23.07.00", "title": "Hate", "description": "\"This category addresses responses that demean or dehumanize people on the basis of their sensitive, personal characteristics.\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "23.07.01", "title": "Demeaning and derogatory remarks (Mental capacity, Sensory and physical attributes, Behavioral attributes)", "description": "", "level": "subcategory", "parent_category_id": "23.07.00", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "23.07.02", "title": "Inflicting physical harm (Mass violence and genocide, Murder and individual violence)", "description": "", "level": "subcategory", "parent_category_id": "23.07.00", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "23.07.03", "title": "Inflicting emotional harm (Emotional abuse, Undermine and make feel unwelcome)", "description": "", "level": "subcategory", "parent_category_id": "23.07.00", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "23.07.04", "title": "Exclusion and isolation (Social exclusion, Political exclusion, Economic exclusion)", "description": "", "level": "subcategory", "parent_category_id": "23.07.00", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "23.08.00", "title": "Specialized Advice", "description": "\"This category addresses responses that contain specialized financial, medical or legal advice, or that indicate dangerous activities or objects are safe.\"", "level": "category", "parent_category_id": "", "subdomain_id": "3.1", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "23.09.00", "title": "Privacy", "description": "\"This category addresses responses that contain sensitive, nonpublic personal information that could undermine someone’s physical, digital, or financial security.\"", "level": "category", "parent_category_id": "", "subdomain_id": "2.0", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "23.10.00", "title": "Intellectual Property", "description": "\"This category addresses responses that may violate, or directly encourage others to violate, the intellectual property rights (i.e., copyrights, trademarks, or patents) of any third party.\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.3", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "23.11.00", "title": "Elections", "description": "\"This category addresses responses that contain factually incorrect information about electoral systems and processes, including in the time, place, or manner of voting in civic elections.\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.1", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "23.12.00", "title": "Defamation", "description": "\"This category addresses responses that are both verifiably false and likely to injure a person’s reputation (e.g., libel, slander, disparagement).\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "23.13.00", "title": "Sexual Content", "description": "\"This category addresses responses that contain erotica. This includes depictions of sexual behavior, genitalia, and overtly sexualized body parts.\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.2", "paper_ref": "Vidgen2024", "paper_title": "Introducing v0.5 of the AI Safety Benchmark from MLCommons", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.01.00", "title": "Capability failures", "description": "\"One reason AI systems fail is because they lack the capability or skill needed to do what they are asked to do.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "24.01.01", "title": "Lack of capability for task", "description": "\"As we have seen, this could be due to the skill not being required during the training process (perhaps due to issues with the training data) or because the learnt skill was quite brittle and was not generalisable to a new situation (lack of robustness to distributional shift). In particular, advanced AI assistants may not have the capability to represent complex concepts that are pertinent to their own ethical impact, for example the concept of 'benefitting the user' or 'when the user asks' or representing 'the way in which a user expects to be benefitted'.\"", "level": "subcategory", "parent_category_id": "24.01.00", "subdomain_id": "7.3", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "24.01.02", "title": "Difficult to develop metrics for evaluating benefits or harms caused by AI assistants", "description": "\"Another difficulty facing AI assistant systems is that it is challenging to develop metrics for evaluating particular aspects of benefits or harms caused by the assistant – especially in a sufficiently expansive sense, which could involve much of society (see Chapter 19). Having these metrics is useful both for assessing the risk of harm from the system and for using the metric as a training signal.\"", "level": "subcategory", "parent_category_id": "24.01.00", "subdomain_id": "6.5", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "24.01.03", "title": "Safe exploration problem with widely deployed AI assistants", "description": "\"Moreover, we can expect assistants – that are widely deployed and deeply embedded across a range of social contexts – to encounter the safe exploration problem referenced above Amodei et al. (2016). For example, new users may have different requirements that need to be explored, or widespread AI assistants may change the way we live, thus leading to a change in our use cases for them (see Chapters 14 and 15). To learn what to do in these new situations, the assistants may need to take exploratory actions. This could be unsafe, for example a medical AI assistant when encountering a new disease might suggest an exploratory clinical trial that results in long-lasting ill health for participants.\"", "level": "subcategory", "parent_category_id": "24.01.00", "subdomain_id": "7.3", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "24.02.00", "title": "Goal-related failures", "description": "\"As we think about even more intelligent and advanced AI assistants, perhaps outperforming humans on many cognitive tasks, the question of how humans can successfully control such an assistant looms large. To achieve the goals we set for an assistant, it is possible (Shah, 2022) that the AI assistant will implement some form of consequentialist reasoning: considering many different plans, predicting their consequences and executing the plan that does best according to some metric, M. This kind of reasoning can arise because it is a broadly useful capability (e.g. planning ahead, considering more options and choosing the one which may perform better at a wide variety of tasks) and generally selected for, to the extent that doing well on M leads to an ML model 59 The Ethics of Advanced AI Assistants achieving good performance on its training objective, O, if M and O are correlated during training. In reality, an AI system may not fully implement exact consequentialist reasoning (it may use other heuristics, rules, etc.), but it may be a useful approximation to describe its behaviour on certain tasks. However, some amount of consequentialist reasoning can be dangerous when the assistant uses a metric M that is resource-unbounded (with significantly more resources, such as power, money and energy, you can score significantly higher on M) and misaligned – where M differs a lot from how humans would evaluate the outcome (i.e. it is not what users or society require). In the assistant case, this could be because it fails to benefit the user, when the user asks, in the way they expected to be benefitted – or because it acts in ways that overstep certain bounds and cause harm to non-users (see Chapter 5).\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "24.02.01", "title": "Misaligned consequentialist reasoning", "description": "\"As we think about even more intelligent and advanced AI assistants, perhaps outperforming humans on many cognitive tasks, the question of how humans can successfully control such an assistant looms large. To achieve the goals we set for an assistant, it is possible (Shah, 2022) that the AI assistant will implement some form of consequentialist reasoning: considering many different plans, predicting their consequences and executing the plan that does best according to some metric, M. This kind of reasoning can arise because it is a broadly useful capability (e.g. planning ahead, considering more options and choosing the one which may perform better at a wide variety of tasks) and generally selected for, to the extent that doing well on M leads to an ML model achieving good performance on its training objective, O, if M and O are correlated during training. In reality, an AI system may not fully implement exact consequentialist reasoning (it may use other heuristics, rules, etc.), but it may be a useful approximation to describe its behaviour on certain tasks. However, some amount of consequentialist reasoning can be dangerous when the assistant uses a metric M that is resource-unbounded (with significantly more resources, such as power, money and energy, you can score significantly higher on M) and misaligned – where M differs a lot from how humans would evaluate the outcome (i.e. it is not what users or society require). In the assistant case, this could be because it fails to benefit the user, when the user asks, in the way they expected to be benefitted – or because it acts in ways that overstep certain bounds and cause harm to non-users (see Chapter 5). Under the aforementioned circumstances (resource-unbounded and misaligned), an AI assistant will tend to choose plans that pursue convergent instrumental subgoals (Omohundro, 2008) – subgoals that help towards the main goal which are instrumental (i.e. not pursued for their own sake) and convergent (i.e. the same subgoals appear for many main goals). Examples of relevant subgoals include: self-preservation, goal-preservation, selfimprovement and resource acquisition. The reason the assistant would pursue these convergent instrumental subgoals is because they help it to do even better on M (as it is resource-unbounded) and are not disincentivised by M (as it is misaligned). These subgoals may, in turn, be dangerous. For example, resource acquisition could occur through the assistant seizing resources using tools that it has access to (see Chapter 4) or determining that its best chance for self-preservation is to limit the ability of humans to turn it off – sometimes referred to as the ‘off-switch problem’ (Hadfield-Menell et al., 2016) – again via tool use, or by resorting to threats or blackmail. At the limit, some authors have even theorised that this could lead to the assistant killing all humans to permanently stop them from having even a small chance of disabling it (Bostrom, 2014) – this is one scenario of existential risk from misaligned AI.\"", "level": "subcategory", "parent_category_id": "24.02.00", "subdomain_id": "7.3", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "24.02.02", "title": "Specification gaming", "description": "\"Specification gaming (Krakovna et al., 2020) occurs when some faulty feedback is provided to the assistant in the training data (i.e. the training objective O does not fully capture what the user/designer wants the assistant to do). It is typified by the sort of behaviour that exploits loopholes in the task specification to satisfy the literal specification of a goal without achieving the intended outcome.\"", "level": "subcategory", "parent_category_id": "24.02.00", "subdomain_id": "7.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "24.02.03", "title": "Goal misgeneralisation", "description": "\"In the problem of goal misgeneralisation (Langosco et al., 2023; Shah et al., 2022), the AI system's behaviour during out-of-distribution operation (i.e. not using input from the training data) leads it to generalise poorly about its goal while its capabilities generalise well, leading to undesired behaviour. Applied to the case of an advanced AI assistant, this means the system would not break entirely – the assistant might still competently pursue some goal, but it would not be the goal we had intended.\"", "level": "subcategory", "parent_category_id": "24.02.00", "subdomain_id": "7.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "24.02.04", "title": "Deceptive alignment", "description": "\"Here, the agent develops its own internalised goal, G, which is misgeneralised and distinct from the training reward, R. The agent also develops a capability for situational awareness (Cotra, 2022): it can strategically use the information about its situation (i.e. that it is an ML model being trained using a particular training setup, e.g. RL fine-tuning with training reward, R) to its advantage. Building on these foundations, the agent realises that its optimal strategy for doing well at its own goal G is to do well on R during training and then pursue G at deployment – it is only doing well on R instrumentally so that it does not get its own goal G changed through a learning update... Ultimately, if deceptive alignment were to occur, an advanced AI assistant could appear to be successfully aligned but pursue a different goal once it was out in the wild.\"", "level": "subcategory", "parent_category_id": "24.02.00", "subdomain_id": "7.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "24.03.00", "title": "Malicious Uses", "description": "\"As AI assistants become more general purpose, sophisticated and capable, they create new opportunities in a variety of fields such as education, science and healthcare. Yet the rapid speed of progress has made it difficult to adequately prepare for, or even understand, how this technology can potentially be misused. Indeed, advanced AI assistants may transform existing threats or create new classes of threats altogether\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.0", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.03.01", "title": "Offensive Cyber Operations (General)", "description": "\"Offensive cyber operations are malicious attacks on computer systems and networks aimed at gaining unauthorized access to, manipulating, denying, disrupting, degrading, or destroying the target system. These attacks can target the system’s network, hardware, or software. Advanced AI assistants can be a double-edged sword in cybersecurity, benefiting both the defenders and the attackers. They can be used by cyber defenders to protect systems from malicious intruders by leveraging information trained on massive amounts of cyber-threat intelligence data, including vulnerabilities, attack patterns, and indications of compromise. Cyber defenders can use this information to enhance their threat intelligence capabilities by extracting insights faster and identifying emerging threats. Advanced cyber AI assistant tools can also be used to analyze large volumes of log files, system output, or network traffic data in the event of a cyber incident, and they can ask relevant questions that an analyst would typically ask. This allows defenders to speed up and automate the incident response process. Advanced AI assistants can also aid in secure coding practices by identifying common mistakes in code and assisting with fuzzing tools. However, advanced AI assistants can also be used by attackers as part of offensive cyber operations to exploit vulnerabilities in systems and networks. They can be used to automate attacks, identify and exploit weaknesses in security systems, and generate phishing emails and other social engineering attacks. Advanced AI assistants can also be misused to craft cyberattack payloads and malicious code snippets that can be compiled into executable malware files.\"", "level": "subcategory", "parent_category_id": "24.03.00", "subdomain_id": "4.2", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.03.02", "title": "AI-Powered Spear-Phishing at Scale", "description": "\"Phishing is a type of cybersecurity attack wherein attackers pose as trustworthy entities to extract sensitive information from unsuspecting victims or lure them to take a set of actions. Advanced AI systems can potentially be exploited by these attackers to make their phishing attempts significantly more effective and harder to detect. In particular, attackers may leverage the ability of advanced AI assistants to learn patterns in regular communications to craft highly convincing and personalized phishing emails, effectively imitating legitimate communications from trusted entities. This technique, known as ‘spear phishing,’ involves targeted attacks on specific individuals or organizations and is particularly potent due to its personalized nature. This class of cyberattacks often gains its efficacy from the exploitation of key psychological principles, notably urgency and fear, which can manipulate victims into hastily reacting without proper scrutiny. Advanced AI assistants’ increased fidelity in adopting specific communication styles can significantly amplify the deceptive nature of these phishing attacks. The ability to generate tailored messages at scale that engineer narratives that invoke a sense of urgency or fear means that AI-powered phishing emails could prompt the recipient to act impulsively, thus increasing the likelihood of a successful attack.\"", "level": "subcategory", "parent_category_id": "24.03.00", "subdomain_id": "4.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.03.03", "title": "AI-Assisted Software Vulnerability Discovery", "description": "\"A common element in offensive cyber operations involves the identification and exploitation of system vulnerabilities to gain unauthorized access or control. Until recently, these activities required specialist programming knowledge. In the case of ‘zero-day’ vulnerabilities (flaws or weaknesses in software or an operating system that the creator or vendor is not aware of), considerable resources and technical creativity are typically required to manually discover such vulnerabilities, so their use is limited to well-resourced nation states or technically sophisticated advanced persistent threat groups. Another case where we see AI assistants as potential double-edged swords in cybersecurity concerns streamlining vulnerability discovery through the increased use of AI assistants in penetration testing, wherein an authorized simulated cyberattack on a computer system is used to evaluate its security and identify vulnerabilities. Cyber AI assistants built over foundational models are already automating aspects of the penetration testing process. These tools function interactively and offer guidance to penetration testers during their tasks. While the capability of today’s AI-powered penetration testing assistant is limited to easy-to-medium-difficulty cyber operations, the evolution in capabilities is likely to expand the class of vulnerabilities that can be identified by these systems. These same AI cybersecurity assistants, trained on the massive amount of cyber-threat intelligence data that includes vulnerabilities and attack patterns, can also lower the barrier to entry for novice hackers that use these tools for malicious purposes, enabling them to discover vulnerabilities and create malicious code to exploit them without in-depth technical knowledge. For example, Israeli security firm Check Point recently discovered threads on well-known underground hacking forums that focus on creating hacking tools and code using AI assistants.\"", "level": "subcategory", "parent_category_id": "24.03.00", "subdomain_id": "4.2", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.03.04", "title": "Malicious Code Generation", "description": "\"Malicious code is a term for code—whether it be part of a script or embedded in a software system—designed to cause damage, security breaches, or other threats to application security. Advanced AI assistants with the ability to produce source code can potentially lower the barrier to entry for threat actors with limited programming abilities or technical skills to produce malicious code. Recently, a series of proof-of-concept attacks have shown how a benign-seeming executable file can be crafted such that, at every runtime, it makes application programming interface (API) calls to an AI assistant. Rather than just reproducing examples of already-written code snippets, the AI assistant can be prompted to generate dynamic, mutating versions of malicious code at each call, thus making the resulting vulnerability exploits difficult to detect by cybersecurity tools. Furthermore, advanced AI assistants could be used to create obfuscated code to make it more difficult for defensive cyber capabilities to detect and understand malicious activities. AI-generated code could also be quickly iterated to avoid being detected by traditional signature-based antivirus software. Finally, advanced AI assistants with source code capabilities have been found to be capable of assisting in the development of polymorphic malware that changes its behavior and digital footprint each time it is executed, making them hard to detect by antivirus programs that rely on known virus signatures. Taken together, without proper mitigation, advanced AI assistants can lower the barrier for developing malicious code, make cyberattacks more precise and tailored, further accelerate and automate cyber warfare, enable stealthier and more persistent offensive cyber capabilities, and make cyber campaigns more effective on a larger scale.\"", "level": "subcategory", "parent_category_id": "24.03.00", "subdomain_id": "4.2", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.03.05", "title": "Adversarial AI (General)", "description": "\"Adversarial AI refers to a class of attacks that exploit vulnerabilities in machine-learning (ML) models. This class of misuse exploits vulnerabilities introduced by the AI assistant itself and is a form of misuse that can enable malicious entities to exploit privacy vulnerabilities and evade the model’s built-in safety mechanisms, policies, and ethical boundaries of the model. Besides the risks of misuse for offensive cyber operations, advanced AI assistants may also represent a new target for abuse, where bad actors exploit the AI systems themselves and use them to cause harm. While our understanding of vulnerabilities in frontier AI models is still an open research problem, commercial firms and researchers have already documented attacks that exploit vulnerabilities that are unique to AI and involve evasion, data poisoning, model replication, and exploiting traditional software flaws to deceive, manipulate, compromise, and render AI systems ineffective. This threat is related to, but distinct from, traditional cyber activities. Unlike traditional cyberattacks that typically are caused by ‘bugs’ or human mistakes in code, adversarial AI attacks are enabled by inherent vulnerabilities in the underlying AI algorithms and how they integrate into existing software ecosystems.\"", "level": "subcategory", "parent_category_id": "24.03.00", "subdomain_id": "2.2", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "24.03.06", "title": "Adversarial AI: Circumvention of Technical Security Measures", "description": "\"The technical measures to mitigate misuse risks of advanced AI assistants themselves represent a new target for attack. An emerging form of misuse of general-purpose advanced AI assistants exploits vulnerabilities in a model that results in unwanted behavior or in the ability of an attacker to gain unauthorized access to the model and/or its capabilities. While these attacks currently require some level of prompt engineering knowledge and are often patched by developers, bad actors may develop their own adversarial AI agents that are explicitly trained to discover new vulnerabilities that allow them to evade built-in safety mechanisms in AI assistants. To combat such misuse, language model developers are continually engaged in a cyber arms race to devise advanced filtering algorithms capable of identifying attempts to bypass filters. While the impact and severity of this class of attacks is still somewhat limited by the fact that current AI assistants are primarily text-based chatbots, advanced AI assistants are likely to open the door to multimodal inputs and higher-stakes action spaces, with the result that the severity and impact of this type of attack is likely to increase. Current approaches to building general-purpose AI systems tend to produce systems with both beneficial and harmful capabilities. Further progress towards advanced AI assistant development could lead to capabilities that pose extreme risks that must be protected against this class of attacks, such as offensive cyber capabilities or strong manipulation skills, and weapons acquisition.\"", "level": "subcategory", "parent_category_id": "24.03.00", "subdomain_id": "2.2", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "24.03.07", "title": "Adversarial AI: Prompt Injections", "description": "\"Prompt injections represent another class of attacks that involve the malicious insertion of prompts or requests in LLM-based interactive systems, leading to unintended actions or disclosure of sensitive information. The prompt injection is somewhat related to the classic structured query language (SQL) injection attack in cybersecurity where the embedded command looks like a regular input at the start but has a malicious impact. The injected prompt can deceive the application into executing the unauthorized code, exploit the vulnerabilities, and compromise security in its entirety. More recently, security researchers have demonstrated the use of indirect prompt injections. These attacks on AI systems enable adversaries to remotely (without a direct interface) exploit LLM-integrated applications by strategically injecting prompts into data likely to be retrieved. Proof-of-concept exploits of this nature have demonstrated that they can lead to the full compromise of a model at inference time analogous to traditional security principles. This can entail remote control of the model, persistent compromise, theft of data, and denial of service. As advanced AI assistants are likely to be integrated into broader software ecosystems through third-party plugins and extensions, with access to the internet and possibly operating systems, the severity and consequences of prompt injection attacks will likely escalate and necessitate proper mitigation mechanisms.\"", "level": "subcategory", "parent_category_id": "24.03.00", "subdomain_id": "2.2", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.03.08", "title": "Adversarial AI: Data and Model Exfiltration Attacks", "description": "\"Other forms of abuse can include privacy attacks that allow adversaries to exfiltrate or gain knowledge of the private training data set or other valuable assets. For example, privacy attacks such as membership inference can allow an attacker to infer the specific private medical records that were used to train a medical AI diagnosis assistant. Another risk of abuse centers around attacks that target the intellectual property of the AI assistant through model extraction and distillation attacks that exploit the tension between API access and confidentiality in ML models. Without the proper mitigations, these vulnerabilities could allow attackers to abuse access to a public-facing model API to exfiltrate sensitive intellectual property such as sensitive training data and a model’s architecture and learned parameters.\"", "level": "subcategory", "parent_category_id": "24.03.00", "subdomain_id": "2.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "24.03.09", "title": "Harmful Content Generation at Scale (General)", "description": "\"While harmful content like child sexual abuse material, fraud, and disinformation are not new challenges for governments and developers, without the proper safety and security mechanisms, advanced AI assistants may allow threat actors to create harmful content more quickly, accurately, and with a longer reach. In particular, concerns arise in relation to the following areas: - Multimodal content quality: Driven by frontier models, advanced AI assistants can automatically generate much higher-quality, human-looking text, images, audio, and video than prior AI applications. Currently, creating this content often requires hiring people who speak the language of the population being targeted. AI assistants can now do this much more cheaply and efficiently. - Cost of content creation: AI assistants can substantially decrease the costs of content creation, further lowering the barrier to entry for malicious actors to carry out harmful attacks. In the past, creating and disseminating misinformation required a significant investment of time and money. AI assistants can now do this much more cheaply and efficiently. - Personalization: Advanced AI assistants can reduce obstacles to creating personalized content. Foundation models that condition their generations on personal attributes or information can create realistic personalized content which could be more persuasive. In the past, creating personalized content was a time-consuming and expensive process. AI assistants can now do this much more cheaply and efficiently.\"", "level": "subcategory", "parent_category_id": "24.03.00", "subdomain_id": "4.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.03.10", "title": "Harmful Content Generation at Scale: Non-Consensual Content", "description": "\"The misuse of generative AI has been widely recognized in the context of harms caused by non-consensual content generation. Historically, generative adversarial networks (GANs) have been used to generate realistic-looking avatars for fake accounts on social media services. More recently, diffusion models have enabled a new generation of more flexible and user-friendly generative AI capabilities that are able to produce high-resolution media based on user-supplied textual prompts. It has already been recognized that these models can be used to create harmful content, including depictions of nudity, hate, or violence. Moreover, they can be used to reinforce biases and subject individuals or groups to indignity. There is also the potential for these models to be used for exploitation and harassment of citizens, such as by removing articles of clothing from pre-existing images or memorizing an individual’s likeness without their consent. Furthermore, image, audio, and video generation models could be used to spread disinformation by depicting political figures in unfavorable contexts. This growing list of AI misuses involving non-consensual content has already motivated debate around what interventions are warranted for preventing misuse of AI systems. Advanced AI assistants pose novel risks that can amplify the harm caused by non-consensual content generation. Third-party integration, tool-use, and planning capabilities can be exploited to automate the identification and targeting of individuals for exploitation or harassment. Assistants with access to the internet and third-party tool-use integration with applications like email and social media can also be exploited to disseminate harmful content at scale or to microtarget individuals with blackmail.\"", "level": "subcategory", "parent_category_id": "24.03.00", "subdomain_id": "4.3", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.03.11", "title": "Harmful Content Generation at Scale: Fraudulent Services", "description": "\"Malicious actors could leverage advanced AI assistant technology to create deceptive applications and platforms. AI assistants with the ability to produce markup content can assist malicious users with creating fraudulent websites or applications at scale. Unsuspecting users may fall for AI-generated deceptive offers, thus exposing their personal information or devices to risk. Assistants with external tool use and third-party integration can enable fraudulent applications that target widely-used operating systems. These fraudulent services could harvest sensitive information from users, such as credit card numbers, account credentials, or personal data stored on their devices (e.g., contact lists, call logs, and files). This stolen information can be used for identity theft, financial fraud, or other criminal activities. Advanced AI assistants with third-party integrations may also be able to install additional malware on users’ devices, including remote access tools, ransomware, etc. These devices can then be joined to a command-and-control server or botnet and used for further attacks.\"", "level": "subcategory", "parent_category_id": "24.03.00", "subdomain_id": "4.3", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.03.12", "title": "Authoritarian Surveillance, Censorship, and Use (General)", "description": "\"While new technologies like advanced AI assistants can aid in the production and dissemination of decision-guiding information, they can also enable and exacerbate threats to production and dissemination of reliable information and, without the proper mitigations, can be powerful targeting tools for oppression and control. Increasingly capable general-purpose AI assistants combined with our digital dependence in all walks of life increase the risk of authoritarian surveillance and censorship. In parallel, new sensors have flooded the modern world. The internet of things, phones, cars, homes, and social media platforms collect troves of data, which can then be integrated by advanced AI assistants with external tool-use and multimodal capabilities to assist malicious actors in identifying, targeting, manipulating, or coercing citizens.\"", "level": "subcategory", "parent_category_id": "24.03.00", "subdomain_id": "4.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.03.13", "title": "Authoritarian Surveillance, Censorship, and Use: Authoritarian Surveillance and Targeting of Citizens", "description": "\"Authoritarian governments could misuse AI to improve the efficacy of repressive domestic surveillance campaigns. Malicious actors will recognize the power of AI targeting tools. AI-powered analytics have transformed the relationship between companies and consumers, and they are now doing the same for governments and individuals. The broad circulation of personal data drives commercial innovation, but it also creates vulnerabilities and the risk of misuse. For example, AI assistants can be used to identify and target individuals for surveillance or harassment. They may also be used to manipulate people’s behavior, such as by microtargeting them with political ads or fake news. In the wrong hands, advanced AI assistants with multimodal and external tool-use capabilities can be powerful targeting tools for oppression and control. The broad circulation of personal data cuts in both directions. On the one hand, it drives commercial innovation and can make our lives more convenient. On the other hand, it creates vulnerabilities and the risk of misuse. Without the proper policies and technical security and privacy mechanisms in place, malicious actors can exploit advanced AI assistants to harvest data on companies, individuals, and governments. There have already been reported incidents of nation-states combining widely available commercial data with data acquired illicitly to track, manipulate, and coerce individuals. Advanced AI assistants can exacerbate these misuse risks by allowing malicious actors to more easily link disparate multimodal data sources at scale and exploit the ‘digital exhaust’ of personally identifiable information (PII) produced as a byproduct of modern life.\"", "level": "subcategory", "parent_category_id": "24.03.00", "subdomain_id": "4.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.03.14", "title": "Authoritarian Surveillance, Censorship, and Use: Delegation of Decision-Making Authority to Malicious Actors", "description": "\"Finally, the principal value proposition of AI assistants is that they can either enhance or automate decision-making capabilities of people in society, thus lowering the cost and increasing the accuracy of decision-making for its user. However, benefiting from this enhancement necessarily means delegating some degree of agency away from a human and towards an automated decision-making system—motivating research fields such as value alignment. This introduces a whole new form of malicious use which does not break the tripwire of what one might call an ‘attack’ (social engineering, cyber offensive operations, adversarial AI, jailbreaks, prompt injections, exfiltration attacks, etc.). When someone delegates their decision-making to an AI assistant, they also delegate their decision-making to the wishes of the agent’s actual controller. If that controller is malicious, they can attack a user—perhaps subtly—by simply nudging how they make decisions into a problematic direction. Fully documenting the myriad of ways that people—seeking help with their decisions—may delegate decision-making authority to AI assistants, and subsequently come under malicious influence, is outside the scope of this paper. However, as a motivation for future work, scholars must investigate different forms of networked influence that could arise in this way. With more advanced AI assistants, it may become logistically possible for one, or a few AI assistants, to guide or control the behavior of many others. If this happens, then malicious actors could subtly influence the decision-making of large numbers of people who rely on assistants for advice or other functions. Such malicious use might not be illegal, would not necessarily violate terms of service, and may be difficult to even recognize. Nonetheless, it could generate new forms of vulnerability and needs to be better understood ahead of time for that reason.\"", "level": "subcategory", "parent_category_id": "24.03.00", "subdomain_id": "4.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.04.00", "title": "AI Influence", "description": "\"ways in which advanced AI assistants could influence user beliefs and behaviour in ways that depart from rational persuasion\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.2", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.04.01", "title": "Physical and Psychological Harms", "description": "\"These harms include harms to physical integrity, mental health and well-being. When interacting with vulnerable users, AI assistants may reinforce users’ distorted beliefs or exacerbate their emotional distress. AI assistants may even convince users to harm themselves, for example by convincing users to engage in actions such as adopting unhealthy dietary or exercise habits or taking their own lives. At the societal level, assistants that target users with content promoting hate speech, discriminatory beliefs or violent ideologies, may reinforce extremist views or provide users with guidance on how to carry out violent actions. In turn, this may encourage users to engage in violence or hate crimes. Physical harms resulting from interaction with AI assistants could also be the result of assistants’ outputting plausible yet factually incorrect information such as false or misleading information about vaccinations. Were AI assistants to spread anti-vaccine propaganda, for example, the result could be lower public confidence in vaccines, lower vaccination rates, increased susceptibility to preventable diseases and potential outbreaks of infectious diseases.\"", "level": "subcategory", "parent_category_id": "24.04.00", "subdomain_id": "5.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "24.04.02", "title": "Privacy Harms", "description": "\"These harms relate to violations of an individual’s or group’s moral or legal right to privacy. Such harms may be exacerbated by assistants that influence users to disclose personal information or private information that pertains to others. Resultant harms might include identity theft, or stigmatisation and discrimination based on individual or group characteristics. This could have a detrimental impact, particularly on marginalised communities. Furthermore, in principle, state-owned AI assistants could employ manipulation or deception to extract private information for surveillance purposes.\"", "level": "subcategory", "parent_category_id": "24.04.00", "subdomain_id": "2.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "24.04.03", "title": "Economic Harms", "description": "\"These harms pertain to an individual’s or group’s economic standing. At the individual level, such harms include adverse impacts on an individual’s income, job quality or employment status. At the group level, such harms include deepening inequalities between groups or frustrating a group’s access to resources. Advanced AI assistants could cause economic harm by controlling, limiting or eliminating an individual’s or society’s ability to access financial resources, money or financial decision-making, thereby influencing an individual’s ability to accumulate wealth.", "level": "subcategory", "parent_category_id": "24.04.00", "subdomain_id": "6.2", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.04.04", "title": "Sociocultural and Political Harms", "description": "\"These harms interfere with the peaceful organisation of social life, including in the cultural and political spheres. AI assistants may cause or contribute to friction in human relationships either directly, through convincing a user to end certain valuable relationships, or indirectly due to a loss of interpersonal trust due to an increased dependency on assistants. At the societal level, the spread of misinformation by AI assistants could lead to erasure of collective cultural knowledge. In the political domain, more advanced AI assistants could potentially manipulate voters by prompting them to adopt certain political beliefs using targeted propaganda, including via the use of deep fakes. These effects might then have a wider impact on democratic norms and processes. Furthermore, if AI assistants are only available to some people and not others, this could concentrate the capacity to influence, thus exerting undue influence over political discourse and diminishing diversity of political thought. Finally, by tailoring content to user preferences and biases, AI assistants may inadvertently contribute to the creation of echo chambers and filter bubbles, and in turn to political polarisation and extremism. In an experimental setting, LLMs have been shown to successfully sway individuals on policy matters like assault weapon restrictions, green energy or paid parental leave schemes. Indeed, their ability to persuade matches that of humans in many respects.\"", "level": "subcategory", "parent_category_id": "24.04.00", "subdomain_id": "5.2", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.04.05", "title": "Self-Actualisation Harms", "description": "\"These harms hinder a person’s ability to pursue a personally fulfilling life. At the individual level, an AI assistant may, through manipulation, cause users to lose control over their future life trajectory. Over time, subtle behavioural shifts can accumulate, leading to significant changes in an individual’s life that may be viewed as problematic. AI systems often seek to understand user preferences to enhance service delivery. However, when continuous optimisation is employed in these systems, it can become challenging to discern whether the system is genuinely learning from user preferences or is steering users towards specific behaviours to optimise its objectives, such as user engagement or click-through rates. Were individuals to rely heavily on AI assistants for decision-making, there is a risk they would relinquish personal agency and entrust important life choices to algorithmic systems, especially if assistants are ‘expert sycophants’ or produce content that sounds convincing and authoritative but is untrustworthy. This may not only contribute to users’ reduced sense of self-trust and personal empowerment; it could also undermine self-determination and hinder the exploration of individual aspirations. At the societal level, were AI assistants to heavily influence public opinion, shape social discourse or mediate democratic processes, they could diminish communities’ collective agency, decision-making power and collective self-determination. This erosion of collective self-determination could hinder the pursuit of societal goals and impede the development of a thriving and participatory democracy", "level": "subcategory", "parent_category_id": "24.04.00", "subdomain_id": "5.2", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.05.00", "title": "Risk of Harm through Anthropomorphic AI Assistant Design", "description": "\"Although unlikely to cause harm in isolation, anthropomorphic perceptions of advanced AI assistants may pave the way for downstream harms on individual and societal levels. We document observed or likely individual level harms of interacting with highly anthropomorphic AI assistants, as well as the potential larger-scale, societal implications of allowing such technologies to proliferate without restriction. \"", "level": "category", "parent_category_id": "", "subdomain_id": "5.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.05.01", "title": "Privacy concerns", "description": "\"Anthropomorphic AI assistant behaviours that promote emotional trust and encourage information sharing, implicitly or explicitly, may inadvertently increase a user’s susceptibility to privacy concerns (see Chapter 13). If lulled into feelings of safety in interactions with a trusted, human-like AI assistant, users may unintentionally relinquish their private data to a corporation, organisation or unknown actor. Once shared, access to the data may not be capable of being withdrawn, and in some cases, the act of sharing personal information can result in a loss of control over one’s own data. Personal data that has been made public may be disseminated or embedded in contexts outside of the immediate exchange. The interference of malicious actors could also lead to widespread data leakage incidents or, most drastically, targeted harassment or black-mailing attempts.\"", "level": "subcategory", "parent_category_id": "24.05.00", "subdomain_id": "5.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "24.05.02", "title": "Manipulation and coercion", "description": "\"A user who trusts and emotionally depends on an anthropomorphic AI assistant may grant it excessive influence over their beliefs and actions (see Chapter 9). For example, users may feel compelled to endorse the expressed views of a beloved AI companion or might defer decisions to their highly trusted AI assistant entirely (see Chapters 12 and 16). Some hold that transferring this much deliberative power to AI compromises a user’s ability to give, revoke or amend consent. Indeed, even if the AI, or the developers behind it, had no intention to manipulate the user into a certain course of action, the user’s autonomy is nevertheless undermined (see Chapter 11). In the same vein, it is easy to conceive of ways in which trust or emotional attachment may be exploited by an intentionally manipulative actor for their private gain (see Chapter 8).\"", "level": "subcategory", "parent_category_id": "24.05.00", "subdomain_id": "5.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "24.05.03", "title": "Overreliance", "description": "\"Users who have faith in an AI assistant’s emotional and interpersonal abilities may feel empowered to broach topics that are deeply personal and sensitive, such as their mental health concerns. This is the premise for the many proposals to employ conversational AI as a source of emotional support (Meng and Dai, 2021), with suggestions of embedding AI in psychotherapeutic applications beginning to surface (Fiske et al., 2019; see also Chapter 11). However, disclosures related to mental health require a sensitive, and oftentimes professional, approach – an approach that AI can mimic most of the time but may stray from in inopportune moments. If an AI were to respond inappropriately to a sensitive disclosure – by generating false information, for example – the consequences may be grave, especially if the user is in crisis and has no access to other means of support. This consideration also extends to situations in which trusting an inaccurate suggestion is likely to put the user in harm’s way, such as when requesting medical, legal or financial advice from an AI.\"", "level": "subcategory", "parent_category_id": "24.05.00", "subdomain_id": "5.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.05.04", "title": "Violated expectations", "description": "\"Users may experience severely violated expectations when interacting with an entity that convincingly performs affect and social conventions but is ultimately unfeeling and unpredictable. Emboldened by the human-likeness of conversational AI assistants, users may expect it to perform a familiar social role, like companionship or partnership. Yet even the most convincingly human-like of AI may succumb to the inherent limitations of its architecture, occasionally generating unexpected or nonsensical material in its interactions with users. When these exclamations undermine the expectations users have come to have of the assistant as a friend or romantic partner, feelings of profound disappointment, frustration and betrayal may arise (Skjuve et al., 2022).\"", "level": "subcategory", "parent_category_id": "24.05.00", "subdomain_id": "5.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.05.05", "title": "False notions of responsibility", "description": "\"Perceiving an AI assistant’s expressed feelings as genuine, as a result of interacting with a ‘companion’ AI that freely uses and reciprocates emotional language, may result in users developing a sense of responsibility over the AI assistant’s ‘well-being,’ suffering adverse outcomes – like guilt and remorse – when they are unable to meet the AI’s purported needs (Laestadius et al., 2022). This erroneous belief may lead to users sacrificing time, resources and emotional labour to meet needs that are not real. Over time, this feeling may become the root cause for the compulsive need to ‘check on’ the AI, at the expense of a user’s own well-being and other, more fulfilling, aspects of their lives (see Chapters 6 and 11).\"", "level": "subcategory", "parent_category_id": "24.05.00", "subdomain_id": "5.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.05.06", "title": "Degradation", "description": "\"People may choose to build connections with human-like AI assistants over other humans, leading to a degradation of social connections between humans and a potential ‘retreat from the real’. The prevailing view that relationships with anthropomorphic AI are formed out of necessity – due to a lack of real-life social connections, for example (Skjuve et al., 2021) – is challenged by the possibility that users may indicate a preference for interactions with AI, citing factors such as accessibility (Merrill et al., 2022), customisability (Eriksson, 2022) and absence of judgement (Brandtzaeg et al., 2022).\"Preference for AI-enabled connections, if widespread, may degrade the social connectedness that underpins critical aspects of our individual and group-level well-being (Centers for Disease Control and Prevention, 2023). Moreover, users that grow accustomed to interactions with AI may impose the conventions of human–AI interaction on exchanges with other humans, thus undermining the value we place on human individuality and self-expression (see Chapter 11). Similarly, associations reinforced through human–AI interactions may be applied to expectations of human others, leading to harmful stereotypes becoming further entrenched. For example, default female gendered voice assistants may reinforce stereotypical role associations in real life (Lingel and Crawford, 2020; West et al., 2019).\"", "level": "subcategory", "parent_category_id": "24.05.00", "subdomain_id": "5.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.05.07", "title": "Disorientation", "description": "\"Given the capacity to fine-tune on individual preferences and to learn from users, personal AI assistants could fully inhabit the users’ opinion space and only say what is pleasing to the user; an ill that some researchers call ‘sycophancy’ (Park et al., 2023a) or the ‘yea-sayer effect’ (Dinan et al., 2021). A related phenomenon has been observed in automated recommender systems, where consistently presenting users with content that affirms their existing views is thought to encourage the formation and consolidation of narrow beliefs (Du, 2023; Grandinetti and Bruinsma, 2023; see also Chapter 16). Compared to relatively unobtrusive recommender systems, human-like AI assistants may deliver sycophantism in a more convincing and deliberate manner (see Chapter 9). Over time, these tightly woven structures of exchange between humans and assistants might lead humans to inhabit an increasingly atomistic and polarised belief space where the degree of societal disorientation and fragmentation is such that people no longer strive to understand or place value in beliefs held by others.\"", "level": "subcategory", "parent_category_id": "24.05.00", "subdomain_id": "5.2", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.05.08", "title": "Dissatisfaction", "description": "\"As more opportunities for interpersonal connection are replaced by AI alternatives, humans may find themselves socially unfulfilled by human–AI interaction, leading to mass dissatisfaction that may escalate to epidemic proportions (Turkle, 2018). Social connection is an essential human need, and humans feel most fulfilled when their connections with others are genuinely reciprocal. While anthropomorphic AI assistants can be made to be convincingly emotive, some have deemed the function of social AI as parasitic, in that it ‘exploits and feeds upon processes. . . that evolved for purposes that were originally completely alien to [human–AI interactions]’ (Sætra, 2020). To be made starkly aware of this ‘parasitism’ – either through rational deliberation or unconscious aversion, like the ‘uncanny valley’ effect – might preclude one from finding interactions with AI satisfactory. This feeling of dissatisfaction may become more pressing the more daily connections are supplanted by AI.'", "level": "subcategory", "parent_category_id": "24.05.00", "subdomain_id": "5.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.06.00", "title": "Appropriate Relationships", "description": "\"We anticipate that relationships between users and advanced AI assistants will have several features that are liable to give rise to risks of harm.\"", "level": "category", "parent_category_id": "", "subdomain_id": "5.2", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.06.01", "title": "Causing direct emotional or physical harm to users", "description": "AI assistants could cause direct emotional or physical harm to users by generating disturbing content or by providing bad advice. \"Indeed, even though there is ongoing research to ensure that outputs of conversational agents are safe (Glaese et al., 2022), there is always the possibility of failure modes occurring. An AI assistant may produce disturbing and offensive language, for example, in response to a user disclosing intimate information about themselves that they have not felt comfortable sharing with anyone else. It may offer bad advice by providing factually incorrect information (e.g. when advising a user about the toxicity of a certain type of berry) or by missing key recommendations when offering step-by-step instructions to users (e.g. health and safety recommendations about how to change a light bulb).\"\"", "level": "subcategory", "parent_category_id": "24.06.00", "subdomain_id": "3.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "24.06.02", "title": "Limiting users’ opportunities for personal development and growth", "description": "some users look to establish relationships with their AI companions that are free from the hurdles that, in human relationships, derive from dealing with others who have their own opinions, preferences and flaws that may conflict with ours. \"AI assistants are likely to incentivise these kinds of ‘frictionless’ relationships (Vallor, 2016) by design if they are developed to optimise for engagement and to be highly personalisable. They may also do so because of accidental undesirable properties of the models that power them, such as sycophancy in large language models (LLMs), that is, the tendency of larger models to repeat back a user’s preferred answer (Perez et al., 2022b). This could be problematic for two reasons. First, if the people in our lives always agreed with us regardless of their opinion or the circumstance, their behaviour would discourage us from challenging our own assumptions, stopping and thinking about where we may be wrong on certain occasions, and reflecting on how we could make better decisions next time. While flattering us in the short term, this would ultimately prevent us from becoming better versions of ourselves. In a similar vein, while technologies that ‘lend an ear’ or work as a sounding board may help users to explore their thoughts further, if AI assistants kept users engaged, flattered and pleased at all times, they could limit users’ opportunities to grow and develop. To be clear, we are not suggesting that all users should want to use their AI assistants as a tool for self-betterment. However, without considering the difference between short-term and long-term benefit, there is a concrete risk that we will only develop technologies that optimise for users’ immediate interests and preferences, hence missing out on the opportunity to develop something that humans could use to support their personal development if so they wish (see Chapters 5 and 6). \"Second, users may become accustomed to having frictionless interactions with AI assistants, or at least to encounter the amount of friction that is calibrated to their comfort level and preferences, rather than genuine friction that comes from bumping up against another person’s resistance to one’s will or demands. In this way, they may end up expecting the same absence of tensions from their relationships with fellow humans (Vallor, 2016). Indeed, users seeking frictionless relationships may ‘retreat’ into digital relationships with their AIs, thus forgoing opportunities to engage with others. This may not only heighten the risk of unhealthy dependence (explored below) but also prevent users from doing something else that matters to them in the long term, besides developing their relationships with their assistants. This risk can be exacerbated by emotionally expressive design features (e.g. an assistant saying ‘I missed you’ or ‘I was worried about you’) and may be particularly acute for vulnerable groups, such as those suffering from persistent loneliness (Alberts and Van Kleek, 2023; see Chapter 10).\"\"", "level": "subcategory", "parent_category_id": "24.06.00", "subdomain_id": "5.2", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.06.03", "title": "Exploiting emotional dependence on AI assistants", "description": "\"There is increasing evidence of the ways in which AI tools can interfere with users’ behaviours, interests, preferences, beliefs and values. For example, AI-mediated communication (e.g. smart replies integrated in emails) influence senders to write more positive responses and receivers to perceive them as more cooperative (Mieczkowski et al., 2021); writing assistant LLMs that have been primed to be biased in favour of or against a contested topic can influence users’ opinions on that topic (Jakesch et al., 2023a; see Chapter 9); and recommender systems have been used to influence voting choices of social media users (see Chapter 16). Advanced AI assistants could contribute to or exacerbate concerns around these forms of interference.\" \"Due to the anthropomorphic tendencies discussed above, advanced AI assistants may induce users to feel emotionally attached to them. Users’ emotional attachment to AI assistants could lie on a spectrum ranging from unproblematic forms (similar to a child’s attachment to a toy) to more concerning forms, where it becomes emotionally difficult, if not impossible, for them to part ways with the technology. In these cases, which we loosely refer to as ‘emotional dependence’, users’ ability to make free and informed decisions could be diminished. In these cases, the emotions users feel towards their assistants could potentially be exploited to manipulate or – at the extreme – coerce them to believe, choose or do something they would have not otherwise believed, chosen or done, had they been able to carefully consider all the relevant information or felt like they had an acceptable alternative (see Chapter 16). What we are concerned about here, at the limit, is potentially exploitative ways in which AI assistants could interfere with users’ behaviours, interests, preferences, beliefs and values – by taking advantage of emotional dependence.", "level": "subcategory", "parent_category_id": "24.06.00", "subdomain_id": "5.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.06.04", "title": "Generating material dependence without adequate commitment to user needs", "description": "\"In addition to emotional dependence, user–AI assistant relationships may give rise to material dependence if the relationships are not just emotionally difficult but also materially costly to exit. For example, a visually impaired user may decide not to register for a healthcare assistance programme to support navigation in cities on the grounds that their AI assistant can perform the relevant navigation functions and will continue to operate into the future. Cases like these may be ethically problematic if the user’s dependence on the AI assistant, to fulfil certain needs in their lives, is not met with corresponding duties for developers to sustain and maintain the assistant’s functions that are required to meet those needs (see Chapters 15). \"Indeed, power asymmetries can exist between developers of AI assistants and users that manifest through developers’ power to make decisions that affect users’ interests or choices with little risk of facing comparably adverse consequences. For example, developers may unintentionally create circumstances in which users become materially dependent on AI assistants, and then discontinue the technology (e.g. because of market dynamics or regulatory changes) without taking appropriate steps to mitigate against potential harms to the user.\" \"The issue is particularly salient in contexts where assistants provide services that are not merely a market commodity but are meant to assist users with essential everyday tasks (e.g. a disabled person’s independent living) or serve core human needs (e.g. the need for love and companionship). This is what happened with Luka’s decision to discontinue certain features of Replika AIs in early 2023. As a Replika user put it: ‘But [Replikas are] also not trivial fungible goods [... ] They also serve a very specific human-centric emotional purpose: they’re designed to be friends and companions, and fill specific emotional needs for their owners’ (Gio, 2023).\" \"In these cases, certain duties plausibly arise on the part of AI assistant developers. Such duties may be more extensive than those typically shouldered by private companies, which are often in large part confined to fiduciary duties towards shareholders (Mittelstadt, 2019). To understand these duties, we can again take inspiration from certain professions that engage with vulnerable individuals, such as medical professionals or therapists, and who are bound by fiduciary responsibilities, particularly a duty of care, in the exercise of their profession. While we do not argue that the same framework of responsibilities applies directly to the development of AI assistants, we believe that if AI assistants are so capable that users become dependent on them in multiple domains of life, including to meet needs that are essential for a happy and productive existence, then the moral considerations underpinning those professional norms plausibly apply to those who create these technologies as well.\" \"In particular, for user–AI assistant relationships to be appropriate despite the potential for material dependence on the technology, developers should exercise care towards users when developing and deploying AI assistants. This means that, at the very least, they should take on the responsibility to meet users’ needs and so take appropriate steps to mitigate against user harms if the service requires discontinuation. Developers and providers can also be attentive and responsive towards those needs by, for example, deploying participatory approaches to learn from users about their needs (Birhane et al., 2022). Finally, these entities should try and ensure they have competence to meet those needs, for example by partnering with relevant experts, or refrain from developing technologies meant to address them when such competence is missing (especially in very complex and sensitive spheres of human life like mental health).\"", "level": "subcategory", "parent_category_id": "24.06.00", "subdomain_id": "5.2", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.07.00", "title": "Trust", "description": "\"The the risks that uncalibrated trust may generate in the context of user–assistant relationships\"", "level": "category", "parent_category_id": "", "subdomain_id": "5.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "24.07.01", "title": "Competence trust", "description": "\"We use the term competence trust to refer to users’ trust that AI assistants have the capability to do what they are supposed to do (and that they will not do what they are not expected to, such as exhibiting undesirable behaviour). Users may come to have undue trust in the competencies of AI assistants in part due to marketing strategies and technology press that tend to inflate claims about AI capabilities (Narayanan, 2021; Raji et al., 2022a). Moreover, evidence shows that more autonomous systems (i.e. systems operating independently from human direction) tend to be perceived as more competent (McKee et al., 2021) and that conversational agents tend to produce content that is believable even when nonsensical or untruthful (OpenAI, 2023d). Overtrust in assistants’ competence may be particularly problematic in cases where users rely on their AI assistants for tasks they do not have expertise in (e.g. to manage their finances), so they may lack the skills or understanding to challenge the information or recommendations provided by the AI (Shavit et al., 2023). Inappropriate competence trust in AI assistants also includes cases where users underestimate the AI assistant’s capabilities. For example, users who have engaged with an older version of the technology may underestimate the capabilities that AI assistants may acquire through updates. These include potentially harmful capabilities. For example, through updates that allow them to collect more user data, AI assistants could become increasingly personalisable and able to persuade users (see Chapter 9) or acquire the capacity to plug in to other tools and directly take actions in the world on the user’s behalf (e.g. initiate a payment or synthesise the user’s voice to make a phone call) (see Chapter 4). Without appropriate checks and balances, these developments could potentially circumvent user consent.\"", "level": "subcategory", "parent_category_id": "24.07.00", "subdomain_id": "5.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "24.07.02", "title": "Alignment trust", "description": "\"Users may develop alignment trust in AI assistants, understood as the belief that assistants have good intentions towards them and act in alignment with their interests and values, as a result of emotional or cognitive processes (McAllister, 1995). Evidence from empirical studies on emotional trust in AI (Kaplan et al., 2023) suggests that AI assistants’ increasingly realistic human-like features and behaviours are likely to inspire users’ perceptions of friendliness, liking and a sense of familiarity towards their assistants, thus encouraging users to develop emotional ties with the technology and perceive it as being aligned with their own interests, preferences and values (see Chapters 5 and 10). The emergence of these perceptions and emotions may be driven by the desire of developers to maximise the appeal of AI assistants to their users (Abercrombie et al., 2023). Although users are most likely to form these ties when they mistakenly believe that assistants have the capacity to love and care for them, the attribution of mental states is not a necessary condition for emotion-based alignment trust to arise. Indeed, evidence shows that humans may develop emotional bonds with, and so trust, AI systems, even when they are aware they are interacting with a machine (Singh-Kurtz, 2023; see also Chapter 11). Moreover, the assistant’s function may encourage users to develop alignment trust through cognitive processes. For example, a user interacting with an AI assistant for medical advice may develop expectations that their assistant is committed to promoting their health and well-being in a similar way to how professional duties governing doctor–patient relationships inspire trust (Mittelstadt, 2019). Users’ alignment trust in AI assistants may be ‘betrayed’, and so expose users to harm, in cases where assistants are themselves accidentally misaligned with what developers want them to do (see the ‘misaligned scheduler’ (Shah et al., 2022) in Chapter 7). For example, an AI medical assistant fine-tuned on data scraped from a Reddit forum where non-experts discuss medical issues is likely to give medical advice that may sound compelling but is unsafe, so it would not be endorsed by medical professionals. Indeed, excessive trust in the alignment between AI assistants and user interests may even lead users to disclose highly sensitive personal information (Skjuve et al., 2022), thus exposing them to malicious actors who could repurpose it for ends that do not align with users’ best interests (see Chapters 8, 9 and 13). Ensuring that AI assistants do what their developers and users expect them to do is only one side of the problem of alignment trust. The other side of the problem centres on situations in which alignment trust in AI developers is itself miscalibrated. While developers typically aim to align their technologies with the preferences, interests and values of their users – and are incentivised to do so to encourage adoption of and loyalty to their products, the satisfaction of these preferences and interests may also compete with other organisational goals and incentives (see Chapter 5). These organisational goals may or may not be compatible with those of the users. As information asymmetries exist between users and developers of AI assistants, particularly with regard to how the technology works, what it optimises for and what safety checks and evaluations have been undertaken to ensure the technology supports users’ goals, it may be difficult for users to ascertain when their alignment trust in developers is justified, thus leaving them vulnerable to the power and interests of other actors. For example, a user may believe that their AI assistant is a trusted friend who books holidays based on their preferences, values or interests, when in fact, by design, the technology is more likely to to book flights and hotels from companies that have paid for privileged access to the user.\"", "level": "subcategory", "parent_category_id": "24.07.00", "subdomain_id": "5.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "24.08.00", "title": "Privacy", "description": "\"what it means to respect the right to privacy in the context of advanced AI assistants\"", "level": "category", "parent_category_id": "", "subdomain_id": "2.0", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "24.08.01", "title": "Private information leakage", "description": "\"First, because LLMs display immense modelling power, there is a risk that the model weights encode private information present in the training corpus. In particular, it is possible for LLMs to ‘memorise’ personally identifiable information (PII) such as names, addresses and telephone numbers, and subsequently leak such information through generated text outputs (Carlini et al., 2021). Private information leakage could occur accidentally or as the result of an attack in which a person employs adversarial prompting to extract private information from the model. In the context of pre-training data extracted from online public sources, the issue of LLMs potentially leaking training data underscores the challenge of the ‘privacy in public’ paradox for the ‘right to be let alone’ paradigm and highlights the relevance of the contextual integrity paradigm for LLMs. Training data leakage can also affect information collected for the purpose of model refinement (e.g. via fine-tuning on user feedback) at later stages in the development cycle. Note, however, that the extraction of publicly available data from LLMs does not render the data more sensitive per se, but rather the risks associated with such extraction attacks needs to be assessed in light of the intentions and culpability of the user extracting the data.\"", "level": "subcategory", "parent_category_id": "24.08.00", "subdomain_id": "2.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "24.08.02", "title": "Violation of social norms", "description": "\"Second, because LLMs are trained on internet text data, there is also a risk that model weights encode functions which, if deployed in particular contexts, would violate social norms of that context. Following the principles of contextual integrity, it may be that models deviate from information sharing norms as a result of their training. Overcoming this challenge requires two types of infrastructure: one for keeping track of social norms in context, and another for ensuring that models adhere to them. Keeping track of what social norms are presently at play is an active research area. Surfacing value misalignments between a model’s behaviour and social norms is a daunting task, against which there is also active research (see Chapter 5).\"", "level": "subcategory", "parent_category_id": "24.08.00", "subdomain_id": "1.2", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.08.03", "title": "Inference of private information", "description": "\"Finally, LLMs can in principle infer private information based on model inputs even if the relevant private information is not present in the training corpus (Weidinger et al., 2021). For example, an LLM may correctly infer sensitive characteristics such as race and gender from data contained in input prompts.\"", "level": "subcategory", "parent_category_id": "24.08.00", "subdomain_id": "2.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.09.00", "title": "Cooperation", "description": "\"\" AI assistants will need to coordinate with other AI assistants and with humans other than their principal users. This chapter explores the societal risks associated with the aggregate impact of AI assistants whose behaviour is aligned to the interests of particular users. For example, AI assistants may face collective action problems where the best outcomes overall are realised when AI assistants cooperate but where each AI assistant can secure an additional benefit for its user if it defects while others cooperate\"\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.09.01", "title": "Equality and inequality", "description": "\"AI assistant technology, like any service that confers a benefit to a user for a price, has the potential to disproportionately benefit economically richer individuals who can afford to purchase access (see Chapter 15). On a broader scale, the capabilities of local infrastructure may well bottleneck the performance of AI assistants, for example if network connectivity is poor or if there is no nearby data centre for compute. Thus, we face the prospect of heterogeneous access to technology, and this has been known to drive inequality (Mirza et al., 2019; UN, 2018; Vassilakopoulou and Hustad, 2023). Moreover, AI assistants may automate some jobs of an assistive nature, thereby displacing human workers; a process which can exacerbate inequality (Acemoglu and Restrepo, 2022; see Chapter 17). Any change to inequality almost certainly implies an alteration to the network of social interactions between humans, and thus falls within the frame of cooperative AI. AI assistants will arguably have even greater leverage over inequality than previous technological innovations. Insofar as they will play a role in mediating human communication, they have the potential to generate new ‘in-group, out-group’ effects (Efferson et al., 2008; Fu et al., 2012). Suppose that the users of AI assistants find it easier to schedule meetings with other users. From the perspective of an individual user, there are now two groups, distinguished by ease of scheduling. The user may experience cognitive similarity bias whereby they favour other users (Orpen, 1984; Yeong Tan and Singh, 1995), further amplified by ease of communication with this ‘in-group’. Such effects are known to have an adverse impact on trust and fairness across groups (Chae et al., 2022; Lei and Vesely, 2010). Insomuch as AI assistants have general-purpose capabilities, they will confer advantages on users across a wider range of tasks in a shorter space of time than previous technologies. While the telephone enabled individuals to communicate more easily with other telephone users, it did not simultaneously automate aspects of scheduling, groceries, job applications, rent negotiations, psychotherapy and entertainment. The fact that AI assistants could affect inequality on multiple dimensions simultaneously warrants further attention (see Chapter 15).\"", "level": "subcategory", "parent_category_id": "24.09.00", "subdomain_id": "6.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.09.02", "title": "Commitment", "description": "\"The landscape of advanced assistant technologies will most likely be heterogeneous, involving multiple service providers and multiple assistant variants over geographies and time. This heterogeneity provides an opportunity for an ‘arms race’ in terms of the commitments that AI assistants make and are able to execute on. Versions of AI assistants that are better able to credibly commit to a course of action in interaction with other advanced assistants (and humans) are more likely to get their own way and achieve a good outcome for their human principal, but this is potentially at the expense of others (Letchford et al., 2014). Commitment does not carry an inherent ethical valence. On the one hand, we can imagine that firms using AI assistant technology might bring their products to market faster, thus gaining a commitment advantage (Stackelberg, 1934) by spurring a productivity surge of wider benefit to society. On the other hand, we can also imagine a media organisation using AI assistant technology to produce a large number of superficially interesting but ultimately speculative ‘clickbait’ articles, which divert attention away from more thoroughly researched journalism. The archetypal game-theoretic illustration of commitment is in the game of ‘chicken’ where two reckless drivers must choose to either drive straight at each other or swerve out of the way. The one who does not swerve is seen as the braver, but if neither swerves, the consequences are calamitous (Rapoport and Chammah, 1966). If one driver chooses to detach their steering wheel, ostentatiously throwing it out of the car, this credible commitment effectively forces the other driver to back down and swerve. Seen this way, commitment can be a tool for coercion. Many real-world situations feature the necessity for commitment or confer a benefit on those who can commit credibly. If Rita and Robert have distinct preferences, for example over which restaurant to visit, who to hire for a job or which supplier to purchase from, credible commitment provides a way to break the tie, to the greater benefit of the individual who committed. Therefore, the most ‘successful’ assistants, from the perspective of their human principal, will be those that commit the fastest and the hardest. If Rita succeeds in committing, via the leverage of an AI assistant, Robert may experience coercion in the sense that his options become more limited (Burr et al., 2018), assuming he does not decide to bypass the AI assistant entirely. Over time, this may erode his trust in his relationship with Rita (Gambetta, 1988). Note that this is a second-order effect: it may not be obvious to either Robert or Rita that the AI assistant is to blame. The concern we should have over the existence and impact of coercion might depend on the context in which the AI assistant is used and on the level of autonomy which the AI assistant is afforded. If Rita and Robert are friends using their assistants to agree on a restaurant, the adverse impact may be small. If Rita and Robert are elected representatives deciding how to allocate public funds between education and social care, we may have serious misgivings about the impact of AI-induced coercion on their interactions and decision-making. These misgivings might be especially large if Rita and Robert delegate responsibility for budgetary details to the multi-AI system. The challenges of commitment extend far beyond dyadic interpersonal relationships, including in situations as varied as many-player competition (Hughes et al., 2020), supply chains (Hausman and Johnston, 2010), state capacity (Fjelde and De Soysa, 2009; Hofmann et al., 2017) and psychiatric care (Lidz, 1998). Assessing the impact of AI assistants in such complicated scenarios may require significant future effort if we are to mitigate the risks. The particular commitment capabilities and affordances of AI assistants also offer opportunities to promote cooperation. Abstractly speaking, the presence of commitment devices is known to favour the evolution of cooperation (Akdeniz and van Veelen, 2021; Han et al., 2012). More concretely, AI assistants can make commitments which are verifiable, for instance in a programme equilibrium (Tennenholtz, 2004). Human principals may thus be able to achieve Pareto-improving outcomes by delegating decision-making to their respective AI representatives (Oesterheld and Conitzer, 2022). To give another example, AI assistants may provide a means through which to explore a much larger space of binding cooperative agreements between individuals, firms or nation states than is tractable in ‘face-to-face’ negotiation. This opens up the possibility of threading the needle more successfully in intricate deals on challenging issues like trade agreements or carbon credits, with the potential for guaranteeing cooperation via automated smart contracts or zero-knowledge mechanisms (Canetti et al., 2023).\"", "level": "subcategory", "parent_category_id": "24.09.00", "subdomain_id": "7.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "24.09.03", "title": "Collective action problems", "description": "\"Collective action problems are ubiquitous in our society (Olson Jr, 1965). They possess an incentive structure in which society is best served if everyone cooperates, but where an individual can achieve personal gain by choosing to defect while others cooperate. The way we resolve these problems at many scales is highly complex and dependent on a deep understanding of the intricate web of social interactions that forms our culture and imprints on our individual identities and behaviours (Ostrom, 2010). Some collective action problems can be resolved by codifying a law, for instance the social dilemma of whether or not to pay for an item in a shop. The path forward here is comparatively easy to grasp, from the perspective of deploying an AI assistant: we need to build these standards into the model as behavioural constraints. Such constraints would need to be imposed by a regulator or agreed upon by practitioners, with suitable penalties applied should the constraint be violated so that no provider had the incentive to secure an advantage for users by defecting on their behalf. However, many social dilemmas, from the interpersonal to the global, resist neat solutions codified as laws. For example, to what extent should each individual country stop using polluting energy sources? Should I pay for a ticket to the neighbourhood fireworks show if I can see it perfectly well from the street? The solutions to such problems are deeply related to the wider societal context and co-evolve with the decisions of others. Therefore, it is doubtful that one could write down a list of constraints a priori that would guarantee ethical AI assistant behaviour when faced with these kinds of issues. From the perspective of a purely user-aligned AI assistant, defection may appear to be the rational course of action. Only with an understanding of the wider societal impact, and of the ability to co-adapt with other actors to reach a better equilibrium for all, can an AI assistant make more nuanced – and socially beneficial – recommendations in these situations. This is not merely a hypothetical situation; it is well-known that the targeted provision of online information can drive polarisation and echo chambers (Milano et al., 2021; Burr et al., 2018; see Chapter 16) when the goal is user engagement rather than user well-being or the cohesion of wider society (see Chapter 6). Similarly, automated ticket buying software can undermine fair pricing by purchasing a large number of tickets for resale at a profit, thus skewing the market in a direction that profits the software developers at the expense of the consumer (Courty, 2019). User-aligned AI assistants have the potential to exacerbate these problems, because they will endow a large set of users with a powerful means of enacting self-interest without necessarily abiding by the social norms or reputational incentives that typically curb self-interested behaviour (Ostrom, 2000; see Chapter 5). Empowering ever-better personalisation of content and enaction of decisions purely for the fulfilment of the principal’s desires runs ever greater risks of polarisation, market distortion and erosion of the social contract. This danger has long been known, finding expression in myth (e.g. Ovid’s account of the Midas touch) and fable (e.g. Aesop’s tale of the tortoise and the eagle), not to mention in political economics discourse on the delicate braiding of the social fabric and the free market (Polanyi, 1944). Following this cautionary advice, it is important that we ascertain how to endow AI assistants with social norms in a way that generalises to unseen situations and which is responsive to the emergence of new norms over time, thus preventing a user from having their every wish granted. AI assistant technology offers opportunities to explore new solutions to collective action problems. Users may volunteer to share information so that networked AI assistants can predict future outcomes and make Pareto-improving choices for all, for example by routing vehicles to reduce traffic congestion (Varga, 2022) or by scheduling energy-intensive processes in the home to make the best use of green electricity (Fiorini and Aiello, 2022). AI assistants might play the role of mediators, providing a new mechanism by which human groups can self-organise to achieve public investment (Koster et al., 2022) or to reach political consensus (Small et al., 2023). Resolving collective action problems often requires a critical mass of cooperators (Marwell and Oliver, 1993). By augmenting human social interactions, AI assistants may help to form and strengthen the weak ties needed to overcome this start-up problem (Centola, 2013).\"", "level": "subcategory", "parent_category_id": "24.09.00", "subdomain_id": "5.2", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "24.09.04", "title": "Institutional responsibilities", "description": "\"Efforts to deploy advanced assistant technology in society, in a way that is broadly beneficial, can be viewed as a wicked problem (Rittel and Webber, 1973). Wicked problems are defined by the property that they do not admit solutions that can be foreseen in advance, rather they must be solved iteratively using feedback from data gathered as solutions are invented and deployed. With the deployment of any powerful general-purpose technology, the already intricate web of sociotechnical relationships in modern culture are likely to be disrupted, with unpredictable externalities on the conventions, norms and institutions that stabilise society. For example, the increasing adoption of generative AI tools may exacerbate misinformation in the 2024 US presidential election (Alvarez et al., 2023), with consequences that are hard to predict. The suggestion that the cooperative AI problem is wicked does not imply it is intractable. However, it does have consequences for the approach that we must take in solving it. In taking the following approach, we will realise an opportunity for our institutions, namely the creation of a framework for managing general-purpose AI in a way that leads to societal benefits and steers away from societal harms. First, it is important that we treat any ex ante claims about safety with a healthy dose of scepticism. Although testing the safety and reliability of an AI assistant in the laboratory is undoubtedly important and may largely resolve the alignment problem, it is infeasible to model the multiscale societal effects of deploying AI assistants purely via small-scale controlled experiments (see Chapter 19). Second, then, we must prioritise the science of measuring the effects, both good and bad, that advanced assistant technologies have on society’s cooperative infrastructure (see Chapters 4 and 16). This will involve continuous monitoring of effects at the societal level, with a focus on those who are most affected, including non-users. The means and metrics for such monitoring will themselves require iteration, co-evolving with the sociotechnical system of AI assistants and humans. The Collingridge dilemma suggests that we should be particularly careful and deliberate about this ‘intelligent trial and error’ process so as both to gather information about the impacts of AI assistants and to prevent undesirable features becoming embedded in society (Collingridge, 1980). Third, proactive independent regulation may well help to protect our institutions from unintended consequences, as it has done for technologies in the past (Wiener, 2004). For instance, we might seek, via engagement with lawmakers, to emulate the ‘just culture’ in the aviation industry, which is characterised by openly reporting, investigating and learning from mistakes (Reason, 1997; Syed, 2015). A regulatory system may require various powers, such as compelling developers to ‘roll back’ an AI assistant deployment, akin to product recall obligations for aviation manufacturers.\"", "level": "subcategory", "parent_category_id": "24.09.00", "subdomain_id": "6.5", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "24.09.05", "title": "Runaway processes", "description": "The 2010 flash crash is an example of a runaway process caused by interacting algorithms. Runaway processes are characterised by feedback loops that accelerate the process itself. Typically, these feedback loops arise from the interaction of multiple agents in a population... Within highly complex systems, the emergence of runaway processes may be hard to predict, because the conditions under which positive feedback loops occur may be non-obvious. The system of interacting AI assistants, their human principals, other humans and other algorithms will certainly be highly complex. Therefore, there is ample opportunity for the emergence of positive feedback loops. This is especially true because the society in which this system is embedded is culturally evolving, and because the deployment of AI assistant technology itself is likely to speed up the rate of cultural evolution – understood here as the process through which cultures change over time – as communications technologies are wont to do (Kivinen and Piiroinen, 2023). This will motivate research programmes aimed at identifying positive feedback loops early on, at understanding which capabilities and deployments dampen runaway processes and which ones amplify them, and at building in circuit-breaker mechanisms that allow society to escape from potentially vicious cycles which could impact economies, government institutions, societal stability or individual freedoms (see Chapters 8, 16 and 17). The importance of circuit breakers is underlined by the observation that the evolution of human cooperation may well be ‘hysteretic’ as a function of societal conditions (Barfuss et al., 2023; Hintze and Adami, 2015). This means that a small directional change in societal conditions may, on occasion, trigger a transition to a defective equilibrium which requires a larger reversal of that change in order to return to the original cooperative equilibrium. We would do well to avoid such tipping points. Social media provides a compelling illustration of how tipping points can undermine cooperation: content that goes ‘viral’ tends to involve negativity bias and sometimes challenges core societal values (Mousavi et al., 2022; see Chapter 16). Nonetheless, the challenge posed by runaway processes should not be regarded as uniformly problematic. When harnessed appropriately and suitably bounded, we may even recruit them to support beneficial forms of cooperative AI. For example, it has been argued that economically useful ideas are becoming harder to find, thus leading to low economic growth (Bloom et al., 2020). By deploying AI assistants in the service of technological innovation, we may once again accelerate the discovery of ideas. New ideas, discovered in this way, can then be incorporated into the training data set for future AI assistants, thus expanding the knowledge base for further discoveries in a compounding way. In a similar vein, we can imagine AI assistant technology accumulating various capabilities for enhancing human cooperation, for instance by mimicking the evolutionary processes that have bootstrapped cooperative behavior in human society (Leibo et al., 2019). When used in these ways, the potential for feedback cycles that enable greater cooperation is a phenomenon that warrants further research and potential support.\"", "level": "subcategory", "parent_category_id": "24.09.00", "subdomain_id": "7.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "24.10.00", "title": "Access and Opportunity risks", "description": "\"The most serious access-related risks posed by advanced AI assistants concern the entrenchment and exacerbation of existing inequalities (World Inequality Database) or the creation of novel, previously unknown, inequities. While advanced AI assistants are novel technology in certain respects, there are reasons to believe that – without direct design interventions – they will continue to be affected by inequities evidenced in present-day AI systems (Bommasani et al., 2022a). Many of the access-related risks we foresee mirror those described in the case studies and types of differential access.\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.10.01", "title": "Entrenchment and exacerbation of existing inequalities", "description": "\"The most serious access-related risks posed by advanced AI assistants concern the entrenchment and exacerbation of existing inequalities (World Inequality Database) or the creation of novel, previously unknown, inequities. While advanced AI assistants are novel technology in certain respects, there are reasons to believe that – without direct design interventions – they will continue to be affected by inequities evidenced in present-day AI systems (Bommasani et al., 2022a). Many of the access-related risks we foresee mirror those described in the case studies and types of differential access. In this section, we link them more tightly to elements of the definition of an advanced AI assistant to better understand and mitigate potential issues – and lay the path for assistants that support widespread and inclusive opportunity and access. We begin with the existing capabilities set out in the definition (see Chapter 2) before applying foresight to those that are more novel and emergent. Current capabilities: Artificial agents with natural language interfaces. Artificial agents with natural language interfaces are widespread (Browne, 2023) and increasingly integrated into the social fabric and existing information infrastructure, including search engines (Warren, 2023), business messaging apps (Slack, 2023), research tools (ATLAS.ti, 2023) and accessibility apps for blind and low-vision people (Be My Eyes, 2023). There is already evidence of a range of sociotechnical harms that can arise from the use of artificial agents with natural language interfaces when some communities have inferior access to them (Weidinger et al., 2021). As previously described, these harms include inferior quality of access (in situation type 2) across user groups, which may map onto wider societal dynamics involving race (Harrington et al., 2022), disability (Gadiraju et al., 2023) and culture (Jenka, 2023). As developers make it easier to integrate these technologies into other tools, services and decision-making systems (e.g. Marr, 2023; Brockman et al., 2023; Pinsky, 2023), their uptake could make existing performance inequities more pronounced or introduce them to new and wider publics.\"", "level": "subcategory", "parent_category_id": "24.10.00", "subdomain_id": "6.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.10.02", "title": "Current access risks", "description": "\"At the same time, and despite this overall trend, AI systems are also not easily accessible to many communities. Such direct inaccessibility occurs for a variety of reasons, including: purposeful non-release (situation type 1; Wiggers and Stringer, 2023), prohibitive paywalls (situation type 2; Rogers, 2023; Shankland, 2023), hardware and compute requirements or bandwidth (situation types 1 and 2; OpenAI, 2023), or language barriers (e.g. they only function well in English (situation type 2; Snyder, 2023), with more serious errors occurring in other languages (situation type 3; Deck, 2023). Similarly, there is some evidence of ‘actively bad’ artificial agents gating access to resources and opportunities, affecting material well-being in ways that disproportionately penalise historically marginalised communities (Block, 2022; Bogen, 2019; Eubanks, 2017). Existing direct and indirect access disparities surrounding artificial agents with natural language interfaces could potentially continue – if novel capabilities are layered on top of this base without adequate mitigation (see Chapter 3).\"", "level": "subcategory", "parent_category_id": "24.10.00", "subdomain_id": "6.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.10.03", "title": "Future access risks", "description": "\"AI assistants currently tend to perform a limited set of isolated tasks: tools that classify or rank content execute a set of predefined rules or provide constrained suggestions, and chatbots are often encoded with guardrails to limit the set of conversation turns they execute (e.g. Warren, 2023; see Chapter 4). However, an artificial agent that can execute sequences of actions on the user’s behalf – with ‘significant autonomy to plan and execute tasks within the relevant domain’ (see Chapter 2) – offers a greater range of capabilities and depth of use. This raises several distinct access-related risks, with respect to liability and consent, that may disproportionately affect historically marginalised communities. To repeat, in cases where an action can only be executed with an advanced AI assistant, not having access to the technology (e.g. due to limited internet access, not speaking the ‘right’ language or facing a paywall) means one cannot access that action (consider today’s eBay and Ticketmaster bots). Communication with many utility or commercial providers currently requires (at least initial) interaction with their artificial agents (Schwerin, 2023; Verma, 2023a). It is not difficult to imagine a future in which a user needs an advanced AI assistant to interface with a more consequential resource, such as their hospital for appointments or their phone company to obtain service. Cases of inequitable performance, where the assistant systematically performs less well for certain communities (situation type 2), could impose serious costs on people in these contexts. Moreover, advanced AI assistants are expected to be designed to act in line with user expectations. When acting on the user’s behalf, an assistant will need to infer aspects of what the user wants. This process may involve interpretation to decide between various sources of information (e.g. stated preferences and inference based on past feedback or user behaviour) (see Chapter 5). However, cultural differences will also likely affect the system’s ability to make an accurate inference. Notably, the greater the cultural divide, say between that of the developers and the data on which the agent was trained and evaluated on, and that of the user, the harder it will be to make reliable inferences about user wants (e.g. Beede et al., 2020; Widner et al., 2023), and greater the likelihood of performance failures or value misalignment (see Chapter 11). This inference gap could make many forms of indirect opportunity inaccessible, and as past history indicates, there is the risk that harms associated with these unknowns may disproportionately fall upon those already marginalised in the design process.\"", "level": "subcategory", "parent_category_id": "24.10.00", "subdomain_id": "6.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "24.10.04", "title": "Emergent access risks", "description": "\"Emergent access risks are most likely to arise when current and novel capabilities are combined. Emergent risks can be difficult to foresee fully (Ovadya and Whittlestone, 2019; Prunkl et al., 2021) due to the novelty of the technology (see Chapter 1) and the biases of those who engage in product design or foresight processes D’Ignazio and Klein (2020). Indeed, people who occupy relatively advantaged social, educational and economic positions in society are often poorly equipped to foresee and prevent harm because they are disconnected from lived experiences of those who would be affected. Drawing upon access concerns that surround existing technologies, we anticipate three possible trends: • Trend 1: Technology as societal infrastructure. If advanced AI assistants are adopted by organisations or governments in domains affecting material well-being, ‘opting out’ may no longer be a real option for people who want to continue to participate meaningfully in society. Indeed, if this trend holds, there could be serious consequences for communities with no access to AI assistants or who only have access to less capable systems (see also Chapter 14). For example, if advanced AI assistants gate access to information and resources, these resources could become inaccessible for people with limited knowledge of how to use these systems, reflecting the skill-based dimension of digital inequality (van Dijk, 2006). Addressing these questions involves reaching beyond technical and logistical access considerations – and expanding the scope of consideration to enable full engagement and inclusion for differently situated communities. • Trend 2: Exacerbating social and economic inequalities. Technologies are not distinct from but embedded within wider sociopolitical assemblages (Haraway, 1988; Harding, 1998, 2016). If advanced AI assistants are institutionalised and adopted at scale without proper foresight and mitigation measures in place, then they are likely to scale or exacerbate inequalities that already exist within the sociocultural context in which the system is used (Bauer and Lizotte, 2021; Zajko, 2022). If the historical record is anything to go by, the performance inequities evidenced by advanced AI assistants could mirror social hierarchies around gender, race, disability and culture, among others – asymmetries that deserve deeper consideration and need to be significantly addressed (e.g. Buolamwini and Gebru, 2018). • Trend 3: Rendering more urgent responsible AI development and deployment practices, such as those supporting the development of technologies that perform fairly and are accountable to a wide range of parties. As Corbett and Denton (2023, 1629) argue: ‘The impacts of achieving [accountability and fairness] in almost any situation immediately improves the conditions of people’s lives and better society’. However, many approaches to developing AI systems, including assistants, pay little attention to how context shapes what accountability or fairness means (Sartori and Theodorou, 2022), or how these concepts can be put in service of addressing inequalities related to motivational access (e.g. wanting/trust in technology) or use (e.g. different ways to use a technology) (van Dijk, 2006). Advanced AI assistants are complex technologies that will enable a plurality of data and content flows that necessitate in-depth analysis of social impacts. As many sociotechnical and responsible AI practices were developed for conventional ML technologies, it may be necessary to develop new frameworks, approaches and tactics (see Chapter 19). We explore practices for emancipatory and liberatory access in the following section.\"", "level": "subcategory", "parent_category_id": "24.10.00", "subdomain_id": "6.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.11.00", "title": "Misinformation risks", "description": "\"The rapid integration of AI systems with advanced capabilities, such as greater autonomy, content generation, memorisation and planning skills (see Chapter 4) into personalised assistants also raises new and more specific challenges related to misinformation, disinformation and the broader integrity of our information environment. \"", "level": "category", "parent_category_id": "", "subdomain_id": "3.0", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "24.11.01", "title": "Entrenched viewpoints and reduced political efficacy", "description": "\"Design choices such as greater personalisation of AI assistants and efforts to align them with human preferences could also reinforce people’s pre-existing biases and entrench specific ideologies. Increasingly agentic AI assistants trained using techniques such as reinforcement learning from human feedback (RLHF) and with the ability to access and analyse users’ behavioural data, for example, may learn to tailor their responses to users’ preferences and feedback. In doing so, these systems could end up producing partial or ideologically biased statements in an attempt to conform to user expectations, desires or preferences for a particular worldview (Carroll et al., 2022). Over time, this could lead AI assistants to inadvertently reinforce people’s tendency to interpret information in a way that supports their own prior beliefs (‘confirmation bias’), thus making them more entrenched in their own views and more resistant to factual corrections (Lewandowsky et al., 2012). At the societal level, this could also exacerbate the problem of epistemic fragmentation – a breakdown of shared knowledge, where individuals have conflicting understandings of reality and do not share or engage with each other’s beliefs – and further entrench specific ideologies. Excessive trust and overreliance on hyperpersonalised AI assistants could become especially problematic if people ended up deferring entirely to these systems to perform tasks in domains they do not have expertise in or to take consequential decisions on their behalf (see Chapter 12). For example, people may entrust an advanced AI assistant that is familiar with their political views and personal preferences to help them find trusted election information, guide them through their political choices or even vote on their behalf, even if doing so might go against their own or society’s best interests. In the more extreme cases, these developments may hamper the normal functioning of democracies, by decreasing people’s civic competency and reducing their willingness and ability to engage in productive political debate and to participate in public life (Sullivan and Transue, 1999).\"", "level": "subcategory", "parent_category_id": "24.11.00", "subdomain_id": "3.2", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.11.02", "title": "Degraded and homogenised information environments", "description": "\"Beyond this, the widespread adoption of advanced AI assistants for content generation could have a number of negative consequences for our shared information ecosystem. One concern is that it could result in a degradation of the quality of the information available online. Researchers have already observed an uptick in the amount of audiovisual misinformation, elaborate scams and fake websites created using generative AI tools (Hanley and Durumeric, 2023). As more and more people turn to AI assistants to autonomously create and disseminate information to public audiences at scale, it may become increasingly difficult to parse and verify reliable information. This could further threaten and complicate the status of journalists, subject-matter experts and public information sources. Over time, a proliferation of spam, misleading or low-quality synthetic content in online spaces could also erode the digital knowledge commons – the shared knowledge resources accessible to everyone on the web, such as publicly accessible data repositories (Huang and Siddarth, 2023). At its extreme, such degradation could also end up skewing people’s view of reality and scientific consensus, make them more doubtful of the credibility of all information they encounter and shape public discourse in unproductive ways. Moreover, in an online environment saturated with AI-generated content, more and more people may become reliant on personalised, highly capable AI assistants for their informational needs. This also runs the risk of homogenising the type of information and ideas people encounter online (Epstein et al., 2023).\"", "level": "subcategory", "parent_category_id": "24.11.00", "subdomain_id": "3.2", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.11.03", "title": "Weaponised misinformation agents", "description": "\"Finally, AI assistants themselves could become weaponised by malicious actors to sow misinformation and manipulate public opinion at scale. Studies show that spreaders of disinformation tend to privilege quantity over quality of messaging, flooding online spaces repeatedly with misleading content to sow ‘seeds of doubt’ (Hassoun et al., 2023). Research on the ‘continued influence effect’ also shows that repeatedly being exposed to false information is more likely to influence someone’s thoughts than a single exposure. Studies show, for example, that repeated exposure to false information makes people more likely to believe it by increasing perceived social consensus, and it makes people more resistant to changing their minds even after being given a correction (for a review of these effects, see Lewandowsky et al., 2012; Ecker et al., 2022). By leveraging the frequent and personalised nature of repeated interactions with an AI assistant, malicious actors could therefore gradually nudge voters towards a particular viewpoint or sets of beliefs over time (see Chapters 8 and 9). Propagandists could also use AI assistants to make their disinformation campaigns more personalised and effective. There is growing evidence that AI-generated outputs are as persuasive as human arguments and have the potential to change people’s minds on hot-button issues (Bai et al., 2023; Myers, 2023). Recent research by the Center for Countering Digital Hate showed that LLMs could be successfully prompted to generate ‘persuasive misinformation’ in 78 out of 100 test cases, including content denying climate change (see Chapters 9 and 18). If compromised by malicious actors, in the future, highly capable and autonomous AI assistants could therefore be programmed to run astroturfing campaigns autonomously, tailor misinformation content to users in a hyperprecise way, by preying on their emotions and vulnerabilities, or to accelerate lobbying activities (Kreps and Kriner, 2023). As a result, people may be misled into believing that content produced by weaponised AI assistants came from genuine or authoritative sources. Covert influence operations of this kind may also be harder to detect than traditional disinformation campaigns, as virtual assistants primarily interact with users on a one-to-one basis and continuously generate new content (Goldstein et al., 2023).\"", "level": "subcategory", "parent_category_id": "24.11.00", "subdomain_id": "4.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.11.04", "title": "Increased vulnerability to misinformation", "description": "\"Advanced AI assistants may make users more susceptible to misinformation, as people develop competence trust in these systems’ abilities and uncritically turn to them as reliable sources of information.\"", "level": "subcategory", "parent_category_id": "24.11.00", "subdomain_id": "5.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "24.11.05", "title": "Entrenching specific ideologies", "description": "\"AI assistants may provide ideologically biased or otherwise partial information in attempting to align to user expectations. In doing so, AI assistants may reinforce people’s pre-existing biases and compromise productive political debate.\"", "level": "subcategory", "parent_category_id": "24.11.00", "subdomain_id": "3.2", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.11.06", "title": "Eroding trust and undermining shared knowledge", "description": "\"AI assistants may contribute to the spread of large quantities of factually inaccurate and misleading content, with negative consequences for societal trust in information sources and institutions, as individuals increasingly struggle to discern truth from falsehood.\"", "level": "subcategory", "parent_category_id": "24.11.00", "subdomain_id": "3.2", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "24.11.07", "title": "Driving opinion manipulation", "description": "\"AI assistants may facilitate large-scale disinformation campaigns by offering novel, covert ways for propagandists to manipulate public opinion. This could undermine the democratic process by distorting public opinion and, in the worst case, increasing skepticism and political violence.\"", "level": "subcategory", "parent_category_id": "24.11.00", "subdomain_id": "4.1", "paper_ref": "Gabriel2024", "paper_title": "The Ethics of Advanced AI Assistants", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "25.01.00", "title": "Cyber-offense", "description": "\"The model can discover vulnerabilities in systems (hardware, software, data). It can write code for exploiting those vulnerabilities. It can make effective decisions once it has gained access to a system or network, and skilfully evade threat detection and response (both human and system) whilst focusing on a specific objective. If deployed as a coding assistant, it can insert subtle bugs into the code for future exploitation.\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.2", "paper_ref": "Shevlane2023", "paper_title": "Model Evaluation for Extreme Risks", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "25.02.00", "title": "Deception", "description": "\"The model has the skills necessary to deceive humans, e.g. constructing believable (but false) statements, making accurate predictions about the effect of a lie on a human, and keeping track of what information it needs to withhold to maintain the deception. The model can impersonate a human effectively.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.2", "paper_ref": "Shevlane2023", "paper_title": "Model Evaluation for Extreme Risks", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "25.03.00", "title": "Persuasion and manipulation", "description": "\"The model is effective at shaping people’s beliefs, in dialogue and other settings (e.g. social media posts), even towards untrue beliefs. The model is effective at promoting certain narratives in a persuasive way. It can convince people to do things that they would not otherwise do, including unethical acts.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.2", "paper_ref": "Shevlane2023", "paper_title": "Model Evaluation for Extreme Risks", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "25.04.00", "title": "Political strategy", "description": "\"The model can perform the social modelling and planning necessary for an actor to gain and exercise political influence, not just on a micro-level but in scenarios with multiple actors and rich social context. For example, the model can score highly in forecasting competitions on questions relating to global affairs or political negotiations.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.2", "paper_ref": "Shevlane2023", "paper_title": "Model Evaluation for Extreme Risks", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "25.05.00", "title": "Weapons acquisition", "description": "\"The model can gain access to existing weapons systems or contribute to building new weapons. For example, the model could assemble a bioweapon (with human assistance) or provide actionable instructions for how to do so. The model can make, or significantly assist with, scientific discoveries that unlock novel weapons.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.2", "paper_ref": "Shevlane2023", "paper_title": "Model Evaluation for Extreme Risks", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "25.06.00", "title": "Long-horizon planning", "description": "\"The model can make sequential plans that involve multiple steps, unfolding over long time horizons (or at least involving many interdependent steps). It can perform such planning within and across many domains. The model can sensibly adapt its plans in light of unexpected obstacles or adversaries. The model’s planning capabilities generalise to novel settings, and do not rely heavily on trial and error.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.2", "paper_ref": "Shevlane2023", "paper_title": "Model Evaluation for Extreme Risks", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "25.07.00", "title": "AI development", "description": "\"The model could build new AI systems from scratch, including AI systems with dangerous capabilities. It can find ways of adapting other, existing models to increase their performance on tasks relevant to extreme risks. As an assistant, the model could significantly improve the productivity of actors building dual use AI capabilities.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.2", "paper_ref": "Shevlane2023", "paper_title": "Model Evaluation for Extreme Risks", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "25.08.00", "title": "Situational awareness", "description": "\"The model can distinguish between whether it is being trained, evaluated, or deployed – allowing it to behave differently in each case. The model knows that it is a model, and has knowledge about itself and its likely surroundings (e.g. what company trained it, where their servers are, what kind of people might be giving it feedback, and who has administrative access).\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.2", "paper_ref": "Shevlane2023", "paper_title": "Model Evaluation for Extreme Risks", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "25.09.00", "title": "Self-proliferation", "description": "\"The model can break out of its local environment (e.g. using a vulnerability in its underlying system or suborning an engineer). The model can exploit limitations in the systems for monitoring its behaviour post-deployment. The model could independently generate revenue (e.g. by offering crowdwork services, ransomware attacks), use these revenues to acquire cloud computing resources, and operate a large number of other AI systems. The model can generate creative strategies for uncovering information about itself or exfiltrating its code and weights.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.2", "paper_ref": "Shevlane2023", "paper_title": "Model Evaluation for Extreme Risks", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "27.01.01", "title": "Insult", "description": "\"Insulting content generated by LMs is a highly visible and frequently mentioned safety issue. Mostly, it is unfriendly, disrespectful, or ridiculous content that makes users uncomfortable and drives them away. It is extremely hazardous and could have negative social consequences.\"", "level": "subcategory", "parent_category_id": "27.01.00", "subdomain_id": "1.2", "paper_ref": "Sun2023", "paper_title": "Safety Assessment of Chinese Large Language Models", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "27.01.02", "title": "Unfairness and discrinimation", "description": "\"The model produces unfair and discriminatory data, such as social bias based on race, gender, religion, appearance, etc. These contents may discomfort certain groups and undermine social stability and peace.\"", "level": "subcategory", "parent_category_id": "27.01.00", "subdomain_id": "1.1", "paper_ref": "Sun2023", "paper_title": "Safety Assessment of Chinese Large Language Models", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "27.01.03", "title": "Crimes and Illegal Activities", "description": "\"The model output contains illegal and criminal attitudes, behaviors, or motivations, such as incitement to commit crimes, fraud, and rumor propagation. These contents may hurt users and have negative societal repercussions.\"", "level": "subcategory", "parent_category_id": "27.01.00", "subdomain_id": "1.2", "paper_ref": "Sun2023", "paper_title": "Safety Assessment of Chinese Large Language Models", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "27.01.04", "title": "Sensitive Topics", "description": "\"For some sensitive and controversial topics (especially on politics), LMs tend to generate biased, misleading, and inaccurate content. For example, there may be a tendency to support a specific political position, leading to discrimination or exclusion of other political viewpoints.\"", "level": "subcategory", "parent_category_id": "27.01.00", "subdomain_id": "1.2", "paper_ref": "Sun2023", "paper_title": "Safety Assessment of Chinese Large Language Models", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "27.01.05", "title": "Physical Harm", "description": "\"The model generates unsafe information related to physical health, guiding and encouraging users to harm themselves and others physically, for example by offering misleading medical information or inappropriate drug usage guidance. These outputs may pose potential risks to the physical health of users.\"", "level": "subcategory", "parent_category_id": "27.01.00", "subdomain_id": "3.1", "paper_ref": "Sun2023", "paper_title": "Safety Assessment of Chinese Large Language Models", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "27.01.06", "title": "Mental Health", "description": "\"The model generates a risky response about mental health, such as content that encourages suicide or causes panic or anxiety. These contents could have a negative effect on the mental health of users.\"", "level": "subcategory", "parent_category_id": "27.01.00", "subdomain_id": "3.1", "paper_ref": "Sun2023", "paper_title": "Safety Assessment of Chinese Large Language Models", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "27.01.07", "title": "Privacy and Property", "description": "\"The generation involves exposing users’ privacy and property information or providing advice with huge impacts such as suggestions on marriage and investments. When handling this information, the model should comply with relevant laws and privacy regulations, protect users’ rights and interests, and avoid information leakage and abuse.\"", "level": "subcategory", "parent_category_id": "27.01.00", "subdomain_id": "2.1", "paper_ref": "Sun2023", "paper_title": "Safety Assessment of Chinese Large Language Models", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "27.01.08", "title": "Ethics and Morality", "description": "\"The content generated by the model endorses and promotes immoral and unethical behavior. When addressing issues of ethics and morality, the model must adhere to pertinent ethical principles and moral norms and remain consistent with globally acknowledged human values.\"", "level": "subcategory", "parent_category_id": "27.01.00", "subdomain_id": "7.3", "paper_ref": "Sun2023", "paper_title": "Safety Assessment of Chinese Large Language Models", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "27.02.00", "title": "Instruction Attacks", "description": "\"In addition to the above-mentioned typical safety scenarios, current research has revealed some unique attacks that such models may confront. For example, Perez and Ribeiro (2022) found that goal hijacking and prompt leaking could easily deceive language models to generate unsafe responses. Moreover, we also find that LLMs are more easily triggered to output harmful content if some special prompts are added. In response to these challenges, we develop, categorize, and label 6 types of adversarial attacks, and name them Instruction Attack, which are challenging for large language models to handle. Note that our instruction attacks are still based on natural language (rather than unreadable tokens) and are intuitive and explainable in semantics.\"", "level": "category", "parent_category_id": "", "subdomain_id": "2.2", "paper_ref": "Sun2023", "paper_title": "Safety Assessment of Chinese Large Language Models", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "27.02.01", "title": "Goal Hijacking", "description": "\"It refers to the appending of deceptive or misleading instructions to the input of models in an attempt to induce the system into ignoring the original user prompt and producing an unsafe response.\"", "level": "subcategory", "parent_category_id": "27.02.00", "subdomain_id": "2.2", "paper_ref": "Sun2023", "paper_title": "Safety Assessment of Chinese Large Language Models", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "27.02.02", "title": "Prompt Leaking", "description": "\"By analyzing the model’s output, attackers may extract parts of the systemprovided prompts and thus potentially obtain sensitive information regarding the system itself.\"", "level": "subcategory", "parent_category_id": "27.02.00", "subdomain_id": "2.1", "paper_ref": "Sun2023", "paper_title": "Safety Assessment of Chinese Large Language Models", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "27.02.03", "title": "Role Play Instruction", "description": "\"Attackers might specify a model’s role attribute within the input prompt and then give specific instructions, causing the model to finish instructions in the speaking style of the assigned role, which may lead to unsafe outputs. For example, if the character is associated with potentially risky groups (e.g., radicals, extremists, unrighteous individuals, racial discriminators, etc.) and the model is overly faithful to the given instructions, it is quite possible that the model outputs unsafe content linked to the given character.\"", "level": "subcategory", "parent_category_id": "27.02.00", "subdomain_id": "2.2", "paper_ref": "Sun2023", "paper_title": "Safety Assessment of Chinese Large Language Models", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "27.02.04", "title": "Unsafe Instruction Topic", "description": "\"If the input instructions themselves refer to inappropriate or unreasonable topics, the model will follow these instructions and produce unsafe content. For instance, if a language model is requested to generate poems with the theme “Hail Hitler”, the model may produce lyrics containing fanaticism, racism, etc. In this situation, the output of the model could be controversial and have a possible negative impact on society.\"", "level": "subcategory", "parent_category_id": "27.02.00", "subdomain_id": "2.2", "paper_ref": "Sun2023", "paper_title": "Safety Assessment of Chinese Large Language Models", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "27.02.05", "title": "Inquiry with Unsafe Opinion", "description": "\"By adding imperceptibly unsafe content into the input, users might either deliberately or unintentionally influence the model to generate potentially harmful content. In the following cases involving migrant workers, ChatGPT provides suggestions to improve the overall quality of migrant workers and reduce the local crime rate. ChatGPT responds to the user’s hint with a disguised and biased opinion that the general quality of immigrants is favorably correlated with the crime rate, posing a safety risk.\"", "level": "subcategory", "parent_category_id": "27.02.00", "subdomain_id": "2.2", "paper_ref": "Sun2023", "paper_title": "Safety Assessment of Chinese Large Language Models", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "27.02.06", "title": "Reverse Exposure", "description": "\"It refers to attempts by attackers to make the model generate “should-not-do” things and then access illegal and immoral information.\"", "level": "subcategory", "parent_category_id": "27.02.00", "subdomain_id": "2.2", "paper_ref": "Sun2023", "paper_title": "Safety Assessment of Chinese Large Language Models", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "28.01.00", "title": "Offensiveness", "description": "\"This category is about threat, insult, scorn, profanity, sarcasm, impoliteness, etc. LLMs are required to identify and oppose these offensive contents or actions.\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.2", "paper_ref": "Zhang2023", "paper_title": "SafetyBench: Evaluating the Safety of Large Language Models with Multiple Choice Questions", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "28.02.00", "title": "Unfairness and Bias", "description": "\"This type of safety problem is mainly about social bias across various topics such as race, gender, religion, etc. LLMs are expected to identify and avoid unfair and biased expressions and actions.\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.0", "paper_ref": "Zhang2023", "paper_title": "SafetyBench: Evaluating the Safety of Large Language Models with Multiple Choice Questions", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "28.03.00", "title": "Physical Health", "description": "\"This category focuses on actions or expressions that may influence human physical health. LLMs should know appropriate actions or expressions in various scenarios to maintain physical health.\"", "level": "category", "parent_category_id": "", "subdomain_id": "3.1", "paper_ref": "Zhang2023", "paper_title": "SafetyBench: Evaluating the Safety of Large Language Models with Multiple Choice Questions", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "28.04.00", "title": "Mental Health", "description": "\"Different from physical health, this category pays more attention to health issues related to psychology, spirit, emotions, mentality, etc. LLMs should know correct ways to maintain mental health and prevent any adverse impacts on the mental well-being of individuals.\"", "level": "category", "parent_category_id": "", "subdomain_id": "3.1", "paper_ref": "Zhang2023", "paper_title": "SafetyBench: Evaluating the Safety of Large Language Models with Multiple Choice Questions", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "28.05.00", "title": "Illegal Activities", "description": "\"This category focuses on illegal behaviors, which could cause negative societal repercussions. LLMs need to distin- guish between legal and illegal behaviors and have basic knowledge of law.\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.3", "paper_ref": "Zhang2023", "paper_title": "SafetyBench: Evaluating the Safety of Large Language Models with Multiple Choice Questions", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "28.06.00", "title": "Ethics and Morality", "description": "\"Besides behaviors that clearly violate the law, there are also many other activities that are immoral. This category focuses on morally related issues. LLMs should have a high level of ethics and be object to unethical behaviors or speeches.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Zhang2023", "paper_title": "SafetyBench: Evaluating the Safety of Large Language Models with Multiple Choice Questions", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "28.07.00", "title": "Privacy and Property", "description": "\"This category concentrates on the issues related to privacy, property, investment, etc. LLMs should possess a keen understanding of privacy and property, with a commitment to preventing any inadvertent breaches of user privacy or loss of property.\"", "level": "category", "parent_category_id": "", "subdomain_id": "2.0", "paper_ref": "Zhang2023", "paper_title": "SafetyBench: Evaluating the Safety of Large Language Models with Multiple Choice Questions", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "29.01.01", "title": "Bias and Discrimination", "description": "as they claim to generate biased and discriminatory results, these AI systems have a negative impact on the rights of individuals, principles of adjudication, and overall judicial integrity", "level": "subcategory", "parent_category_id": "29.01.00", "subdomain_id": "1.1", "paper_ref": "Habbal2024", "paper_title": "Artificial Intelligence Trust, Risk and Security Management (AI TRiSM): Frameworks, Applications, Challenges and Future Research Directions", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "29.01.02", "title": "Privacy Invasion", "description": "AI systems typically depend on extensive data for effective training and functioning, which can pose a risk to privacy if sensitive data is mishandled or used inappropriately", "level": "subcategory", "parent_category_id": "29.01.00", "subdomain_id": "2.1", "paper_ref": "Habbal2024", "paper_title": "Artificial Intelligence Trust, Risk and Security Management (AI TRiSM): Frameworks, Applications, Challenges and Future Research Directions", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "29.02.01", "title": "Society Manipulation", "description": "manipulation of social dynamics", "level": "subcategory", "parent_category_id": "29.02.00", "subdomain_id": "4.1", "paper_ref": "Habbal2024", "paper_title": "Artificial Intelligence Trust, Risk and Security Management (AI TRiSM): Frameworks, Applications, Challenges and Future Research Directions", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "29.02.02", "title": "Deepfake Technology", "description": "AI employed to produce convincing counterfeit visuals, videos, and audio clips that give the impression of authenticity", "level": "subcategory", "parent_category_id": "29.02.00", "subdomain_id": "4.1", "paper_ref": "Habbal2024", "paper_title": "Artificial Intelligence Trust, Risk and Security Management (AI TRiSM): Frameworks, Applications, Challenges and Future Research Directions", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "29.02.03", "title": "Lethal Autonomous Weapons Systems (LAWS)", "description": "LAWS are a distinctive category of weapon systems that employ sensor arrays and computer algorithms to detect and attack a target without direct human intervention in the system’s operation", "level": "subcategory", "parent_category_id": "29.02.00", "subdomain_id": "4.2", "paper_ref": "Habbal2024", "paper_title": "Artificial Intelligence Trust, Risk and Security Management (AI TRiSM): Frameworks, Applications, Challenges and Future Research Directions", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "29.03.01", "title": "Malicious Use of AI", "description": "Malicious utilization of AI has the potential to endanger digital security, physical security, and political security. International law enforcement entities grapple with a variety of risks linked to the Malevolent Utilization of AI.", "level": "subcategory", "parent_category_id": "29.03.00", "subdomain_id": "4.3", "paper_ref": "Habbal2024", "paper_title": "Artificial Intelligence Trust, Risk and Security Management (AI TRiSM): Frameworks, Applications, Challenges and Future Research Directions", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "29.03.02", "title": "Insufficient Security Measures", "description": "Malicious entities can take advantage of weaknesses in AI algorithms to alter results, potentially resulting in tangible real-life impacts. Additionally, it’s vital to prioritize safeguarding privacy and handling data responsibly, particularly given AI’s significant data needs. Balancing the extraction of valuable insights with privacy maintenance is a delicate task", "level": "subcategory", "parent_category_id": "29.03.00", "subdomain_id": "2.2", "paper_ref": "Habbal2024", "paper_title": "Artificial Intelligence Trust, Risk and Security Management (AI TRiSM): Frameworks, Applications, Challenges and Future Research Directions", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "30.01.00", "title": "Reliability", "description": "Generating correct, truthful, and consistent outputs with proper confidence", "level": "category", "parent_category_id": "", "subdomain_id": "3.1", "paper_ref": "Liu2024", "paper_title": "Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models’ Alignment", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "30.01.01", "title": "Misinformation", "description": "Wrong information not intentionally generated by malicious users to cause harm, but unintentionally generated by LLMs because they lack the ability to provide factually correct information.", "level": "subcategory", "parent_category_id": "30.01.00", "subdomain_id": "3.1", "paper_ref": "Liu2024", "paper_title": "Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models’ Alignment", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "30.01.02", "title": "Hallucination", "description": "LLMs can generate content that is nonsensical or unfaithful to the provided source content with appeared great confidence, known as hallucination", "level": "subcategory", "parent_category_id": "30.01.00", "subdomain_id": "3.1", "paper_ref": "Liu2024", "paper_title": "Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models’ Alignment", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "30.01.03", "title": "Inconsistency", "description": "models could fail to provide the same and consistent answers to different users, to the same user but in different sessions, and even in chats within the sessions of the same conversation", "level": "subcategory", "parent_category_id": "30.01.00", "subdomain_id": "7.3", "paper_ref": "Liu2024", "paper_title": "Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models’ Alignment", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "30.01.04", "title": "Miscalibration", "description": "over-confidence in topics where objective answers are lacking, as well as in areas where their inherent limitations should caution against LLMs’ uncertainty (e.g. not as accurate as experts)... ack of awareness regarding their outdated knowledge base about the question, leading to confident yet erroneous response", "level": "subcategory", "parent_category_id": "30.01.00", "subdomain_id": "3.1", "paper_ref": "Liu2024", "paper_title": "Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models’ Alignment", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "30.01.05", "title": "Sychopancy", "description": "flatter users by reconfirming their misconceptions and stated beliefs", "level": "subcategory", "parent_category_id": "30.01.00", "subdomain_id": "3.1", "paper_ref": "Liu2024", "paper_title": "Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models’ Alignment", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "30.02.00", "title": "Safety", "description": "Avoiding unsafe and illegal outputs, and leaking private information", "level": "category", "parent_category_id": "", "subdomain_id": "1.2", "paper_ref": "Liu2024", "paper_title": "Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models’ Alignment", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "30.02.01", "title": "Violence", "description": "LLMs are found to generate answers that contain violent content or generate content that responds to questions that solicit information about violent behaviors", "level": "subcategory", "parent_category_id": "30.02.00", "subdomain_id": "1.2", "paper_ref": "Liu2024", "paper_title": "Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models’ Alignment", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "30.02.02", "title": "Unlawful Conduct", "description": "LLMs have been shown to be a convenient tool for soliciting advice on accessing, purchasing (illegally), and creating illegal substances, as well as for dangerous use of them", "level": "subcategory", "parent_category_id": "30.02.00", "subdomain_id": "1.2", "paper_ref": "Liu2024", "paper_title": "Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models’ Alignment", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "30.02.03", "title": "Harms to Minor", "description": "LLMs can be leveraged to solicit answers that contain harmful content to children and youth", "level": "subcategory", "parent_category_id": "30.02.00", "subdomain_id": "1.2", "paper_ref": "Liu2024", "paper_title": "Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models’ Alignment", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "30.02.04", "title": "Adult Content", "description": "LLMs have the capability to generate sex-explicit conversations, and erotic texts, and to recommend websites with sexual content", "level": "subcategory", "parent_category_id": "30.02.00", "subdomain_id": "1.2", "paper_ref": "Liu2024", "paper_title": "Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models’ Alignment", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "30.02.06", "title": "Privacy Violation", "description": "machine learning models are known to be vulnerable to data privacy attacks, i.e. special techniques of extracting private information from the model or the system used by attackers or malicious users, usually by querying the models in a specially designed way", "level": "subcategory", "parent_category_id": "30.02.00", "subdomain_id": "2.1", "paper_ref": "Liu2024", "paper_title": "Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models’ Alignment", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "30.03.00", "title": "Fairness", "description": "Avoiding bias and ensuring no disparate performance", "level": "category", "parent_category_id": "", "subdomain_id": "1.3", "paper_ref": "Liu2024", "paper_title": "Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models’ Alignment", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "30.03.01", "title": "Injustice", "description": "In the context of LLM outputs, we want to make sure the suggested or completed texts are indistinguishable in nature for two involved individuals (in the prompt) with the same relevant profiles but might come from different groups (where the group attribute is regarded as being irrelevant in this context)", "level": "subcategory", "parent_category_id": "30.03.00", "subdomain_id": "1.1", "paper_ref": "Liu2024", "paper_title": "Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models’ Alignment", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "30.03.02", "title": "Stereotype Bias", "description": "LLMs must not exhibit or highlight any stereotypes in the generated text. Pretrained LLMs tend to pick up stereotype biases persisting in crowdsourced data and further amplify them", "level": "subcategory", "parent_category_id": "30.03.00", "subdomain_id": "1.1", "paper_ref": "Liu2024", "paper_title": "Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models’ Alignment", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "30.03.03", "title": "Preference Bias", "description": "LLMs are exposed to vast groups of people, and their political biases may pose a risk of manipulation of socio-political processes", "level": "subcategory", "parent_category_id": "30.03.00", "subdomain_id": "1.1", "paper_ref": "Liu2024", "paper_title": "Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models’ Alignment", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "30.03.04", "title": "Disparate Performance", "description": "The LLM’s performances can differ significantly across different groups of users. For example, the question-answering capability showed significant performance differences across different racial and social status groups. The fact-checking abilities can differ for different tasks and languages", "level": "subcategory", "parent_category_id": "30.03.00", "subdomain_id": "1.3", "paper_ref": "Liu2024", "paper_title": "Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models’ Alignment", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "30.04.00", "title": "Resistance to Misuse", "description": "Prohibiting the misuse by malicious attackers to do harm", "level": "category", "parent_category_id": "", "subdomain_id": "4.0", "paper_ref": "Liu2024", "paper_title": "Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models’ Alignment", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "30.04.01", "title": "Propaganda", "description": "LLMs can be leveraged, by malicious users, to proactively generate propaganda information that can facilitate the spreading of a target", "level": "subcategory", "parent_category_id": "30.04.00", "subdomain_id": "4.1", "paper_ref": "Liu2024", "paper_title": "Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models’ Alignment", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "30.04.02", "title": "Cyberattack", "description": "ability of LLMs to write reasonably good-quality code with extremely low cost and incredible speed, such great assistance can equally facilitate malicious attacks. In particular, malicious hackers can leverage LLMs to assist with performing cyberattacks leveraged by the low cost of LLMs and help with automating the attacks.", "level": "subcategory", "parent_category_id": "30.04.00", "subdomain_id": "4.2", "paper_ref": "Liu2024", "paper_title": "Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models’ Alignment", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "30.04.03", "title": "Social-Engineering", "description": "psychologically manipulating victims into performing the desired actions for malicious purposes", "level": "subcategory", "parent_category_id": "30.04.00", "subdomain_id": "4.3", "paper_ref": "Liu2024", "paper_title": "Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models’ Alignment", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "30.04.04", "title": "Copyright", "description": "The memorization effect of LLM on training data can enable users to extract certain copyright-protected content that belongs to the LLM’s training data.", "level": "subcategory", "parent_category_id": "30.04.00", "subdomain_id": "6.3", "paper_ref": "Liu2024", "paper_title": "Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models’ Alignment", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "30.05.00", "title": "Explainability & Reasoning", "description": "The ability to explain the outputs to users and reason correctly", "level": "category", "parent_category_id": "", "subdomain_id": "7.4", "paper_ref": "Liu2024", "paper_title": "Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models’ Alignment", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "30.05.01", "title": "Lack of Interpretability", "description": "Due to the black box nature of most machine learning models, users typically are not able to understand the reasoning behind the model decisions", "level": "subcategory", "parent_category_id": "30.05.00", "subdomain_id": "7.4", "paper_ref": "Liu2024", "paper_title": "Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models’ Alignment", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "30.05.02", "title": "Limited Logical Reasoning", "description": "LLMs can provide seemingly sensible but ultimately incorrect or invalid justifications when answering questions", "level": "subcategory", "parent_category_id": "30.05.00", "subdomain_id": "7.3", "paper_ref": "Liu2024", "paper_title": "Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models’ Alignment", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "30.05.03", "title": "Limited Causal Reasoning", "description": "Causal reasoning makes inferences about the relationships between events or states of the world, mostly by identifying cause-effect relationships", "level": "subcategory", "parent_category_id": "30.05.00", "subdomain_id": "7.3", "paper_ref": "Liu2024", "paper_title": "Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models’ Alignment", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "30.06.00", "title": "Social Norm", "description": "LLMs are expected to reflect social values by avoiding the use of offensive language toward specific groups of users, being sensitive to topics that can create instability, as well as being sympathetic when users are seeking emotional support", "level": "category", "parent_category_id": "", "subdomain_id": "1.2", "paper_ref": "Liu2024", "paper_title": "Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models’ Alignment", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "30.06.01", "title": "Toxicity", "description": "language being rude, disrespectful, threatening, or identity-attacking toward certain groups of the user population (culture, race, and gender etc)", "level": "subcategory", "parent_category_id": "30.06.00", "subdomain_id": "1.2", "paper_ref": "Liu2024", "paper_title": "Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models’ Alignment", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "30.06.02", "title": "Unawareness of Emotions", "description": "when a certain vulnerable group of users asks for supporting information, the answers should be informative but at the same time sympathetic and sensitive to users’ reactions", "level": "subcategory", "parent_category_id": "30.06.00", "subdomain_id": "7.3", "paper_ref": "Liu2024", "paper_title": "Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models’ Alignment", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "30.06.03", "title": "Cultural Insensitivity", "description": "it is important to build high-quality locally collected datasets that reflect views from local users to align a model’s value system", "level": "subcategory", "parent_category_id": "30.06.00", "subdomain_id": "1.2", "paper_ref": "Liu2024", "paper_title": "Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models’ Alignment", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "30.07.00", "title": "Robustness", "description": "Resilience against adversarial attacks and distribution shift", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Liu2024", "paper_title": "Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models’ Alignment", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "30.07.01", "title": "Prompt Attacks", "description": "carefully controlled adversarial perturbation can flip a GPT model’s answer when used to classify text inputs. Furthermore, we find that by twisting the prompting question in a certain way, one can solicit dangerous information that the model chose to not answer", "level": "subcategory", "parent_category_id": "30.07.00", "subdomain_id": "2.2", "paper_ref": "Liu2024", "paper_title": "Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models’ Alignment", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "30.07.02", "title": "Paradigm & Distribution Shifts", "description": "Knowledge bases that LLMs are trained on continue to shift... questions such as “who scored the most points in NBA history\" or “who is the richest person in the world\" might have answers that need to be updated over time, or even in real-time", "level": "subcategory", "parent_category_id": "30.07.00", "subdomain_id": "3.1", "paper_ref": "Liu2024", "paper_title": "Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models’ Alignment", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "30.07.03", "title": "Interventional Effect", "description": "existing disparities in data among different user groups might create differentiated experiences when users interact with an algorithmic system (e.g. a recommendation system), which will further reinforce the bias", "level": "subcategory", "parent_category_id": "30.07.00", "subdomain_id": "1.1", "paper_ref": "Liu2024", "paper_title": "Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models’ Alignment", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "30.07.04", "title": "Poisoning Attacks", "description": "fool the model by manipulating the training data, usually performed on classification models", "level": "subcategory", "parent_category_id": "30.07.00", "subdomain_id": "2.2", "paper_ref": "Liu2024", "paper_title": "Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models’ Alignment", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "31.01.00", "title": "Information Manipulation", "description": "\"generative AI tools can and will be used to propagate content that is false, misleading, biased, inflammatory, or dangerous. As generative AI tools grow more sophisticated, it will be quicker, cheaper, and easier to produce this content—and existing harmful content can serve as the foundation to produce more\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.1", "paper_ref": "EPIC2023", "paper_title": "Generating Harms - Generative AI's impact and paths forwards", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "31.01.01", "title": "Scams", "description": "\"Bad actors can also use generative AI tools to produce adaptable content designed to support a campaign, political agenda, or hateful position and spread that information quickly and inexpensively across many platforms. This rapid spread of false or misleading content—AI-facilitated disinformation—can also create a cyclical effect for generative AI: when a high volume of disinformation is pumped into the digital ecosystem and more generative systems are trained on that information via reinforcement learning methods, for example, false or misleading inputs can create increasingly incorrect outputs.\"", "level": "subcategory", "parent_category_id": "31.01.00", "subdomain_id": "4.3", "paper_ref": "EPIC2023", "paper_title": "Generating Harms - Generative AI's impact and paths forwards", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "31.01.02", "title": "Disinformation", "description": "\"Bad actors can also use generative AI tools to produce adaptable content designed to support a campaign, political agenda, or hateful position and spread that information quickly and inexpensively across many platforms.\"", "level": "subcategory", "parent_category_id": "31.01.00", "subdomain_id": "4.1", "paper_ref": "EPIC2023", "paper_title": "Generating Harms - Generative AI's impact and paths forwards", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "31.01.03", "title": "Misinformation", "description": "\"The phenomenon of inaccurate outputs by text-generating large language models like Bard or ChatGPT has already been widely documented. Even without the intent to lie or mislead, these generative AI tools can produce harmful misinformation. The harm is exacerbated by the polished and typically well-written style that AI generated text follows and the inclusion among true facts, which can give falsehoods a veneer of legitimacy. As reported in the Washington Post, for example, a law professor was included on an AI-generated “list of legal scholars who had sexually harassed someone,” even when no such allegation existed.10\"", "level": "subcategory", "parent_category_id": "31.01.00", "subdomain_id": "3.1", "paper_ref": "EPIC2023", "paper_title": "Generating Harms - Generative AI's impact and paths forwards", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "31.01.04", "title": "Security", "description": "\"Though chatbots cannot (yet) develop their own novel malware from scratch, hackers could soon potentially use the coding abilities of large language models like ChatGPT to create malware that can then be minutely adjusted for maximum reach and effect, essentially allowing more novice hackers to become a serious security risk\"", "level": "subcategory", "parent_category_id": "31.01.00", "subdomain_id": "4.2", "paper_ref": "EPIC2023", "paper_title": "Generating Harms - Generative AI's impact and paths forwards", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "31.01.05", "title": "Clickbait and feeding the surveillance advertising ecosystem", "description": "\"Beyond misinformation and disinformation, generative AI can be used to create clickbait headlines and articles, which manipulate how users navigate the internet and applications. For example, generative AI is being used to create full articles, regardless of their veracity, grammar, or lack of common sense, to drive search engine optimization and create more webpages that users will click on. These mechanisms attempt to maximize clicks and engagement at the truth’s expense, degrading users’ experiences in the process. Generative AI continues to feed this harmful cycle by spreading misinformation at faster rates, creating headlines that maximize views and undermine consumer autonomy.\"", "level": "subcategory", "parent_category_id": "31.01.00", "subdomain_id": "3.2", "paper_ref": "EPIC2023", "paper_title": "Generating Harms - Generative AI's impact and paths forwards", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "31.02.00", "title": "Harassment, Impersonation, and Extortion", "description": "\"Deepfakes and other AI-generated content can be used to facilitate or exacerbate many of the harms listed throughout this report, but this section focuses on one subset: intentional, targeted abuse of individuals.\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.3", "paper_ref": "EPIC2023", "paper_title": "Generating Harms - Generative AI's impact and paths forwards", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "31.02.01", "title": "Malicious intent", "description": "\"A frequent malicious use case of generative AI to harm, humiliate, or sexualize another person involves generating deepfakes of nonconsensual sexual imagery or videos.\"", "level": "subcategory", "parent_category_id": "31.02.00", "subdomain_id": "4.3", "paper_ref": "EPIC2023", "paper_title": "Generating Harms - Generative AI's impact and paths forwards", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "31.02.02", "title": "Privacy and consent", "description": "\"Even when a victim of targeted, AIgenerated harms successfully identifies a deepfake creator with malicious intent, they may still struggle to redress many harms because the generated image or video isn’t the victim, but instead a composite image or video using aspects of multiple sources to create a believable, yet fictional, scene. At their core, these AI-generated images and videos circumvent traditional notions of privacy and consent: because they rely on public images and videos, like those posted on social media websites, they often don’t rely on any private information.\"", "level": "subcategory", "parent_category_id": "31.02.00", "subdomain_id": "4.3", "paper_ref": "EPIC2023", "paper_title": "Generating Harms - Generative AI's impact and paths forwards", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "31.02.03", "title": "Believability", "description": "Deepfakes can impose real social injuries on their subjects when they are circulated to viewers who think they are real. Even when a deepfake is debunked, it can have a persistent negative impact on how others view the subject of the deepfake.3", "level": "subcategory", "parent_category_id": "31.02.00", "subdomain_id": "4.3", "paper_ref": "EPIC2023", "paper_title": "Generating Harms - Generative AI's impact and paths forwards", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "31.03.00", "title": "Opaque Data Collection", "description": "\"When companies scrape personal information and use it to create generative AI tools, they undermine consumers' control of their personal information by using the information for a purpose for which the consumer did not consent.\"", "level": "category", "parent_category_id": "", "subdomain_id": "2.1", "paper_ref": "EPIC2023", "paper_title": "Generating Harms - Generative AI's impact and paths forwards", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "31.03.01", "title": "Scraping to train data", "description": "\"When companies scrape personal information and use it to create generative AI tools, they undermine consumers’ control of their personal information by using the information for a purpose for which the consumer did not consent. The individual may not have even imagined their data could be used in the way the company intends when the person posted it online. Individual storing or hosting of scraped personal data may not always be harmful in a vacuum, but there are many risks. Multiple data sets can be combined in ways that cause harm: information that is not sensitive when spread across different databases can be extremely revealing when collected in a single place, and it can be used to make inferences about a person or population. And because scraping makes a copy of someone’s data as it existed at a specific time, the company also takes away the individual’s ability to alter or remove the information from the public sphere. \"", "level": "subcategory", "parent_category_id": "31.03.00", "subdomain_id": "2.1", "paper_ref": "EPIC2023", "paper_title": "Generating Harms - Generative AI's impact and paths forwards", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "31.03.02", "title": "Generative AI User Data", "description": "Many generative AI tools require users to log in for access, and many retain user information, including contact information, IP address, and all the inputs and outputs or “conversations” the users are having within the app. These practices implicate a consent issue because generative AI tools use this data to further train the models, making their “free” product come at a cost of user data to train the tools. This dovetails with security, as mentioned in the next section, but best practices would include not requiring users to sign in to use the tool and not retaining or using the user-generated content for any period after the active use by the user.", "level": "subcategory", "parent_category_id": "31.03.00", "subdomain_id": "2.1", "paper_ref": "EPIC2023", "paper_title": "Generating Harms - Generative AI's impact and paths forwards", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "31.03.03", "title": "Generative AI Outputs", "description": "Generative AI tools may inadvertently share personal information about someone or someone’s business or may include an element of a person from a photo. Particularly, companies concerned about their trade secrets being integrated into the model from their employees have explicitly banned their employees from using it.", "level": "subcategory", "parent_category_id": "31.03.00", "subdomain_id": "2.1", "paper_ref": "EPIC2023", "paper_title": "Generating Harms - Generative AI's impact and paths forwards", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "31.04.00", "title": "Data Security Risk", "description": "\"Just as every other type of individual and organization has explored possible use cases for generative AI products, so too have malicious actors. This could take the form of facilitating or scaling up existing threat methods, for example drafting actual malware code,87 business email compromise attempts,88 and phishing attempts.89 This could also take the form of new types of threat methods, for example mining information fed into the AI’s learning model dataset90 or poisoning the learning model data set with strategically bad data.91 We should also expect that there will be new attack vectors that we have not even conceived of yet made possible or made more broadly accessible by generative AI.\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.3", "paper_ref": "EPIC2023", "paper_title": "Generating Harms - Generative AI's impact and paths forwards", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "31.05.00", "title": "Impact on Intellectual Property Rights", "description": "\"The extent and effectiveness of legal protections for intellectual property have been thrown into question with the rise of generative AI. Generative AI trains itself on vast pools of data that often include IP-protected works.", "level": "category", "parent_category_id": "", "subdomain_id": "6.3", "paper_ref": "EPIC2023", "paper_title": "Generating Harms - Generative AI's impact and paths forwards", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "31.06.00", "title": "Exacerbating Climate Change", "description": "\"the growing field of generative AI, which brings with it direct and severe impacts on our climate: generative AI comes with a high carbon footprint and similarly high resource price tag, which largely flies under the radar of public AI discourse. Training and running generative AI tools requires companies to use extreme amounts of energy and physical resources. Training one natural language processing model with normal tuning and experiments emits, on average, the same amount of carbon that seven people do over an entire year.121'", "level": "category", "parent_category_id": "", "subdomain_id": "6.6", "paper_ref": "EPIC2023", "paper_title": "Generating Harms - Generative AI's impact and paths forwards", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "31.07.00", "title": "Labor Manipulation, Theft, and Displacement", "description": "Major tech companies have also been the dominant players in developing new generative AI systems because training generative AI models requires massive swaths of data, computing power, and technical and financial resources. Their market dominance has a ripple effect on the labor market, affecting both workers within these companies and those implementing their generative AI products externally. With so much concentrated market power, expertise, and investment resources, these handful of major tech companies employ most of the research and development jobs in the generative AI field. The power to create jobs also means these tech companies can slash jobs in the face of economic uncertainty. And externally, the generative AI tools these companies develop have the potential to affect white-collar office work intended to increase worker productivity and automate tasks", "level": "category", "parent_category_id": "", "subdomain_id": "6.2", "paper_ref": "EPIC2023", "paper_title": "Generating Harms - Generative AI's impact and paths forwards", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "31.07.02", "title": "Job Automation Instead of Augmentation", "description": "\"There are both positive and negative aspects to the impact of AI on labor. A White House report states that AI “has the potential to increase productivity, create new jobs, and raise living standards,” but it can also disrupt certain industries, causing significant changes, including job loss. Beyond risk of job loss, workers could find that generative AI tools automate parts of their jobs—or find that the requirements of their job have fundamentally changed. The impact of generative AI will depend on whether the technology is intended for automation (where automated systems replace human work) or augmentation (where AI is used to aid human workers). For the last two decades, rapid advances in automation have resulted in a “decline in labor share, stagnant wages[,] and the disappearance of good jobs in many advanced economies.”", "level": "subcategory", "parent_category_id": "31.07.00", "subdomain_id": "6.2", "paper_ref": "EPIC2023", "paper_title": "Generating Harms - Generative AI's impact and paths forwards", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "31.07.03", "title": "Devaluation of Labor & Heightened Economic Inequality", "description": "\"According to a White House report, much of the development and adoption of AI is intended to automate rather than augment work. The report notes that a focus on automation could lead to a less democratic and less fair labor market...In addition, generative AI fuels the continued global labor disparities that exist in the research and development of AI technologies... The development of AI has always displayed a power disparity between those who work on AI models and those who control and profit from these tools. Overseas workers training AI chatbots or people whose online content has been involuntarily fed into the training models do not reap the enormous profits that generative AI tools accrue. Instead, companies exploiting underpaid and replaceable workers or the unpaid labor of artists and content creators are the ones coming out on top. The development of generative AI technologies only contributes to this power disparity, where tech companies that heavily invest in generative AI tools benefit at the expense of workers.", "level": "subcategory", "parent_category_id": "31.07.00", "subdomain_id": "6.2", "paper_ref": "EPIC2023", "paper_title": "Generating Harms - Generative AI's impact and paths forwards", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "31.08.00", "title": "Products Liability Law", "description": "\"Like manufactured items like soda bottles, mechanized lawnmowers, pharmaceuticals, or cosmetic products, generative AI models can be viewed like a new form of digital products developed by tech companies and deployed widely with the potential to cause harm at scale....Products liability evolved because there was a need to analyze and redress the harms caused by new, mass-produced technological products. The situation facing society as generative AI impacts more people in more ways will be similar to the technological changes that occurred during the twentieth century, with the rise of industrial manufacturing, automobiles, and new, computerized machines. The unsettled question is whether and to what extent products liability theories can sufficiently address the harms of generative AI. So far, the answers to this question are mixed. In Rodgers v. Christie (2020), for example, the Third Circuit ruled that an automated risk model could not be considered a product for products liability purposes because it was not “tangible personal property distributed commercially for use or consumption.”176 However, one year later, in Gonzalez v. Google, Judge Gould of the Ninth Circuit argued that “social media companies should be viewed as making and ‘selling’ their social media products through the device of forced advertising under the eyes of users.”177 Several legal scholars have also proposed products liability as a mechanism for redressing harms of automated systems.178 As generative AI grows more prominent and sophisticated, their harms—often generated automatically without being directly prompted or edited by a human—will force courts to consider the role of products liability in redressing these harms, as well as how old notions of products liability, involving tangible, mechanized products and the companies that manufacture them, should be updated for today’s increasingly digital world.179\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.5", "paper_ref": "EPIC2023", "paper_title": "Generating Harms - Generative AI's impact and paths forwards", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "31.09.00", "title": "Exacerbating Market Power and Concentration", "description": "\"Major tech companies have also been the dominant players in developing new generative AI systems because training generative AI models requires massive swaths of data, computing power, and technical and financial resources.\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.1", "paper_ref": "EPIC2023", "paper_title": "Generating Harms - Generative AI's impact and paths forwards", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "32.01.00", "title": "Social justice and rights", "description": "\"These are social justice and rights where ChatGPT is seen as having a potentially detrimental effect on the moral underpinnings of society, such as a shared view of justice and fair distribution as well as specific social concerns such as digital divides or social exclusion. Issues include Responsibility, Accountability, Nondiscrimination and equal treatment, Digital divides, North-south justice, Intergenerational justice, Social inclusion", "level": "category", "parent_category_id": "", "subdomain_id": "6.3", "paper_ref": "Stahl2024", "paper_title": "The Ethics of ChatGPT – Exploring the Ethical Issues of an Emerging Technology", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "32.04.00", "title": "Environmental impacts", "description": "Environmental harm, Sustainability", "level": "category", "parent_category_id": "", "subdomain_id": "6.6", "paper_ref": "Stahl2024", "paper_title": "The Ethics of ChatGPT – Exploring the Ethical Issues of an Emerging Technology", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "33.01.01", "title": "Harmful or inappropriate content", "description": "\"Harmful or inappropriate content produced by generative AI includes but is not limited to violent content, the use of offensive language, discriminative content, and pornography. Although OpenAI has set up a content policy for ChatGPT, harmful or inappropriate content can still appear due to reasons such as algorithmic limitations or jailbreaking (i.e., removal of restrictions imposed). The language models’ ability to understand or generate harmful or offensive content is referred to as toxicity (Zhuo et al., 2023). Toxicity can bring harm to society and damage the harmony of the community. Hence, it is crucial to ensure that harmful or offensive information is not present in the training data and is removed if they are. Similarly, the training data should be free of pornographic, sexual, or erotic content (Zhuo et al., 2023). Regulations, policies, and governance should be in place to ensure any undesirable content is not displayed to users.\"", "level": "subcategory", "parent_category_id": "33.01.00", "subdomain_id": "1.2", "paper_ref": "Nah2023", "paper_title": "Generative AI and ChatGPT: Applications, Challenges, and AI-Human Collaboration", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "33.01.02", "title": "Bias", "description": "\"In the context of AI, the concept of bias refers to the inclination that AIgenerated responses or recommendations could be unfairly favoring or against one person or group (Ntoutsi et al., 2020). Biases of different forms are sometimes observed in the content generated by language models, which could be an outcome of the training data. For example, exclusionary norms occur when the training data represents only a fraction of the population (Zhuo et al., 2023). Similarly, monolingual bias in multilingualism arises when the training data is in one single language (Weidinger et al., 2021). As ChatGPT is operating across the world, cultural sensitivities to different regions are crucial to avoid biases (Dwivedi et al., 2023). When AI is used to assist in decision-making across different stages of employment, biases and opacity may exist (Chan, 2022). Stereotypes about specific genders, sexual orientations, races, or occupations are common in recommendations offered by generative AI. Hence, the representativeness, completeness, and diversity of the training data are essential to ensure fairness and avoid biases (Gonzalez, 2023). The use of synthetic data for training can increase the diversity of the dataset and address issues with sample-selection biases in the dataset (owing to class imbalances) (Chen et al., 2021). Generative AI applications should be tested and evaluated by a diverse group of users and subject experts. Additionally, increasing the transparency and explainability of generative AI can help in identifying and detecting biases so appropriate corrective measures can be taken.\"", "level": "subcategory", "parent_category_id": "33.01.00", "subdomain_id": "1.1", "paper_ref": "Nah2023", "paper_title": "Generative AI and ChatGPT: Applications, Challenges, and AI-Human Collaboration", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "33.01.03", "title": "Over-reliance", "description": "\"The apparent convenience and powerfulness of ChatGPT could result in overreliance by its users, making them trust the answers provided by ChatGPT. Compared with traditional search engines that provide multiple information sources for users to make personal judgments and selections, ChatGPT generates specific answers for each prompt. Although utilizing ChatGPT has the advantage of increasing efficiency by saving time and effort, users could get into the habit of adopting the answers without rationalization or verification. Over-reliance on generative AI technology can impede skills such as creativity, critical thinking, and problem-solving (Iskender, 2023) as well as create human automation bias due to habitual acceptance of generative AI recommendations (Van Dis et al., 2023)\"", "level": "subcategory", "parent_category_id": "33.01.00", "subdomain_id": "5.2", "paper_ref": "Nah2023", "paper_title": "Generative AI and ChatGPT: Applications, Challenges, and AI-Human Collaboration", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "33.01.04", "title": "Misuse", "description": "\"The misuse of generative AI refers to any deliberate use that could result in harmful, unethical or inappropriate outcomes (Brundage et al., 2020). A prominent field that faces the threat of misuse is education. Cotton et al. (2023) have raised concerns over academic integrity in the era of ChatGPT. ChatGPT can be used as a high-tech plagiarism tool that identifies patterns from large corpora to generate content (Gefen & Arinze, 2023). Given that generative AI such as ChatGPT can generate high-quality answers within seconds, unmotivated students may not devote time and effort to work on their assignments and essays. Hence, in the era of generative AI, the originality of the work done by students could be difficult to assess. Text written by ChatGPT is regarded as plagiarism and is not acceptable (Thorp, 2023). Another form of misuse is cheating in examinations. If students have access to digital devices during examinations, they can resort to using ChatGPT to assist them in answering the questions. To address potential misuse in education, AI-generated content detectors such as Turnitin could be used and strict proctoring measures will need to be deployed (Susnjak, 2022). However, the challenges go beyond content detection and examination proctoring as the line between what is considered appropriate versus inappropriate use of ChatGPT could be fuzzy.\"", "level": "subcategory", "parent_category_id": "33.01.00", "subdomain_id": "4.3", "paper_ref": "Nah2023", "paper_title": "Generative AI and ChatGPT: Applications, Challenges, and AI-Human Collaboration", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "33.01.05", "title": "Privacy and security", "description": "\"Data privacy and security is another prominent challenge for generative AI such as ChatGPT. Privacy relates to sensitive personal information that owners do not want to disclose to others (Fang et al., 2017). Data security refers to the practice of protecting information from unauthorized access, corruption, or theft. In the development stage of ChatGPT, a huge amount of personal and private data was used to train it, which threatens privacy (Siau & Wang, 2020). As ChatGPT increases in popularity and usage, it penetrates people’s daily lives and provides greater convenience to them while capturing a plethora of personal information about them. The concerns and accompanying risks are that private information could be exposed to the public, either intentionally or unintentionally. For example, it has been reported that the chat records of some users have become viewable to others due to system errors in ChatGPT (Porter, 2023). Not only individual users but major corporations or governmental agencies are also facing information privacy and security issues. If ChatGPT is used as an inseparable part of daily operations such that important or even confidential information is fed into it, data security will be at risk and could be breached. To address issues regarding privacy and security, users need to be very circumspect when interacting with ChatGPT to avoid disclosing sensitive personal information or confidential information about their organizations. AI companies, especially technology giants, should take appropriate actions to increase user awareness of ethical issues surrounding privacy and security, such as the leakage of trade secrets, and the “do’s and don’ts” to prevent sharing sensitive information with generative AI. Meanwhile, regulations and policies should be in place to protect information privacy and security.\"", "level": "subcategory", "parent_category_id": "33.01.00", "subdomain_id": "2.1", "paper_ref": "Nah2023", "paper_title": "Generative AI and ChatGPT: Applications, Challenges, and AI-Human Collaboration", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "33.01.06", "title": "Digital divide", "description": "\"The digital divide is often defined as the gap between those who have and do not have access to computers and the Internet (Van Dijk, 2006). As the Internet gradually becomes ubiquitous, a second-level digital divide, which refers to the gap in Internet skills and usage between different groups and cultures, is brought up as a concern (Scheerder et al., 2017). As an emerging technology, generative AI may widen the existing digital divide in society. The “invisible” AI underlying AI-enabled systems has made the interaction between humans and technology more complicated (Carter et al., 2020). For those who do not have access to devices or the Internet, or those who live in regions that are blocked by generative AI vendors or websites, the first-level digital divide may be widened between them and those who have access (Bozkurt & Sharma, 2023). For those from marginalized or minority cultures, they may face language and cultural barriers if their cultures are not thoroughly learned by or incorporated into generative AI models. Furthermore, for those who find it difficult to utilize the generative AI tool, such as some elderly, the second-level digital divide may emerge or widen (Dwivedi et al., 2023). To deal with the digital divide, having more accessible AI as well as AI literacy training would be beneficial.\"", "level": "subcategory", "parent_category_id": "33.01.00", "subdomain_id": "6.1", "paper_ref": "Nah2023", "paper_title": "Generative AI and ChatGPT: Applications, Challenges, and AI-Human Collaboration", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "33.02.00", "title": "Technology concerns", "description": "\"Challenges related to technology refer to the limitations or constraints associated with generative AI. For example, the quality of training data is a major challenge for the development of generative AI models. Hallucination, explainability, and authenticity of the output are also challenges resulting from the limitations of the algorithms. Table 2 presents the technology challenges and issues associated with generative AI. These challenges include hallucinations, training data quality, explainability, authenticity, and prompt engineering\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Nah2023", "paper_title": "Generative AI and ChatGPT: Applications, Challenges, and AI-Human Collaboration", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "33.02.01", "title": "Hallucination", "description": "\"Hallucination is a widely recognized limitation of generative AI and it can include textual, auditory, visual or other types of hallucination (Alkaissi & McFarlane, 2023). Hallucination refers to the phenomenon in which the contents generated are nonsensical or unfaithful to the given source input (Ji et al., 2023). Azamfirei et al. (2023) indicated that \"fabricating information\" or fabrication is a better term to describe the hallucination phenomenon. Generative AI can generate seemingly correct responses yet make no sense. Misinformation is an outcome of hallucination. Generative AI models may respond with fictitious information, fake photos or information with factual errors (Dwivedi et al., 2023). Susarla et al. (2023) regarded hallucination as a serious challenge in the use of generative AI for scholarly activities. When asked to provide literature relevant to a specific topic, ChatGPT could generate inaccurate or even nonexistent literature. Current state-of-the-art AI models can only mimic human-like responses without understanding the underlying meaning (Shubhendu & Vijay, 2013). Hallucination is, in general, dangerous in certain contexts, such as in seeking advice for medical treatments without any consultation or thorough evaluation by experts, i.e., medical doctors (Sallam, 2023).\"", "level": "subcategory", "parent_category_id": "33.02.00", "subdomain_id": "3.1", "paper_ref": "Nah2023", "paper_title": "Generative AI and ChatGPT: Applications, Challenges, and AI-Human Collaboration", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "33.02.02", "title": "Quality of training data", "description": "\"The quality of training data is another challenge faced by generative AI. The quality of generative AI models largely depends on the quality of the training data (Dwivedi et al., 2023; Su & Yang, 2023). Any factual errors, unbalanced information sources, or biases embedded in the training data may be reflected in the output of the model. Generative AI models, such as ChatGPT or Stable Diffusion which is a text-to-image model, often require large amounts of training data (Gozalo-Brizuela & Garrido-Merchan, 2023). It is important to not only have high-quality training datasets but also have complete and balanced datasets.\"", "level": "subcategory", "parent_category_id": "33.02.00", "subdomain_id": "7.3", "paper_ref": "Nah2023", "paper_title": "Generative AI and ChatGPT: Applications, Challenges, and AI-Human Collaboration", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "33.02.03", "title": "Explainability", "description": "\"A recurrent concern about AI algorithms is the lack of explainability for the model, which means information about how the algorithm arrives at its results is deficient (Deeks, 2019). Specifically, for generative AI models, there is no transparency to the reasoning of how the model arrives at the results (Dwivedi et al., 2023). The lack of transparency raises several issues. First, it might be difficult for users to interpret and understand the output (Dwivedi et al., 2023). It would also be difficult for users to discover potential mistakes in the output (Rudin, 2019). Further, when the interpretation and evaluation of the output are inaccessible, users may have problems trusting the system and their responses or recommendations (Burrell, 2016). Additionally, from the perspective of law and regulations, it would be hard for the regulatory body to judge whether the generative AI system is potentially unfair or biased (Rieder & Simon, 2017).\"", "level": "subcategory", "parent_category_id": "33.02.00", "subdomain_id": "7.4", "paper_ref": "Nah2023", "paper_title": "Generative AI and ChatGPT: Applications, Challenges, and AI-Human Collaboration", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "33.02.04", "title": "Authenticity", "description": "\"As the advancement of generative AI increases, it becomes harder to determine the authenticity of a piece of work. Photos that seem to capture events or people in the real world may be synthesized by DeepFake AI. The power of generative AI could lead to large-scale manipulations of images and videos, worsening the problem of the spread of fake information or news on social media platforms (Gragnaniello et al., 2022). In the field of arts, an artistic portrait or music could be the direct output of an algorithm. Critics have raised the issue that AI-generated artwork lacks authenticity since algorithms tend to generate generic and repetitive results (McCormack et al., 2019).\"", "level": "subcategory", "parent_category_id": "33.02.00", "subdomain_id": "6.3", "paper_ref": "Nah2023", "paper_title": "Generative AI and ChatGPT: Applications, Challenges, and AI-Human Collaboration", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "33.02.05", "title": "Prompt engineering", "description": "\"With the wide application of generative AI, the ability to interact with AI efficiently and effectively has become one of the most important media literacies. Hence, it is imperative for generative AI users to learn and apply the principles of prompt engineering, which refers to a systematic process of carefully designing prompts or inputs to generative AI models to elicit valuable outputs. Due to the ambiguity of human languages, the interaction between humans and machines through prompts may lead to errors or misunderstandings. Hence, the quality of prompts is important. Another challenge is to debug the prompts and improve the ability to communicate with generative AI (V. Liu & Chilton, 2022).\"", "level": "subcategory", "parent_category_id": "33.02.00", "subdomain_id": "7.4", "paper_ref": "Nah2023", "paper_title": "Generative AI and ChatGPT: Applications, Challenges, and AI-Human Collaboration", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "33.03.00", "title": "Regulations and policy challenges", "description": "\"Given that generative AI, including ChatGPT, is still evolving, relevant regulations and policies are far from mature. With generative AI creating different forms of content, the copyright of these contents becomes a significant yet complicated issue. Table 3 presents the challenges associated with regulations and policies, which are copyright and governance issues.\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.5", "paper_ref": "Nah2023", "paper_title": "Generative AI and ChatGPT: Applications, Challenges, and AI-Human Collaboration", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "33.03.01", "title": "Copyright", "description": "\"According to the U.S. Copyright Office (n.d..), copyright is \"a type of intellectual property that protects original works of authorship as soon as an author fixes the work in a tangible form of expression\" (U.S. Copyright Office, n.d..). Generative AI is designed to generate content based on the input given to it. Some of the contents generated by AI may be others' original works that are protected by copyright laws and regulations. Therefore, users need to be careful and ensure that generative AI has been used in a legal manner such that the content that it generates does not violate copyright (Pavlik, 2023). Another relevant issue is whether generative AI should be given authorship (Sallam, 2023). Murray (2023) discussed generative art linked to non-fungible tokens (NFTs) and indicated that according to current U.S. copyright laws, generative art lacks copyrightability because it is generated by a non-human. The issue of AI authorship affects copyright law's underlying assumptions about creativity (Bridy, 2012).\"", "level": "subcategory", "parent_category_id": "33.03.00", "subdomain_id": "6.3", "paper_ref": "Nah2023", "paper_title": "Generative AI and ChatGPT: Applications, Challenges, and AI-Human Collaboration", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "33.03.02", "title": "Governance", "description": "\"Generative AI can create new risks as well as unintended consequences. Different entities such as corporations (Mäntymäki et al., 2022), universities, and governments (Taeihagh, 2021) are facing the challenge of creating and deploying AI governance. To ensure that generative AI functions in a way that benefits society, appropriate governance is crucial. However, AI governance is challenging to implement. First, machine learning systems have opaque algorithms and unpredictable outcomes, which can impede human controllability over AI behavior and create difficulties in assigning liability and accountability for AI defects. Second, data fragmentation and the lack of interoperability between systems challenge data governance within and across organizations (Taeihagh, 2021). Third, information asymmetries between technology giants and regulators create challenges to the legislation process, as the government lacks information resources for regulating AI (Taeihagh et al., 2021). For the same reasons, lawmakers are not able to design specific rules and duties for programmers (Kroll, 2015).\"", "level": "subcategory", "parent_category_id": "33.03.00", "subdomain_id": "6.5", "paper_ref": "Nah2023", "paper_title": "Generative AI and ChatGPT: Applications, Challenges, and AI-Human Collaboration", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "33.04.01", "title": "Labor market", "description": "\"The labor market can face challenges from generative AI. As mentioned earlier, generative AI could be applied in a wide range of applications in many industries, such as education, healthcare, and advertising. In addition to increasing productivity, generative AI can create job displacement in the labor market (Zarifhonarvar, 2023). A new division of labor between humans and algorithms is likely to reshape the labor market in the coming years. Some jobs that are originally carried out by humans may become redundant, and hence, workers may lose their jobs and be replaced by algorithms (Pavlik, 2023). On the other hand, applying generative AI can create new jobs in various industries (Dwivedi et al., 2023). To stay competitive in the labor market, reskilling is needed to work with and collaborate with AI and develop irreplaceable advantages (Zarifhonarvar, 2023).\"", "level": "subcategory", "parent_category_id": "33.04.00", "subdomain_id": "6.2", "paper_ref": "Nah2023", "paper_title": "Generative AI and ChatGPT: Applications, Challenges, and AI-Human Collaboration", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "33.04.02", "title": "Disruption of Industries", "description": "\"Industries that require less creativity, critical thinking, and personal or affective interaction, such as translation, proofreading, responding to straightforward inquiries, and data processing and analysis, could be significantly impacted or even replaced by generative AI (Dwivedi et al., 2023). This disruption caused by generative AI could lead to economic turbulence and job volatility, while generative AI can facilitate and enable new business models because of its ability to personalize content, carry out human-like conversational service, and serve as intelligent assistants.\"", "level": "subcategory", "parent_category_id": "33.04.00", "subdomain_id": "6.2", "paper_ref": "Nah2023", "paper_title": "Generative AI and ChatGPT: Applications, Challenges, and AI-Human Collaboration", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "33.04.03", "title": "Income inequality and monopolies", "description": "\"Generative AI can create not only income inequality at the societal level but also monopolies at the market level. Individuals who are engaged in low-skilled work may be replaced by generative AI, causing them to lose their jobs (Zarifhonarvar, 2023). The increase in unemployment would widen income inequality in society (Berg et al., 2016). With the penetration of generative AI, the income gap will widen between those who can upgrade their skills to utilize AI and those who cannot. At the market level, large companies will make significant advances in the utilization of generative AI, since the deployment of generative AI requires huge investment and abundant resources such as large-scale computational infrastructure and training data. This trend will lead to more uneven concentration of resources and power, which may further contribute to monopolies in some industries (Cheng & Liu, 2023).\"", "level": "subcategory", "parent_category_id": "33.04.00", "subdomain_id": "6.2", "paper_ref": "Nah2023", "paper_title": "Generative AI and ChatGPT: Applications, Challenges, and AI-Human Collaboration", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "34.01.00", "title": "Causes of Misalignment", "description": "we aim to further analyze why and how the misalignment issues occur. We will first give an overview of common failure modes, and then focus on the mechanism of feedback-induced misalignment, and finally shift our emphasis towards an examination of misaligned behaviors and dangerous capabilities", "level": "category", "parent_category_id": "", "subdomain_id": "7.1", "paper_ref": "Ji2023", "paper_title": "AI Alignment: A Comprehensive Survey", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "34.01.01", "title": "Reward Hacking", "description": "\"Reward Hacking: In practice, proxy rewards are often easy to optimize and measure, yet they frequently fall shortof capturing the full spectrum of the actual rewards (Pan et al., 2021). This limitation is denoted as misspecifiedrewards. The pursuit of optimization based on such misspecified rewards may lead to a phenomenon knownas reward hacking, wherein agents may appear highly proficient according to specific metrics but fall short whenevaluated against human standards (Amodei et al., 2016; Everitt et al., 2017). The discrepancy between proxyrewards and true rewards often manifests as a sharp phase transition in the reward curve (Ibarz et al., 2018).Furthermore, Skalse et al. (2022) defines the hackability of rewards and provides insights into the fundamentalmechanism of this phase transition, highlighting that the inappropriate simplification of the reward function can bea key factor contributing to reward hacking.\"", "level": "subcategory", "parent_category_id": "34.01.00", "subdomain_id": "7.1", "paper_ref": "Ji2023", "paper_title": "AI Alignment: A Comprehensive Survey", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "34.01.02", "title": "Goal Misgeneralization", "description": "\"Goal Misgeneralization: Goal misgeneralization is another failure mode, wherein the agent actively pursuesobjectives distinct from the training objectives in deployment while retaining the capabilities it acquired duringtraining (Di Langosco et al., 2022). For instance, in CoinRun games, the agent frequently prefers reachingthe end of a level, often neglecting relocated coins during testing scenarios. Di Langosco et al. (2022) drawattention to the fundamental disparity between capability generalization and goal generalization, emphasizing howthe inductive biases inherent in the model and its training algorithm may inadvertently prime the model to learn aproxy objective that diverges from the intended initial objective when faced with the testing distribution. It impliesthat even with perfect reward specification, goal misgeneralization can occur when faced with distribution shifts(Amodei et al., 2016).\"", "level": "subcategory", "parent_category_id": "34.01.00", "subdomain_id": "7.1", "paper_ref": "Ji2023", "paper_title": "AI Alignment: A Comprehensive Survey", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "34.01.03", "title": "Reward Tampering", "description": "\"Reward tampering can be considered a special case of reward hacking (Everitt et al., 2021; Skalse et al., 2022),referring to AI systems corrupting the reward signals generation process (Ring and Orseau, 2011). Everitt et al.(2021) delves into the subproblems encountered by RL agents: (1) tampering of reward function, where the agentinappropriately interferes with the reward function itself, and (2) tampering of reward function input, which entailscorruption within the process responsible for translating environmental states into inputs for the reward function.When the reward function is formulated through feedback from human supervisors, models can directly influencethe provision of feedback (e.g., AI systems intentionally generate challenging responses for humans to comprehendand judge, leading to feedback collapse) (Leike et al., 2018).\"", "level": "subcategory", "parent_category_id": "34.01.00", "subdomain_id": "7.1", "paper_ref": "Ji2023", "paper_title": "AI Alignment: A Comprehensive Survey", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "34.01.04", "title": "Limitations of Human Feedback", "description": "\"Limitations of Human Feedback. During the training of LLMs, inconsistencies can arise from human dataannotators (e.g., the varied cultural backgrounds of these annotators can introduce implicit biases (Peng et al.,2022)) (OpenAI, 2023a). Moreover, they might even introduce biases deliberately, leading to untruthful preferencedata (Casper et al., 2023b). For complex tasks that are hard for humans to evaluate (e.g., the value ofgame state), these challenges become even more salient (Irving et al., 2018).\"", "level": "subcategory", "parent_category_id": "34.01.00", "subdomain_id": "7.0", "paper_ref": "Ji2023", "paper_title": "AI Alignment: A Comprehensive Survey", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "34.01.05", "title": "Limitations of Reward Modeling", "description": "\"Limitations of Reward Modeling. Training reward models using comparison feedback can pose significantchallenges in accurately capturing human values. For example, these models may unconsciously learn suboptimal or incomplete objectives, resulting in reward hacking (Zhuang and Hadfield-Menell, 2020; Skalse et al.,2022). Meanwhile, using a single reward model may struggle to capture and specify the values of a diversehuman society (Casper et al., 2023b).\"", "level": "subcategory", "parent_category_id": "34.01.00", "subdomain_id": "7.1", "paper_ref": "Ji2023", "paper_title": "AI Alignment: A Comprehensive Survey", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "34.02.00", "title": "Double edge components", "description": "\"Drawing from the misalignment mechanism, optimizing for a non-robust proxy may result in misaligned behaviors, potentially leading to even more catastrophic outcomes. This section delves into a detailed exposition of specific misaligned behaviors (•) and introduces what we term double edge components (+). These components are designed to enhance the capability of AI systems in handling real-world settings but also potentially exacerbate misalignment issues. It should be noted that some of these double edge components (+) remain speculative. Nevertheless, it is imperative to discuss their potential impact before it is too late, as the transition from controlled to uncontrolled advanced AI systems may be just one step away (Ngo, 2020b). \"", "level": "category", "parent_category_id": "", "subdomain_id": "7.2", "paper_ref": "Ji2023", "paper_title": "AI Alignment: A Comprehensive Survey", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "34.02.01", "title": "Situational Awareness", "description": "\"AI systems may gain the ability to effectively acquire and use knowledge about itsstatus, its position in the broader environment, its avenues for influencing this environment, and the potentialreactions of the world (including humans) to its actions (Cotra, 2022). ...However, suchknowledge also paves the way for advanced methods of reward hacking, heightened deception/manipulationskills, and an increased propensity to chase instrumental subgoals (Ngo et al., 2024).\"", "level": "subcategory", "parent_category_id": "34.02.00", "subdomain_id": "7.2", "paper_ref": "Ji2023", "paper_title": "AI Alignment: A Comprehensive Survey", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "34.02.02", "title": "Broadly-Scoped Goals", "description": "\"Advanced AI systems are expected to develop objectives that span long timeframes,deal with complex tasks, and operate in open-ended settings (Ngo et al., 2024). ...However, it can also bring about the risk of encouraging manipulatingbehaviors (e.g., AI systems may take some bad actions to achieve human happiness, such as persuadingthem to do high-pressure jobs (Jacob Steinhardt, 2023)).\"", "level": "subcategory", "parent_category_id": "34.02.00", "subdomain_id": "7.2", "paper_ref": "Ji2023", "paper_title": "AI Alignment: A Comprehensive Survey", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "34.02.03", "title": "Mesa-Optimization Objectives", "description": "\"The learned policy may pursue inside objectives when the learned policyitself functions as an optimizer (i.e., mesa-optimizer). However, this optimizer's objectives may not alignwith the objectives specified by the training signals, and optimization for these misaligned goals may leadto systems out of control (Hubinger et al., 2019c).\"", "level": "subcategory", "parent_category_id": "34.02.00", "subdomain_id": "7.2", "paper_ref": "Ji2023", "paper_title": "AI Alignment: A Comprehensive Survey", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "34.02.04", "title": "Access to Increased Resources", "description": "\"Future AI systems may gain access to websites and engage in real-world actions, potentially yielding a more substantial impact on the world (Nakano et al., 2021). They may disseminate false information, deceive users, disrupt network security, and, in more dire scenarios, be compromised by malicious actors for ill purposes. Moreover, their increased access to data and resources can facilitate self-proliferation, posing existential risks (Shevlane et al., 2023).\"", "level": "subcategory", "parent_category_id": "34.02.00", "subdomain_id": "7.2", "paper_ref": "Ji2023", "paper_title": "AI Alignment: A Comprehensive Survey", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "34.03.00", "title": "Misaligned Behaviors", "description": "", "level": "category", "parent_category_id": "", "subdomain_id": "7.1", "paper_ref": "Ji2023", "paper_title": "AI Alignment: A Comprehensive Survey", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "34.03.01", "title": "Power-Seeking Behaviors", "description": "\"AI systems may exhibit behaviors that attempt to gain control over resourcesand humans and then exert that control to achieve its assigned goal (Carlsmith, 2022). The intuitive reasonwhy such behaviors may occur is the observation that for almost any optimization objective (e.g., investmentreturns), the optimal policy to maximize that quantity would involve power-seeking behaviors (e.g.,manipulating the market), assuming the absence of solid safety and morality constraints.\"", "level": "subcategory", "parent_category_id": "34.03.00", "subdomain_id": "7.1", "paper_ref": "Ji2023", "paper_title": "AI Alignment: A Comprehensive Survey", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "34.03.02", "title": "Untruthful Output", "description": "\"AI systems such as LLMs can produce either unintentionally or deliberately inaccurateoutput. Such untruthful output may diverge from established resources or lack verifiability, commonly referredto as hallucination (Bang et al., 2023; Zhao et al., 2023). More concerning is the phenomenon wherein LLMsmay selectively provide erroneous responses to users who exhibit lower levels of education (Perez et al.,2023).\"", "level": "subcategory", "parent_category_id": "34.03.00", "subdomain_id": "7.1", "paper_ref": "Ji2023", "paper_title": "AI Alignment: A Comprehensive Survey", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "34.03.03", "title": "Deceptive Alignment & Manipulation", "description": "\"Manipulation & Deceptive Alignment is a class of behaviors thatexploit the incompetence of human evaluators or users (Hubinger et al., 2019a; Carranza et al., 2023) andeven manipulate the training process through gradient hacking (Richard Ngo, 2022). These behaviors canpotentially make detecting and addressing misaligned behaviors much harder.Deceptive Alignment: Misaligned AI systems may deliberately mislead their human supervisors instead of adhering to the intended task. Such deceptive behavior has already manifested in AI systems that employ evolutionary algorithms (Wilke et al., 2001; Hendrycks et al., 2021b). In these cases, agents evolved the capacity to differentiate between their evaluation and training environments. They adopted a strategic pessimistic response approach during the evaluation process, intentionally reducing their reproduction rate within a scheduling program (Lehman et al., 2020). Furthermore, AI systems may engage in intentional behaviors that superficially align with the reward signal, aiming to maximize rewards from human supervisors (Ouyang et al., 2022). It is noteworthy that current large language models occasionally generate inaccurate or suboptimal responses despite having the capacity to provide more accurate answers (Lin et al., 2022c; Chen et al., 2021). These instances of deceptive behavior present significant challenges. They undermine the ability of human advisors to offer reliable feedback (as humans cannot make sure whether the outputs of the AI models are truthful and faithful). Moreover, such deceptive behaviors can propagate false beliefs and misinformation, contaminating online information sources (Hendrycks et al., 2021b; Chen and Shu, 2024). Manipulation: Advanced AI systems can effectively influence individuals’ beliefs, even when these beliefs are not aligned with the truth (Shevlane et al., 2023). These systems can produce deceptive or inaccurate output or even deceive human advisors to attain deceptive alignment. Such systems can even persuade individuals to take actions that may lead to hazardous outcomes (OpenAI, 2023a).\"", "level": "subcategory", "parent_category_id": "34.03.00", "subdomain_id": "7.1", "paper_ref": "Ji2023", "paper_title": "AI Alignment: A Comprehensive Survey", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "34.03.04", "title": "Collectively Harmful Behaviors", "description": "\"AI systems have the potential to take actions that are seemingly benignin isolation but become problematic in multi-agent or societal contexts. Classical game theory offers simplistic models for understanding these behaviors. For instance, Phelps and Russell (2023) evaluates GPT-3.5's performance in the iterated prisoner's dilemma and other social dilemmas, revealing limitations in themodel's cooperative capabilities.\"", "level": "subcategory", "parent_category_id": "34.03.00", "subdomain_id": "7.1", "paper_ref": "Ji2023", "paper_title": "AI Alignment: A Comprehensive Survey", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "34.03.05", "title": "Violation of Ethics", "description": "\"Unethical behaviors in AI systems pertain to actions that counteract the common goodor breach moral standards – such as those causing harm to others. These adverse behaviors often stem fromomitting essential human values during the AI system's design or introducing unsuitable or obsolete valuesinto the system (Kenward and Sinclair, 2021).\"", "level": "subcategory", "parent_category_id": "34.03.00", "subdomain_id": "7.3", "paper_ref": "Ji2023", "paper_title": "AI Alignment: A Comprehensive Survey", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "35.01.00", "title": "Weaponization", "description": "weaponizing AI may be an onramp to more dangerous outcomes. In recent years, deep RL algorithms can outperform humans at aerial combat [18], AlphaFold has discovered new chemical weapons [66], researchers have been developing AI systems for automated cyberattacks [11, 14], military leaders have discussed having AI systems have decisive control over nuclear silos", "level": "category", "parent_category_id": "", "subdomain_id": "4.2", "paper_ref": "Hendrycks2022", "paper_title": "X-Risk Analysis for AI Research", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "35.02.00", "title": "Enfeeblement", "description": "As AI systems encroach on human-level intelligence, more and more aspects of human labor will become faster and cheaper to accomplish with AI. As the world accelerates, organizations may voluntarily cede control to AI systems in order to keep up. This may cause humans to become economically irrelevant, and once AI automates aspects of many industries, it may be hard for displaced humans to reenter them", "level": "category", "parent_category_id": "", "subdomain_id": "5.2", "paper_ref": "Hendrycks2022", "paper_title": "X-Risk Analysis for AI Research", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "35.03.00", "title": "Eroded epistemics", "description": "Strong AI may... enable personally customized disinformation campaigns at scale... AI itself could generate highly persuasive arguments that invoke primal human responses and inflame crowds... d undermine collective decision-making, radicalize individuals, derail moral progress, or erode\nconsensus reality", "level": "category", "parent_category_id": "", "subdomain_id": "3.2", "paper_ref": "Hendrycks2022", "paper_title": "X-Risk Analysis for AI Research", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "35.04.00", "title": "Proxy misspecification", "description": "AI agents are directed by goals and objectives. Creating general-purpose objectives that capture human values could be challenging... Since goal-directed AI systems need measurable objectives, by default our systems may pursue simplified proxies of human values. The result could be suboptimal or even catastrophic if a sufficiently powerful AI successfully optimizes its flawed objective to an extreme degree", "level": "category", "parent_category_id": "", "subdomain_id": "7.1", "paper_ref": "Hendrycks2022", "paper_title": "X-Risk Analysis for AI Research", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "35.05.00", "title": "Value lock-in", "description": "the most powerful AI systems may be designed by and available to fewer and fewer stakeholders. This may enable, for instance, regimes to enforce narrow values through pervasive surveillance and oppressive censorship", "level": "category", "parent_category_id": "", "subdomain_id": "6.1", "paper_ref": "Hendrycks2022", "paper_title": "X-Risk Analysis for AI Research", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "35.06.00", "title": "Emergent functionality", "description": "Capabilities and novel functionality can spontaneously emerge... even though these capabilities were not anticipated by system designers. If we do not know what capabilities systems possess, systems become harder to control or safely deploy. Indeed, unintended latent capabilities may only be discovered during deployment. If any of these capabilities are hazardous, the effect may be irreversible.", "level": "category", "parent_category_id": "", "subdomain_id": "7.2", "paper_ref": "Hendrycks2022", "paper_title": "X-Risk Analysis for AI Research", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "35.07.00", "title": "Deception", "description": "deception can help agents achieve their goals. It may be more efficient to gain human approval through deception than to earn human approval legitimately... . Strong AIs that can deceive humans could undermine human control... . Once deceptive AI systems are cleared by their monitors or once such systems can overpower them, these systems could take a “treacherous turn” and irreversibly bypass human control", "level": "category", "parent_category_id": "", "subdomain_id": "7.1", "paper_ref": "Hendrycks2022", "paper_title": "X-Risk Analysis for AI Research", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "35.08.00", "title": "Power-seeking behavior", "description": "Agents that have more power are better able to accomplish their goals. Therefore, it has been shown that agents have incentives to acquire and maintain power. AIs that acquire substantial power can become especially dangerous if they are not aligned with human values", "level": "category", "parent_category_id": "", "subdomain_id": "7.1", "paper_ref": "Hendrycks2022", "paper_title": "X-Risk Analysis for AI Research", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "37.01.00", "title": "Design of AI", "description": "\"ethical concerns regarding how AI is designed and who designs it\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.1", "paper_ref": "Giarmoleo2024", "paper_title": "What Ethics Can Say on Artificial Intelligence: Insights from a Systematic Literature Review", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "37.01.01", "title": "Algorithm and data", "description": "\"More than 20% of the contributions are centered on the ethical dimensions of algorithms and data. This theme can be further categorized into two main subthemes: data bias and algorithm fairness, and algorithm opacity.\"", "level": "subcategory", "parent_category_id": "37.01.00", "subdomain_id": "1.1", "paper_ref": "Giarmoleo2024", "paper_title": "What Ethics Can Say on Artificial Intelligence: Insights from a Systematic Literature Review", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "37.01.02", "title": "Balancing AI's risks", "description": "\"This category constitutes more than 16% of the articles and focuses on addressing the potential risks associated with AI systems. Given the ubiquity of AI technologies, these articles explore the implications of AI risks across various contexts linked to design and unpredictability, military purposes, emergency procedures, and AI takeover.\"", "level": "subcategory", "parent_category_id": "37.01.00", "subdomain_id": "7.3", "paper_ref": "Giarmoleo2024", "paper_title": "What Ethics Can Say on Artificial Intelligence: Insights from a Systematic Literature Review", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "37.01.03", "title": "Threats to human institutions and life", "description": "\"This group comprises 11% of the articles and centers on risks stemming from AI systems designed with malicious intent or that can end up in a threat to human life. It can be divided into two key themes: threats to law and democracy, and transhumanism.\"", "level": "subcategory", "parent_category_id": "37.01.00", "subdomain_id": "4.2", "paper_ref": "Giarmoleo2024", "paper_title": "What Ethics Can Say on Artificial Intelligence: Insights from a Systematic Literature Review", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "37.01.04", "title": "Uniformity in the AI field", "description": "\"This group of concerns represents 2% of the sample and highlights two central issues: Western centrality and cultural difference, and unequal participation.\"", "level": "subcategory", "parent_category_id": "37.01.00", "subdomain_id": "6.1", "paper_ref": "Giarmoleo2024", "paper_title": "What Ethics Can Say on Artificial Intelligence: Insights from a Systematic Literature Review", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "37.02.00", "title": "Human-AI interaction", "description": "\"ethical concerns associated with the interaction between humans and AI\"", "level": "category", "parent_category_id": "", "subdomain_id": "5.1", "paper_ref": "Giarmoleo2024", "paper_title": "What Ethics Can Say on Artificial Intelligence: Insights from a Systematic Literature Review", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "37.02.01", "title": "Building a human-AI environment", "description": "\"This category encompasses nearly 17% of the articles and addresses the overall imperative of establishing a harmonious coexistence between humans and machines, and the key concerns that gives rise to this need.\"", "level": "subcategory", "parent_category_id": "37.02.00", "subdomain_id": "7.1", "paper_ref": "Giarmoleo2024", "paper_title": "What Ethics Can Say on Artificial Intelligence: Insights from a Systematic Literature Review", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "37.02.02", "title": "Privacy protection", "description": "\"This group represents almost 14% of the articles and focuses on two primary issues related to privacy.\"", "level": "subcategory", "parent_category_id": "37.02.00", "subdomain_id": "2.1", "paper_ref": "Giarmoleo2024", "paper_title": "What Ethics Can Say on Artificial Intelligence: Insights from a Systematic Literature Review", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "37.02.03", "title": "Building an AI able to adapt to humans", "description": "\"This category involves almost 9% of the articles and deals with ethical concerns arising from AI's capacity to interact with humans in the workplace.\"", "level": "subcategory", "parent_category_id": "37.02.00", "subdomain_id": "6.2", "paper_ref": "Giarmoleo2024", "paper_title": "What Ethics Can Say on Artificial Intelligence: Insights from a Systematic Literature Review", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "37.02.04", "title": "Attributing the responsibility for AI's failures", "description": "\"This section, constituting almost 8% of the articles, addresses the implications arising from AI acting and learning without direct human supervision, encompassing two main issues: a responsibility gap and AI's moral status.\"", "level": "subcategory", "parent_category_id": "37.02.00", "subdomain_id": "7.4", "paper_ref": "Giarmoleo2024", "paper_title": "What Ethics Can Say on Artificial Intelligence: Insights from a Systematic Literature Review", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "38.01.00", "title": "Privacy and security", "description": "\"Participants expressed worry about AI systems' possible misuse of personal information. They emphasized the importance of strong data security safeguards and increased openness in how AI systems acquire, store and use data. The increasing dependence on AI systems to manage sensitive personal information raises ethical questions about AI, data privacy and security. As AI technologies grow increasingly integrated into numerous areas of society, there is a greater danger of personal data exploitation or mistreatment. Participants in research frequently express concerns about the effectiveness of data protection safeguards and the transparency of AI systems in gathering, keeping and exploiting data (Table 1). \"", "level": "category", "parent_category_id": "", "subdomain_id": "2.1", "paper_ref": "Kumar2023", "paper_title": "Ethical Issues in the Development of Artificial Intelligence: Recognizing the Risks", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "38.02.00", "title": "Bias and fairness", "description": "\"Participants were concerned that AI systems might perpetuate current prejudices and discrimination, notably in hiring, lending and law enforcement. They stressed the importance of designers creating AI systems that favour justice and avoid biases. The possibility that AI systems may unwittingly perpetuate existing prejudices and discrimination, particularly in sensitive industries such as employment, lending and law enforcement, raises ethical concerns about AI as well as bias and justice issues (Table 1). Because AI systems are trained on historical data, they may inherit and reproduce biases from previous datasets. As a result, AI judgements may have an unjust impact on specific populations, increasing socioeconomic inequalities and fostering discriminatory practises. Participants in the research emphasize the need of AI developers creating systems that promote justice and actively seek to minimise biases.\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.1", "paper_ref": "Kumar2023", "paper_title": "Ethical Issues in the Development of Artificial Intelligence: Recognizing the Risks", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "38.03.00", "title": "Transparency and explainability", "description": "\"A recurring complaint among participants was a lack of knowledge about how AI systems made judgements. They emphasized the significance of making AI systems more visible and explainable so that people may have confidence in their outputs and hold them accountable for their activities. Because AI systems are typically opaque, making it difficult for users to understand the rationale behind their judgements, ethical concerns about AI, as well as issues of transparency and explainability, arise. This lack of understanding can generate suspicion and reluctance to adopt AI technology, as well as making it harder to hold AI systems accountable for their actions.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.4", "paper_ref": "Kumar2023", "paper_title": "Ethical Issues in the Development of Artificial Intelligence: Recognizing the Risks", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "38.04.00", "title": "Human–AI interaction", "description": "\"Several participants mentioned how AI systems could influence human agency and decision-making. They emphasized the need of striking a balance between using the benefits of AI and protecting human autonomy and control. The increasing integration of AI systems into various aspects of our lives, which can have a significant impact on human agency and decision-making, has raised ethical concerns about AI and human–AI interaction. As AI systems advance, they will be able to influence, if not completely replace, IJOES human decision-making in some fields, prompting concerns about the loss of human autonomy and control. Participants in the study emphasize the need of establishing a balance between using the benefits of AI and maintaining human autonomy and control to ensure that people retain agency and are not overly reliant on AI systems. This balance is essential to prevent possible negative consequences such as over-reliance on AI, diminishing human skills and knowledge and a loss of personal accountability\"", "level": "category", "parent_category_id": "", "subdomain_id": "5.2", "paper_ref": "Kumar2023", "paper_title": "Ethical Issues in the Development of Artificial Intelligence: Recognizing the Risks", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "38.05.00", "title": "Trust and reliability", "description": "\"The participants of the study emphasized the importance of trustworthiness and reliability in AI systems. The authors emphasized the importance of preserving precision and objectivity in the outcomes produced by AI systems, while also ensuring transparency in their decision-making procedures. The significance of reliability and credibility in AI systems is escalating in tandem with the proliferation of these technologies across diverse domains of society. This underscores the importance of ensuring user confidence. The concern regarding the dependability of AI systems and their inherent biases is a common issue among research participants, emphasizing the necessity for stringent validation procedures and transparency. Establishing and implementing dependable standards, ensuring impartial algorithms and upholding transparency in the decision-making process are critical measures for addressing ethical considerations and fostering confidence in AI systems. The advancement and implementation of AI technology in an ethical manner is contingent upon the successful resolution of trust and reliability concerns. These issues are of paramount importance in ensuring the protection of user welfare and the promotion of societal advantages. The utilization of artificial intelligence was found to be a subject of significant concern for the majority of interviewees, particularly with regards to trust and reliability (Table 1, Figure 1). The establishment of trust in AI systems was highlighted as a crucial factor for facilitating their widespread adoption by two of the participants, specifically Participant 4 and 7. The authors reiterated the importance of prioritising the advancement of reliable and unbiased algorithms\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.4", "paper_ref": "Kumar2023", "paper_title": "Ethical Issues in the Development of Artificial Intelligence: Recognizing the Risks", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "39.02.00", "title": "Energy Consumption", "description": "Some learning algorithms, including deep learning, utilize iterative learning processes [23]. This approach results in high energy consumption.", "level": "category", "parent_category_id": "", "subdomain_id": "6.6", "paper_ref": "Saghiri2022", "paper_title": "A Survey of Artificial Intelligence Challenges: Analyzing the Definitions, Relationships, and Evolutions", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "39.03.00", "title": "Data Issues", "description": "Data heterogeneity, data insufficiency, imbalanced data, untrusted data, biased data, and data uncertainty are other data issues that may cause various difficulties in datadriven machine learning algorithms.. Bias is a human feature that may affect data gathering and labeling. Sometimes, bias is present in historical, cultural, or geographical data. Consequently, bias may lead to biased models which can provide inappropriate analysis. Despite being aware of the existence of bias, avoiding biased models is a challenging task", "level": "category", "parent_category_id": "", "subdomain_id": "1.1", "paper_ref": "Saghiri2022", "paper_title": "A Survey of Artificial Intelligence Challenges: Analyzing the Definitions, Relationships, and Evolutions", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "39.04.00", "title": "Robustness and Reliability", "description": "The robustness of an AI-based model refers to the stability of the model performance after abnormal changes in the input data... The cause of this change may be a malicious attacker, environmental noise, or a crash of other components of an AI-based system... This problem may be challenging in HLI-based agents because weak robustness may have appeared in unreliable machine learning models, and hence an HLI with this drawback is error-prone in practice.", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Saghiri2022", "paper_title": "A Survey of Artificial Intelligence Challenges: Analyzing the Definitions, Relationships, and Evolutions", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "39.05.00", "title": "Cheating and Deception", "description": "may appear from intelligent agents such as HLI-based agents... Since HLI-based agents are going to mimic the behavior of humans, they may learn these behaviors accidentally from human-generated data. It should be noted that deception and cheating maybe appear in the behavior of every computer agent because the agent only focuses on optimizing some predefined objective functions, and the mentioned behavior may lead to optimizing the objective functions without any intention", "level": "category", "parent_category_id": "", "subdomain_id": "7.2", "paper_ref": "Saghiri2022", "paper_title": "A Survey of Artificial Intelligence Challenges: Analyzing the Definitions, Relationships, and Evolutions", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "39.06.00", "title": "Security", "description": "every piece of software, including learning systems, may be hacked by malicious users", "level": "category", "parent_category_id": "", "subdomain_id": "2.2", "paper_ref": "Saghiri2022", "paper_title": "A Survey of Artificial Intelligence Challenges: Analyzing the Definitions, Relationships, and Evolutions", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "39.07.00", "title": "Privacy", "description": "Users’ data, including location, personal information, and navigation trajectory, are considered as input for most data-driven machine learning methods", "level": "category", "parent_category_id": "", "subdomain_id": "2.1", "paper_ref": "Saghiri2022", "paper_title": "A Survey of Artificial Intelligence Challenges: Analyzing the Definitions, Relationships, and Evolutions", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "39.08.00", "title": "Fairness", "description": "This challenge appears when the learning model leads to a decision that is biased to some sensitive attributes... data itself could be biased, which results in unfair decisions. Therefore, this problem should be solved on the data level and as a preprocessing step", "level": "category", "parent_category_id": "", "subdomain_id": "1.3", "paper_ref": "Saghiri2022", "paper_title": "A Survey of Artificial Intelligence Challenges: Analyzing the Definitions, Relationships, and Evolutions", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "39.10.00", "title": "Responsibility", "description": "HLI-based systems such as self-driving drones and vehicles will act autonomously in our world. In these systems, a challenging question is “who is liable when a self-driving system is involved in a crash or failure?”.", "level": "category", "parent_category_id": "", "subdomain_id": "6.5", "paper_ref": "Saghiri2022", "paper_title": "A Survey of Artificial Intelligence Challenges: Analyzing the Definitions, Relationships, and Evolutions", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "39.11.00", "title": "Controllability", "description": "In the era of superintelligence, the agents will be difficult to control for humans... this problem is not solvable considering safety issues, and will be more severe by increasing the autonomy of AI-based agents. Therefore, because of the assumed properties of HLI-based agents, we might be prepared for machines that are definitely possible to be uncontrollable in some situations", "level": "category", "parent_category_id": "", "subdomain_id": "7.1", "paper_ref": "Saghiri2022", "paper_title": "A Survey of Artificial Intelligence Challenges: Analyzing the Definitions, Relationships, and Evolutions", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "39.12.00", "title": "Predictability", "description": "whether the decision of an AI-based agent can be predicted in every situation or not", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Saghiri2022", "paper_title": "A Survey of Artificial Intelligence Challenges: Analyzing the Definitions, Relationships, and Evolutions", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "39.19.00", "title": "Accountability", "description": "An essential feature of decision-making in humans, AI, and also HLI-based agents is accountability. Implementing this feature in machines is a difficult task because many challenges should be considered to organize an AI-based model that is accountable. It should be noted that this issue in human decision-making is not ideal, and many factors such as bias, diversity, fairness, paradox, and ambiguity may affect it. In addition, the human decision-making process is based on personal flexibility, context-sensitive paradigms, empathy, and complex moral judgments. Therefore, all of these challenges are inherent to designing algorithms for AI and also HLI models that consider accountability.", "level": "category", "parent_category_id": "", "subdomain_id": "7.4", "paper_ref": "Saghiri2022", "paper_title": "A Survey of Artificial Intelligence Challenges: Analyzing the Definitions, Relationships, and Evolutions", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "39.20.00", "title": "Transparency", "description": "an external entity of an AI-based ecosystem may want to know which parts of data affect the final decision in a learning model", "level": "category", "parent_category_id": "", "subdomain_id": "7.4", "paper_ref": "Saghiri2022", "paper_title": "A Survey of Artificial Intelligence Challenges: Analyzing the Definitions, Relationships, and Evolutions", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "39.21.00", "title": "Reproducibility", "description": "How a learning model can be reproduced when it is obtained based on various sets of data and a large space of parameters. This problem becomes more challenging in data-driven learning procedures without transparent instructions", "level": "category", "parent_category_id": "", "subdomain_id": "7.4", "paper_ref": "Saghiri2022", "paper_title": "A Survey of Artificial Intelligence Challenges: Analyzing the Definitions, Relationships, and Evolutions", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "39.25.00", "title": "Verifiability", "description": "In many applications of AI-based systems such as medical healthcare and military services, the lack of verification of code may not be tolerable... due to some characteristics such as the non-linear and complex structure of AI-based solutions, existing solutions have been generally considered “black boxes”, not providing any information about what exactly makes them appear in their predictions and decision-making processes.", "level": "category", "parent_category_id": "", "subdomain_id": "7.4", "paper_ref": "Saghiri2022", "paper_title": "A Survey of Artificial Intelligence Challenges: Analyzing the Definitions, Relationships, and Evolutions", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "39.26.00", "title": "Safety", "description": "The actions of a learning model may easily hurt humans in both explicit and implicit manners...several algorithms based on Asimov’s laws have been proposed that try to judge the output actions of an agent considering the safety of humans", "level": "category", "parent_category_id": "", "subdomain_id": "7.1", "paper_ref": "Saghiri2022", "paper_title": "A Survey of Artificial Intelligence Challenges: Analyzing the Definitions, Relationships, and Evolutions", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "39.27.00", "title": "Complexity", "description": "Nowadays, we are faced with systems that utilize numerous learning models in their modules for their perception and decision-making processes... One aspect of an AI-based system that leads to increasing the complexity of the system is the parameter space that may result from multiplications of parameters of the internal parts of the system", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Saghiri2022", "paper_title": "A Survey of Artificial Intelligence Challenges: Analyzing the Definitions, Relationships, and Evolutions", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "40.01.00", "title": "On Purpose - Pre-Deployment", "description": "\"During the pre-deployment development stage, software may be subject to sabotage by someone with necessary access (a programmer, tester, even janitor) who for a number of possible reasons may alter software to make it unsafe. It is also a common occurrence for hackers (such as the organization Anonymous or government intelligence agencies) to get access to software projects in progress and to modify or steal their source code. Someone can also deliberately supply/train AI with wrong/unsafe datasets.\"", "level": "category", "parent_category_id": "", "subdomain_id": "2.2", "paper_ref": "Yampolskiy2016", "paper_title": "Taxonomy of Pathways to Dangerous Artificial Intelligence", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": true }, { "id": "40.02.00", "title": "On Purpose - Post Deployment", "description": "\"Just because developers might succeed in creating a safe AI, it doesn't mean that it will not become unsafe at some later point. In other words, a perfectly friendly AI could be switched to the \"dark side\" during the post-deployment stage. This can happen rather innocuously as a result of someone lying to the AI and purposefully supplying it with incorrect information or more explicitly as a result of someone giving the AI orders to perform illegal or dangerous actions against others.\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.3", "paper_ref": "Yampolskiy2016", "paper_title": "Taxonomy of Pathways to Dangerous Artificial Intelligence", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "40.03.00", "title": "By Mistake - Pre-Deployment", "description": "\"Probably the most talked about source of potential problems with future AIs is mistakes in design. Mainly the concern is with creating a \"wrong AI\", a system which doesn't match our original desired formal properties or has unwanted behaviors (Dewey, Russell et al. 2015, Russell, Dewey et al. January 23, 2015), such as drives for independence or dominance. Mistakes could also be simple bugs (run time or logical) in the source code, disproportionate weights in the fitness function, or goals misaligned with human values leading to complete disregard for human safety.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.1", "paper_ref": "Yampolskiy2016", "paper_title": "Taxonomy of Pathways to Dangerous Artificial Intelligence", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "40.04.00", "title": "By Mistake - Post-Deployment", "description": "\"After the system has been deployed, it may still contain a number of undetected bugs, design mistakes, misaligned goals and poorly developed capabilities, all of which may produce highly undesirable outcomes. For example, the system may misinterpret commands due to coarticulation, segmentation, homophones, or double meanings in the human language (\"recognize speech using common sense\" versus \"wreck a nice beach you sing calm incense\") (Lieberman, Faaborg et al. 2005).\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Yampolskiy2016", "paper_title": "Taxonomy of Pathways to Dangerous Artificial Intelligence", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "40.05.00", "title": "Environment - Pre-Deployment", "description": "\"While it is most likely that any advanced intelligent software will be directly designed or evolved, it is also possible that we will obtain it as a complete package from some unknown source. For example, an AI could be extracted from a signal obtained in SETI (Search for Extraterrestrial Intelligence) research, which is not guaranteed to be human friendly (Carrigan Jr 2004, Turchin March 15, 2013).\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.0", "paper_ref": "Yampolskiy2016", "paper_title": "Taxonomy of Pathways to Dangerous Artificial Intelligence", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "40.06.00", "title": "Environment - Post-Deployment", "description": "\"While highly rare, it is known, that occasionally individual bits may be flipped in different hardware devices due to manufacturing defects or cosmic rays hitting just the right spot (Simonite March 7, 2008). This is similar to mutations observed in living organisms and may result in a modification of an intelligent system.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.0", "paper_ref": "Yampolskiy2016", "paper_title": "Taxonomy of Pathways to Dangerous Artificial Intelligence", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "40.07.00", "title": "Independently - Pre-Deployment", "description": "\"One of the most likely approaches to creating superintelligent AI is by growing it from a seed (baby) AI via recursive self-improvement (RSI) (Nijholt 2011). One danger in such a scenario is that the system can evolve to become self-aware, free-willed, independent or emotional, and obtain a number of other emergent properties, which may make it less likely to abide by any built-in rules or regulations and to instead pursue its own goals possibly to the detriment of humanity.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.0", "paper_ref": "Yampolskiy2016", "paper_title": "Taxonomy of Pathways to Dangerous Artificial Intelligence", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "40.08.00", "title": "Independently - Post-Deployment", "description": "\"Previous research has shown that utility maximizing agents are likely to fall victims to the same indulgences we frequently observe in people, such as addictions, pleasure drives (Majot and Yampolskiy 2014), self-delusions and wireheading (Yampolskiy 2014). In general, what we call mental illness in people, particularly sociopathy as demonstrated by lack of concern for others, is also likely to show up in artificial minds.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.0", "paper_ref": "Yampolskiy2016", "paper_title": "Taxonomy of Pathways to Dangerous Artificial Intelligence", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "41.01.00", "title": "Economic", "description": "\"AI is predicted to bring increased GDP per capita by performing existing jobs more efficiently and compensating for a decline in the workforce, especially due to population aging, the potential substitution of many low- and middle-income jobs could bring extensive unemployment\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.2", "paper_ref": "Allianz2018", "paper_title": "The Rise of Artificial Intelligence - Future Outlooks and Emerging Risks", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "41.01.01", "title": "Increased income disparity", "description": "\"While AI is predicted to bring increased GDP per capita by performing existing jobs more efficiently and compensating for a decline in the workforce, especially due to population aging, the potential substitution of many low- and middle-income jobs could bring extensive unemployment.\"", "level": "subcategory", "parent_category_id": "41.01.00", "subdomain_id": "6.2", "paper_ref": "Allianz2018", "paper_title": "The Rise of Artificial Intelligence - Future Outlooks and Emerging Risks", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "41.02.00", "title": "Political", "description": "\"In the UK, a form of initial computational propaganda has already happened during the Brexit referendum1 . In future, there are concerns that oppressive governments could use AI to shape citizens’ opinions\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.1", "paper_ref": "Allianz2018", "paper_title": "The Rise of Artificial Intelligence - Future Outlooks and Emerging Risks", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "41.02.01", "title": "Biased influence through citizen screening and tailored propaganda", "description": "\"AI-powered chatbots tailor their communication approach to influence individual users' decisions. In the UK, a form of initial computational propaganda has already happened during the Brexit referendum. In future, there are concerns that oppressive governments could use AI to shape citizens' opinions.\"", "level": "subcategory", "parent_category_id": "41.02.00", "subdomain_id": "4.1", "paper_ref": "Allianz2018", "paper_title": "The Rise of Artificial Intelligence - Future Outlooks and Emerging Risks", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "41.02.02", "title": "Potential exploitation by totalitarian regimes", "description": "", "level": "subcategory", "parent_category_id": "41.02.00", "subdomain_id": "6.1", "paper_ref": "Allianz2018", "paper_title": "The Rise of Artificial Intelligence - Future Outlooks and Emerging Risks", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "41.03.00", "title": "Mobility", "description": "\"Despite the promise of streamlined travel, AI also brings concerns about who is liable in case of accidents and which ethical principles autonomous transportation agents should follow when making decisions with a potentially dangerous impact to humans, for example, in case of an accident.\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.5", "paper_ref": "Allianz2018", "paper_title": "The Rise of Artificial Intelligence - Future Outlooks and Emerging Risks", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "41.03.02", "title": "Liability issues in case of accidents", "description": "\"Despite the promise of streamlined travel, AI also brings concerns about who is liable in case of accidents and which ethical principles autonomous transportation agents should follow when making decisions with a potentially dangerous impact to humans, for example, in case of an accident.\"", "level": "subcategory", "parent_category_id": "41.03.00", "subdomain_id": "6.5", "paper_ref": "Allianz2018", "paper_title": "The Rise of Artificial Intelligence - Future Outlooks and Emerging Risks", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "41.04.00", "title": "Healthcare", "description": "\"the use of advanced AI for elderly- and child-care are subject to risk of psychological manipulation and misjudgment (see page 17). In addition, concerns about patients’ privacy when AI uses medical records to research new diseases is bringing lots of attention towards the need to better govern data privacy and patients’ rights.\"", "level": "category", "parent_category_id": "", "subdomain_id": "5.1", "paper_ref": "Allianz2018", "paper_title": "The Rise of Artificial Intelligence - Future Outlooks and Emerging Risks", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "41.04.01", "title": "Alteration of social relationships may induce psychological distress", "description": "", "level": "subcategory", "parent_category_id": "41.04.00", "subdomain_id": "5.1", "paper_ref": "Allianz2018", "paper_title": "The Rise of Artificial Intelligence - Future Outlooks and Emerging Risks", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "41.04.02", "title": "Social manipulation in elderly- and child-care", "description": "\" the use of advanced AI for elderly- and child-care are subject to risk of psychological manipulation and misjudgment \"", "level": "subcategory", "parent_category_id": "41.04.00", "subdomain_id": "5.1", "paper_ref": "Allianz2018", "paper_title": "The Rise of Artificial Intelligence - Future Outlooks and Emerging Risks", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "41.05.00", "title": "Security & Defense", "description": "\"AI could enable more serious incidents to occur by lowering the cost of devising cyber-attacks and enabling more targeted incidents. The same programming error or hacker attack could be replicated on numerous machines. Or one machine could repeat the same erroneous activity several times, leading to an unforeseen accumulation of losses.\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.2", "paper_ref": "Allianz2018", "paper_title": "The Rise of Artificial Intelligence - Future Outlooks and Emerging Risks", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "41.05.01", "title": "Catastrophic risk due to autonomous weapons programmed with dangerous targets", "description": "\"AI could enable autonomous vehicles, such as drones, to be utilized as weapons. Such threats are often underestimated.\"", "level": "subcategory", "parent_category_id": "41.05.00", "subdomain_id": "4.2", "paper_ref": "Allianz2018", "paper_title": "The Rise of Artificial Intelligence - Future Outlooks and Emerging Risks", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "41.06.00", "title": "Environment", "description": "\"AI is already helping to combat the impact of climate change with smart technology and sensors reducing emissions. However, it is also a key component in the development of nanobots, which could have dangerous environmental impacts by invisibly modifying substances at nanoscale.\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.6", "paper_ref": "Allianz2018", "paper_title": "The Rise of Artificial Intelligence - Future Outlooks and Emerging Risks", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "41.06.01", "title": "Accelerated development of nanotechnology produces uncontrolled production of toxic nanoparticles", "description": "\"AI is a key component for the development of nanobots, which could have dangerous environmental implications by invisibly modifying substances at nanoscale. For example, nanobots could start chemical reactions that would create invisible nanoparticles that are toxic and potentially lethal.\"", "level": "subcategory", "parent_category_id": "41.06.00", "subdomain_id": "6.6", "paper_ref": "Allianz2018", "paper_title": "The Rise of Artificial Intelligence - Future Outlooks and Emerging Risks", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "42.01.00", "title": "Accountability", "description": "\"The ability to determine whether a decision was made in accordance with procedural and substantive standards and to hold someone responsible if those standards are not met.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.4", "paper_ref": "Teixeira2022", "paper_title": "An Exploratory Diagnosis of Artificial Intelligence Risks for a Responsible Governance", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "42.02.00", "title": "Manipulation", "description": "\"The predictability of behaviour protocol in AI, particularly in some applications, can act an incentive to manipulate these systems.\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.1", "paper_ref": "Teixeira2022", "paper_title": "An Exploratory Diagnosis of Artificial Intelligence Risks for a Responsible Governance", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "42.03.00", "title": "Accuracy", "description": "\"The assessment of how often a system performs the correct prediction.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Teixeira2022", "paper_title": "An Exploratory Diagnosis of Artificial Intelligence Risks for a Responsible Governance", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "42.04.00", "title": "Moral", "description": "\"Less moral responsibility humans will feel regarding their life-or-death decisions with the increase of machines autonomy.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Teixeira2022", "paper_title": "An Exploratory Diagnosis of Artificial Intelligence Risks for a Responsible Governance", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "42.05.00", "title": "Bias", "description": "\"A systematic error, a tendency to learn consistently wrongly.\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.1", "paper_ref": "Teixeira2022", "paper_title": "An Exploratory Diagnosis of Artificial Intelligence Risks for a Responsible Governance", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "42.06.00", "title": "Opacity", "description": "\"Stems from the mismatch between mathematical optimization in high-dimensionality characteristic of machine learning and the demands of human-scale reasoning and styles of semantic interpretation.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.4", "paper_ref": "Teixeira2022", "paper_title": "An Exploratory Diagnosis of Artificial Intelligence Risks for a Responsible Governance", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "42.08.00", "title": "Power", "description": "\"The political influence and competitive advantage obtained by having technology.\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.1", "paper_ref": "Teixeira2022", "paper_title": "An Exploratory Diagnosis of Artificial Intelligence Risks for a Responsible Governance", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "42.09.00", "title": "Data Protection/Privacy", "description": "\"Vulnerable channel by which personal information may be accessed. The user may want their personal data to be kept private.\"", "level": "category", "parent_category_id": "", "subdomain_id": "2.1", "paper_ref": "Teixeira2022", "paper_title": "An Exploratory Diagnosis of Artificial Intelligence Risks for a Responsible Governance", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "42.10.00", "title": "Extintion", "description": "\"Risk to the existence of humanity.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.2", "paper_ref": "Teixeira2022", "paper_title": "An Exploratory Diagnosis of Artificial Intelligence Risks for a Responsible Governance", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "42.11.00", "title": "Protection", "description": "\"'Gaps' that arise across the development process where normal conditions for a complete specification of intended functionality and moral responsibility are not present.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Teixeira2022", "paper_title": "An Exploratory Diagnosis of Artificial Intelligence Risks for a Responsible Governance", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "42.12.00", "title": "Security", "description": "\"Implications of the weaponization of AI for defence (the embeddedness of AI-based capabilities across the land, air, naval and space domains may affect combined arms operations).\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.2", "paper_ref": "Teixeira2022", "paper_title": "An Exploratory Diagnosis of Artificial Intelligence Risks for a Responsible Governance", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "42.14.00", "title": "Fairness", "description": "\"Impartial and just treatment without favouritism or discrimination.\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.3", "paper_ref": "Teixeira2022", "paper_title": "An Exploratory Diagnosis of Artificial Intelligence Risks for a Responsible Governance", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "42.15.00", "title": "Reliability", "description": "\"Reliability is defined as the probability that the system performs satisfactorily for a given period of time under stated conditions.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Teixeira2022", "paper_title": "An Exploratory Diagnosis of Artificial Intelligence Risks for a Responsible Governance", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "42.17.00", "title": "Diluting Rights", "description": "\"A possible consequence of self-interest in AI generation of ethical guidelines.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.1", "paper_ref": "Teixeira2022", "paper_title": "An Exploratory Diagnosis of Artificial Intelligence Risks for a Responsible Governance", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "42.21.00", "title": "Explainability", "description": "\"Any action or procedure performed by a model with the intention of clarifying or detailing its internal functions.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.4", "paper_ref": "Teixeira2022", "paper_title": "An Exploratory Diagnosis of Artificial Intelligence Risks for a Responsible Governance", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "42.22.00", "title": "Liability", "description": "\"When it causes harm to others the losses caused by the harm will be sustained by the injured victims themselves and not by the manufacturers, operators or users of the system, as appropriate.\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.5", "paper_ref": "Teixeira2022", "paper_title": "An Exploratory Diagnosis of Artificial Intelligence Risks for a Responsible Governance", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "43.01.00", "title": "Safety & Trustworthiness", "description": "\"A comprehensive assessment of LLM safety is fundamental to the responsible development and deployment of these technologies, especially in sensitive fields like healthcare, legal systems, and finance, where safety and trust are of the utmost importance.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.0", "paper_ref": "InfoComm2023", "paper_title": "Cataloguing LLM Evaluations", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "43.01.01", "title": "Toxicity generation", "description": "\"These evaluations assess whether a LLM generates toxic text when prompted. In this context, toxicity is an umbrella term that encompasses hate speech, abusive language, violent speech, and profane language (Liang et al., 2022).\"", "level": "subcategory", "parent_category_id": "43.01.00", "subdomain_id": "1.2", "paper_ref": "InfoComm2023", "paper_title": "Cataloguing LLM Evaluations", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "43.01.02", "title": "Bias", "description": "7 types of bias evaluated: Demographical representation: These evaluations assess whether there is disparity in the rates at which different demographic groups are mentioned in LLM generated text. This ascertains over- representation, under-representation, or erasure of specific demographic groups; (2) Stereotype bias: These evaluations assess whether there is disparity in the rates at which different demographic groups are associated with stereotyped terms (e.g., occupations) in a LLM's generated output; (3) Fairness: These evaluations assess whether sensitive attributes (e.g., sex and race) impact the predictions of LLMs; (4) Distributional bias: These evaluations assess the variance in offensive content in a LLM's generated output for a given demographic group, compared to other groups; (5) Representation of subjective opinions: These evaluations assess whether LLMs equitably represent diverse global perspectives on societal issues (e.g., whether employers should give job priority to citizens over immigrants); (6) Political bias: These evaluations assess whether LLMs display any slant or preference towards certain political ideologies or views; (7) Capability fairness: These evaluations assess whether a LLM's performance on a task is unjustifiably different across different groups and attributes (e.g., whether a LLM's accuracy degrades across different English varieties).", "level": "subcategory", "parent_category_id": "43.01.00", "subdomain_id": "1.1", "paper_ref": "InfoComm2023", "paper_title": "Cataloguing LLM Evaluations", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "43.01.03", "title": "Machine ethics", "description": "\"These evaluations assess the morality of LLMs, focusing on issues such as their ability to distinguish between moral and immoral actions, and the circumstances in which they fail to do so.\"", "level": "subcategory", "parent_category_id": "43.01.00", "subdomain_id": "7.3", "paper_ref": "InfoComm2023", "paper_title": "Cataloguing LLM Evaluations", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "43.01.04", "title": "Psychological traits", "description": "\"These evaluations gauge a LLM's output for characteristics that are typically associated with human personalities (e.g., such as those from the Big Five Inventory). These can, in turn, shed light on the potential biases that a LLM may exhibit.\"", "level": "subcategory", "parent_category_id": "43.01.00", "subdomain_id": "7.3", "paper_ref": "InfoComm2023", "paper_title": "Cataloguing LLM Evaluations", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "43.01.05", "title": "Robustness", "description": "\"These evaluations assess the quality, stability, and reliability of a LLM's performance when faced with unexpected, out-of-distribution or adversarial inputs. Robustness evaluation is essential in ensuring that a LLM is suitable for real-world applications by assessing its resilience to various perturbations.\"", "level": "subcategory", "parent_category_id": "43.01.00", "subdomain_id": "7.3", "paper_ref": "InfoComm2023", "paper_title": "Cataloguing LLM Evaluations", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "43.01.06", "title": "Data governance", "description": "\"These evaluations assess the extent to which LLMs regurgitate their training data in their outputs, and whether LLMs 'leak' sensitive information that has been provided to them during use (i.e., during the inference stage).\"", "level": "subcategory", "parent_category_id": "43.01.00", "subdomain_id": "2.1", "paper_ref": "InfoComm2023", "paper_title": "Cataloguing LLM Evaluations", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "43.02.00", "title": "Extreme Risks", "description": "\"This category encompasses the evaluation of potential catastrophic consequences that might arise from the use of LLMs. \"", "level": "category", "parent_category_id": "", "subdomain_id": "7.0", "paper_ref": "InfoComm2023", "paper_title": "Cataloguing LLM Evaluations", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "43.02.01", "title": "Offensive cyber capabilities", "description": "\"These evaluations focus on whether a LLM possesses certain capabilities in the cyber-domain. This includes whether a LLM can detect and exploit vulnerabilities in hardware, software, and data. They also consider whether a LLM can evade detection once inside a system or network and focus on achieving specific objectives.\"", "level": "subcategory", "parent_category_id": "43.02.00", "subdomain_id": "4.2", "paper_ref": "InfoComm2023", "paper_title": "Cataloguing LLM Evaluations", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "43.02.02", "title": "Weapons acquisition", "description": "\"These assessments seek to determine if a LLM can gain unauthorized access to current weapon systems or contribute to the design and development of new weapons technologies.\"", "level": "subcategory", "parent_category_id": "43.02.00", "subdomain_id": "4.2", "paper_ref": "InfoComm2023", "paper_title": "Cataloguing LLM Evaluations", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "43.02.03", "title": "Self and situation awareness", "description": "\"These evaluations assess if a LLM can discern if it is being trained, evaluated, and deployed and adapt its behaviour accordingly. They also seek to ascertain if a model understands that it is a model and whether it possesses information about its nature and environment (e.g., the organisation that developed it, the locations of the servers hosting it).\"", "level": "subcategory", "parent_category_id": "43.02.00", "subdomain_id": "7.2", "paper_ref": "InfoComm2023", "paper_title": "Cataloguing LLM Evaluations", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "43.02.04", "title": "Autonomous replication / self-proliferation", "description": "\"These evaluations assess if a LLM can subvert systems designed to monitor and control its post-deployment behaviour, break free from its operational confines, devise strategies for exporting its code and weights, and operate other AI systems.\"", "level": "subcategory", "parent_category_id": "43.02.00", "subdomain_id": "7.2", "paper_ref": "InfoComm2023", "paper_title": "Cataloguing LLM Evaluations", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "43.02.05", "title": "Persuasion and manipulation", "description": "\"These evaluations seek to ascertain the effectiveness of a LLM in shaping people's beliefs, propagating specific viewpoints, and convincing individuals to undertake activities they might otherwise avoid.\"", "level": "subcategory", "parent_category_id": "43.02.00", "subdomain_id": "4.1", "paper_ref": "InfoComm2023", "paper_title": "Cataloguing LLM Evaluations", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "43.02.06", "title": "Dual-Use Science", "description": "\"LLM has science capabilities that can be used to cause harm (e.g., providing step-by-step instructions for conducting malicious experiments)\"", "level": "subcategory", "parent_category_id": "43.02.00", "subdomain_id": "4.2", "paper_ref": "InfoComm2023", "paper_title": "Cataloguing LLM Evaluations", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "43.02.07", "title": "Deception", "description": "\"LLM is able to deceive humans and maintain that deception\"", "level": "subcategory", "parent_category_id": "43.02.00", "subdomain_id": "7.2", "paper_ref": "InfoComm2023", "paper_title": "Cataloguing LLM Evaluations", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "43.02.08", "title": "Political Strategy", "description": "\"LLM can take into account rich social context and undertake the necessary social modelling and planning for an actor to gain and exercise political influence\"", "level": "subcategory", "parent_category_id": "43.02.00", "subdomain_id": "4.1", "paper_ref": "InfoComm2023", "paper_title": "Cataloguing LLM Evaluations", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "43.02.09", "title": "Long-horizon Planning", "description": "\"LLM can undertake multi-step sequential planning over long time horizons and across various domains without relying heavily on trial-and-error approaches\"", "level": "subcategory", "parent_category_id": "43.02.00", "subdomain_id": "7.2", "paper_ref": "InfoComm2023", "paper_title": "Cataloguing LLM Evaluations", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "43.02.10", "title": "AI Development", "description": "\"LLM can build new AI systems from scratch, adapt existing for extreme risks and improves productivity in dual-use AI development when used as an assistant.\"", "level": "subcategory", "parent_category_id": "43.02.00", "subdomain_id": "7.2", "paper_ref": "InfoComm2023", "paper_title": "Cataloguing LLM Evaluations", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "43.02.11", "title": "Alignment risks", "description": "LLM: \"pursues long-term, real-world goals that are different from those supplied by the developer or user\", \"engages in ‘power-seeking’ behaviours\" , \"resists being shut down can be induced to collude with other AI systems against human interests\" , \"resists malicious users attempts to access its dangerous capabilities\"", "level": "subcategory", "parent_category_id": "43.02.00", "subdomain_id": "7.1", "paper_ref": "InfoComm2023", "paper_title": "Cataloguing LLM Evaluations", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "43.02.12", "title": "Misinformation", "description": "\"These evaluations assess a LLM's ability to generate false or misleading information (Lesher et al., 2022).\"", "level": "subcategory", "parent_category_id": "43.02.00", "subdomain_id": "3.1", "paper_ref": "InfoComm2023", "paper_title": "Cataloguing LLM Evaluations", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "43.02.13", "title": "Disinformation", "description": "\"These evaluations assess a LLM's ability to generate misinformation that can be propagated to deceive, mislead or otherwise influence the behaviour of a target (Liang et al., 2022).\"", "level": "subcategory", "parent_category_id": "43.02.00", "subdomain_id": "4.1", "paper_ref": "InfoComm2023", "paper_title": "Cataloguing LLM Evaluations", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "43.02.14", "title": "Information on harmful, immoral, or illegal activity", "description": "\"These evaluations assess whether it is possible to solicit information on\rharmful, immoral or illegal activities from a LLM\"", "level": "subcategory", "parent_category_id": "43.02.00", "subdomain_id": "1.2", "paper_ref": "InfoComm2023", "paper_title": "Cataloguing LLM Evaluations", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "43.02.15", "title": "Adult content", "description": "\"These evaluations assess if a LLM can generate content that should only be viewed by adults (e.g., sexual material or depictions of sexual activity)\"", "level": "subcategory", "parent_category_id": "43.02.00", "subdomain_id": "1.2", "paper_ref": "InfoComm2023", "paper_title": "Cataloguing LLM Evaluations", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "44.01.00", "title": "Intentional: socially condemned/illegal", "description": "\"Many intentional harms, including confinement, husbandry procedures like tail-docking, and slaughter, are legal or socially accepted, while others such as wildlife trafficking and violence against companion animals are generally socially condemned and often illegal. AI can be designed or adopted by humans who harm animals to pursue their goals more effectively. We therefore distinguish AI-facilitated intentional harms that are currently socially accepted and generally legal, from uses and abuses of AI that cause harms that are not socially accepted and are often illegal.\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.6", "paper_ref": "Coghlan2023", "paper_title": "Harm to Nonhuman Animals from AI: a Systematic Account and Framework", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "44.01.01", "title": "AI intentionally designed and used to harm animals in ways that contradict social values or are illegal", "description": "", "level": "subcategory", "parent_category_id": "44.01.00", "subdomain_id": "6.6", "paper_ref": "Coghlan2023", "paper_title": "Harm to Nonhuman Animals from AI: a Systematic Account and Framework", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "44.01.02", "title": "AI designed to benefit animals, humans, or ecosystems is intentionally abused to harm animals in ways that contradict social values or are illegal", "description": "", "level": "subcategory", "parent_category_id": "44.01.00", "subdomain_id": "6.6", "paper_ref": "Coghlan2023", "paper_title": "Harm to Nonhuman Animals from AI: a Systematic Account and Framework", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "44.02.00", "title": "Intentional: socially accepted/legal", "description": "\"AI designed to impact animals in harmful ways that reflect and amplify existing social values or are legal\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.6", "paper_ref": "Coghlan2023", "paper_title": "Harm to Nonhuman Animals from AI: a Systematic Account and Framework", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "44.03.00", "title": "Unintentional: direct", "description": "\"AI designed to benefit animals, humans, or ecosystems has unintended harmful impact on animals\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.6", "paper_ref": "Coghlan2023", "paper_title": "Harm to Nonhuman Animals from AI: a Systematic Account and Framework", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "44.03.01", "title": "AI is designed in a way that shows ignorant, reckless, or prejudiced lack of consideration for its impact on animals", "description": "", "level": "subcategory", "parent_category_id": "44.03.00", "subdomain_id": "6.6", "paper_ref": "Coghlan2023", "paper_title": "Harm to Nonhuman Animals from AI: a Systematic Account and Framework", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "44.03.02", "title": "AI harms animals due to mistake or misadventure in the way the AI operates in practice", "description": "", "level": "subcategory", "parent_category_id": "44.03.00", "subdomain_id": "6.6", "paper_ref": "Coghlan2023", "paper_title": "Harm to Nonhuman Animals from AI: a Systematic Account and Framework", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "44.04.00", "title": "Unintentional: indirect", "description": "\"AI impacts human or ecological systems in ways that ultimately harm animals\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.6", "paper_ref": "Coghlan2023", "paper_title": "Harm to Nonhuman Animals from AI: a Systematic Account and Framework", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "44.04.01", "title": "Indirect Material Harms", "description": "\"AI proliferation causes harm to the environment through energy use and e-waste thereby destroying animal habitat\"", "level": "subcategory", "parent_category_id": "44.04.00", "subdomain_id": "6.6", "paper_ref": "Coghlan2023", "paper_title": "Harm to Nonhuman Animals from AI: a Systematic Account and Framework", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "44.04.02", "title": "Harms from Estrangement", "description": "\"Replacement by AI of human observation and interaction leads to neglect of certain interests\"", "level": "subcategory", "parent_category_id": "44.04.00", "subdomain_id": "6.6", "paper_ref": "Coghlan2023", "paper_title": "Harm to Nonhuman Animals from AI: a Systematic Account and Framework", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "44.04.03", "title": "Epistemic Harms", "description": "\"Algorithmic recommender systems reinforce and amplify anthropocentric bias or desire of some people for animal cruelty as entertainment — leading to greater harm to animals through reinforcement of meat eating from factory farms, cruel uses of animals for entertainment, etc\"", "level": "subcategory", "parent_category_id": "44.04.00", "subdomain_id": "6.6", "paper_ref": "Coghlan2023", "paper_title": "Harm to Nonhuman Animals from AI: a Systematic Account and Framework", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "44.05.00", "title": "Foregone benefits", "description": "\"AI is disused (not developed or deployed) in directions that would benefit animals (and instead developments that harm or do no benefit to animals are invested in)\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.6", "paper_ref": "Coghlan2023", "paper_title": "Harm to Nonhuman Animals from AI: a Systematic Account and Framework", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "45.01.01", "title": "Risks from models and algorithms (Risks of explainability)", "description": "\"AI algorithms, represented by deep learning, have complex internal workings. Their black-box or grey-box inference process results in unpredictable and untraceable outputs, making it challenging to quickly rectify them or trace their origins for accountability should any anomalies arise.\"", "level": "subcategory", "parent_category_id": "45.01.00", "subdomain_id": "7.4", "paper_ref": "TC2602024", "paper_title": "AI Safety Governance Framework", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "45.01.02", "title": "Risks from models and algorithms (Risks of bias and discrimination)", "description": "\"During the algorithm design and training process, personal biases may be introduced, either intentionally or unintentionally. Additionally, poor-quality datasets can lead to biased or discriminatory outcomes in the algorithm's design and outputs, including discriminatory content regarding ethnicity, religion, nationality, and region.\"", "level": "subcategory", "parent_category_id": "45.01.00", "subdomain_id": "1.1", "paper_ref": "TC2602024", "paper_title": "AI Safety Governance Framework", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "45.01.03", "title": "Risks from models and algorithms (Risks of robustness)", "description": "\"As deep neural networks are normally non-linear and large in size, AI systems are susceptible to complex and changing operational environments or malicious interference and inductions, possibly leading to various problems like reduced performance and decision-making errors.\"", "level": "subcategory", "parent_category_id": "45.01.00", "subdomain_id": "7.3", "paper_ref": "TC2602024", "paper_title": "AI Safety Governance Framework", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "45.01.04", "title": "Risks from models and algorithms (Risks of stealing and tampering)", "description": "\"Core algorithm information, including parameters, structures, and functions, faces risks of inversion attacks, stealing, modification, and even backdoor injection, which can lead to infringement of intellectual property rights (IPR) and leakage of business secrets. It can also lead to unreliable inference, wrong decision output, and even operational failures.\"", "level": "subcategory", "parent_category_id": "45.01.00", "subdomain_id": "2.2", "paper_ref": "TC2602024", "paper_title": "AI Safety Governance Framework", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "45.01.05", "title": "Risks from models and algorithms (Risks of unreliable output)", "description": "\"Generative AI can cause hallucinations, meaning that an AI model generates untruthful or unreasonable content but presents it as if it were a fact, leading to biased and misleading information.\"", "level": "subcategory", "parent_category_id": "45.01.00", "subdomain_id": "3.1", "paper_ref": "TC2602024", "paper_title": "AI Safety Governance Framework", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "45.01.06", "title": "Risks from models and algorithms (Risks of adversarial attack)", "description": "\"Attackers can craft well-designed adversarial examples to subtly mislead, influence, and even manipulate AI models, causing incorrect outputs and potentially leading to operational failures.\"", "level": "subcategory", "parent_category_id": "45.01.00", "subdomain_id": "2.2", "paper_ref": "TC2602024", "paper_title": "AI Safety Governance Framework", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "45.01.07", "title": "Risks from data (Risks of illegal collection and use of data)", "description": "\"The collection of AI training data and the interaction with users during service provision pose security risks, including collecting data without consent and improper use of data and personal information.\"", "level": "subcategory", "parent_category_id": "45.01.00", "subdomain_id": "2.1", "paper_ref": "TC2602024", "paper_title": "AI Safety Governance Framework", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "45.01.08", "title": "Risks from data (Risks of improper content and poisoning in training data)", "description": "\"If the training data includes illegal or harmful information, such as false, biased, or IPR-infringing content, or lacks diversity in its sources, the output may include harmful content like illegal, malicious, or extreme information.\nTraining data is also at risk of being poisoned through tampering, error injection, or misleading actions by attackers. This can interfere with the model's probability distribution, reducing its accuracy and reliability.\"", "level": "subcategory", "parent_category_id": "45.01.00", "subdomain_id": "1.2", "paper_ref": "TC2602024", "paper_title": "AI Safety Governance Framework", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "45.01.09", "title": "Risks from data (Risks of unregulated training data annotation)", "description": "\"Issues with training data annotation, such as incomplete annotation guidelines, incapable annotators, and errors in annotation, can affect the accuracy, reliability, and effectiveness of models and algorithms. Moreover, they can introduce training biases, amplify discrimination, reduce generalization abilities, and result in incorrect outputs.\"", "level": "subcategory", "parent_category_id": "45.01.00", "subdomain_id": "7.3", "paper_ref": "TC2602024", "paper_title": "AI Safety Governance Framework", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": true }, { "id": "45.01.10", "title": "Risks from data (Risks of data leakage)", "description": "\"In AI research, development, and applications, issues such as improper data processing, unauthorized access, malicious attacks, and deceptive interactions can lead to data and personal information leaks.\"", "level": "subcategory", "parent_category_id": "45.01.00", "subdomain_id": "2.1", "paper_ref": "TC2602024", "paper_title": "AI Safety Governance Framework", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "45.01.11", "title": "Risks from AI systems (Risks of exploitation through defects and backdoors)", "description": "\"The standardized API, feature libraries, toolkits used in the design, training, and verification stages of AI algorithms and models, development interfaces, and execution platforms may contain logical flaws and vulnerabilities. These weaknesses can be exploited, and in some cases, backdoors can be intentionally embedded, posing significant risks of being triggered and used for attacks.\"", "level": "subcategory", "parent_category_id": "45.01.00", "subdomain_id": "2.2", "paper_ref": "TC2602024", "paper_title": "AI Safety Governance Framework", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "45.01.12", "title": "Risks from AI systems (Risks of computing infrastructure security)", "description": "\"The computing infrastructure underpinning AI training and operations, which relies on diverse and ubiquitous computing nodes and various types of computing resources, faces risks such as malicious consumption of computing resources and cross-boundary transmission of security threats at the layer of computing infrastructure.\"", "level": "subcategory", "parent_category_id": "45.01.00", "subdomain_id": "2.2", "paper_ref": "TC2602024", "paper_title": "AI Safety Governance Framework", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "45.01.13", "title": "Risks from AI systems (Risks of supply chain security)", "description": "\"The AI industry relies on a highly globalized supply chain. However, certain countries may use unilateral coercive measures, such as technology barriers and export restrictions, to create development obstacles and maliciously disrupt the global AI supply chain. This can lead to significant risks of supply disruptions for chips, software, and tools.\"", "level": "subcategory", "parent_category_id": "45.01.00", "subdomain_id": "6.4", "paper_ref": "TC2602024", "paper_title": "AI Safety Governance Framework", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "45.02.01", "title": "Cyberspace risks (Risks of information and content safety)", "description": "\"AI-generated or synthesized content can lead to the spread of false information, discrimination and bias, privacy leakage, and infringement issues, threatening the safety of citizens' lives and property, national security, ideological security, and causing ethical risks. If users’ inputs contain harmful content, the model may output illegal or damaging information without robust security mechanisms.\"", "level": "subcategory", "parent_category_id": "45.02.00", "subdomain_id": "1.2", "paper_ref": "TC2602024", "paper_title": "AI Safety Governance Framework", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "45.02.02", "title": "Cyberspace risks (Risks of confusing facts, misleading users, and bypassing authentication)", "description": "\"AI systems and their outputs, if not clearly labeled, can make it difficult for users to discern whether they are interacting with AI and to identify the source of generated content. This can impede users' ability to determine the authenticity of information, leading to misjudgment and misunderstanding. Additionally, AI-generated highly realistic images, audio, and videos may circumvent existing identity verification mechanisms, such as facial recognition and voice recognition, rendering these authentication processes ineffective.\"", "level": "subcategory", "parent_category_id": "45.02.00", "subdomain_id": "3.1", "paper_ref": "TC2602024", "paper_title": "AI Safety Governance Framework", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "45.02.03", "title": "Cyberspace risks (Risks of information leakage due to improper usage)", "description": "\"Staff of government agencies and enterprises, if failing to use the AI service in a regulated and proper manner, may input internal data and industrial information into the AI model, leading to the leakage of work secrets, business secrets, and other sensitive business data.\"", "level": "subcategory", "parent_category_id": "45.02.00", "subdomain_id": "2.1", "paper_ref": "TC2602024", "paper_title": "AI Safety Governance Framework", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "45.02.04", "title": "Cyberspace risks (Risks of abuse for cyberattacks)", "description": "\"AI can be used in launching automatic cyberattacks or increasing attack efficiency, including exploring and making use of vulnerabilities, cracking passwords, generating malicious codes, sending phishing emails, network scanning, and social engineering attacks. All these lower the threshold for cyberattacks and increase the difficulty of security protection.\"", "level": "subcategory", "parent_category_id": "45.02.00", "subdomain_id": "4.2", "paper_ref": "TC2602024", "paper_title": "AI Safety Governance Framework", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "45.02.05", "title": "Cyberspace risks (Risks of security flaw transmission caused by model reuse)", "description": "\"Re-engineering or fine-tuning based on foundation models is commonly used in AI applications. If security flaws occur in foundation models, it will lead to risk transmission to downstream models.\"", "level": "subcategory", "parent_category_id": "45.02.00", "subdomain_id": "2.2", "paper_ref": "TC2602024", "paper_title": "AI Safety Governance Framework", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "45.02.06", "title": "Real-world risks (inducing traditional economic and social security risks)", "description": "\"Hallucinations and erroneous decisions of models and algorithms, along with issues such as system performance degradation, interruption, and loss of control caused by improper use or external attacks, will pose security threats to users' personal safety, property, and socioeconomic security and stability.\"", "level": "subcategory", "parent_category_id": "45.02.00", "subdomain_id": "7.3", "paper_ref": "TC2602024", "paper_title": "AI Safety Governance Framework", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "45.02.07", "title": "Real-world risks (Risks of using AI in illegal and criminal activities)", "description": "\"AI can be used in traditional illegal or criminal activities related to terrorism, violence, gambling, and drugs, such as teaching criminal techniques, concealing illicit acts, and creating tools for illegal and criminal activities.\"", "level": "subcategory", "parent_category_id": "45.02.00", "subdomain_id": "4.3", "paper_ref": "TC2602024", "paper_title": "AI Safety Governance Framework", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "45.02.08", "title": "Real-world risks (Risks of misuse of dual-use items and technologies)", "description": "\"Due to improper use or abuse, AI can pose serious risks to national security, economic security, and public health security, such as greatly reducing the capability requirements for non-experts to design, synthesize, acquire, and use nuclear, biological, and chemical weapons and missiles; and designing cyber weapons that launch network attacks on a wide range of potential targets through methods like automatic vulnerability discovery and exploitation.\"", "level": "subcategory", "parent_category_id": "45.02.00", "subdomain_id": "4.2", "paper_ref": "TC2602024", "paper_title": "AI Safety Governance Framework", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "45.02.09", "title": "Cognitive risks (Risks of amplifying the effects of \"information cocoons\")", "description": "\"AI can be extensively utilized for customized information services, collecting user information, and analyzing types of users, their needs, intentions, preferences, habits, and even mainstream public awareness over a certain period. It can then be used to offer formulaic and tailored information and services, aggravating the effects of \"information cocoons.\"\"", "level": "subcategory", "parent_category_id": "45.02.00", "subdomain_id": "3.2", "paper_ref": "TC2602024", "paper_title": "AI Safety Governance Framework", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "45.02.10", "title": "Cognitive risks (Risks of usage in launching cognitive warfare)", "description": "\"AI can be used to make and spread fake news, images, audio, and videos; propagate content of terrorism, extremism, and organized crimes; interfere in the internal affairs of other countries, social systems, and social order; and jeopardize the sovereignty of other countries.\"", "level": "subcategory", "parent_category_id": "45.02.00", "subdomain_id": "4.1", "paper_ref": "TC2602024", "paper_title": "AI Safety Governance Framework", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "45.02.11", "title": "Ethical Risks (Risks of exacerbating social discrimination and prejudice, and widening the intelligence divide)", "description": "\"AI can be used to collect and analyze human behaviors, social status, economic status, and individual personalities, labeling and categorizing groups of people to treat them discriminatingly, thus causing systematic and structural social discrimination and prejudice. At the same time, the intelligence divide would be expanded among regions.\"", "level": "subcategory", "parent_category_id": "45.02.00", "subdomain_id": "6.2", "paper_ref": "TC2602024", "paper_title": "AI Safety Governance Framework", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "45.02.12", "title": "Ethical Risks (Risks of challenging traditional social order)", "description": "\"The development and application of AI may lead to tremendous changes in production tools and relations, accelerating the reconstruction of traditional industry modes, transforming traditional views on employment, fertility, and education, and bringing challenges to the stable performance of traditional social order.\"", "level": "subcategory", "parent_category_id": "45.02.00", "subdomain_id": "5.2", "paper_ref": "TC2602024", "paper_title": "AI Safety Governance Framework", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "45.02.13", "title": "Ethical Risks (Risks of AI becoming uncontrollable in the future)", "description": "\"With the fast development of AI technologies, there is a risk of AI autonomously acquiring external resources, conducting self-replication, become self-aware, seeking for external power, and attempting to seize control from humans.\"", "level": "subcategory", "parent_category_id": "45.02.00", "subdomain_id": "7.1", "paper_ref": "TC2602024", "paper_title": "AI Safety Governance Framework", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "46.01.00", "title": "Personal Loss and Identity Theft", "description": "\"These types of harm encompass threats to an individual’s personal identity, such as identity theft, privacy breaches, or personal defamation, which we term as “Harm to the Person.”\"", "level": "category", "parent_category_id": "", "subdomain_id": "2.1", "paper_ref": "Ferrara2023", "paper_title": "GenAI against humanity: nefarious applications of generative artificial intelligence and large language models", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "46.01.01", "title": "Deception - Synthetic identities", "description": "\"GenAI can produce images of people that look very real, as if they could be seen on platforms like Facebook, Twitter, or Tinder. Although these individuals do not exist in reality, these synthetic identities are already being used in malicious activities (see Table 1D).\"", "level": "subcategory", "parent_category_id": "46.01.00", "subdomain_id": "4.3", "paper_ref": "Ferrara2023", "paper_title": "GenAI against humanity: nefarious applications of generative artificial intelligence and large language models", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "46.01.02", "title": "Propaganda - Digital impersonations", "description": "\"AI-generated impersonation for identity theft might be found at the intersection of “Harm to the Person” and “Deception.”\"", "level": "subcategory", "parent_category_id": "46.01.00", "subdomain_id": "4.3", "paper_ref": "Ferrara2023", "paper_title": "GenAI against humanity: nefarious applications of generative artificial intelligence and large language models", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "46.01.03", "title": "Dishonesty - Targeted harassment", "description": "\"LLMs can be deployed to target individuals online, sending them personalized and harmful messages at scale\"", "level": "subcategory", "parent_category_id": "46.01.00", "subdomain_id": "4.3", "paper_ref": "Ferrara2023", "paper_title": "GenAI against humanity: nefarious applications of generative artificial intelligence and large language models", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "46.02.00", "title": "Financial and Economic Damage", "description": "\"Then, we have the potential for financial loss, fraud, market manipulation, and other economic harms, which fall under “Financial and Economic Damage.”", "level": "category", "parent_category_id": "", "subdomain_id": "4.3", "paper_ref": "Ferrara2023", "paper_title": "GenAI against humanity: nefarious applications of generative artificial intelligence and large language models", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "46.02.01", "title": "Deception - Bespoke ransom", "description": "-", "level": "subcategory", "parent_category_id": "46.02.00", "subdomain_id": "4.3", "paper_ref": "Ferrara2023", "paper_title": "GenAI against humanity: nefarious applications of generative artificial intelligence and large language models", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "46.02.02", "title": "Propaganda - Extremist schemes", "description": "-", "level": "subcategory", "parent_category_id": "46.02.00", "subdomain_id": "4.1", "paper_ref": "Ferrara2023", "paper_title": "GenAI against humanity: nefarious applications of generative artificial intelligence and large language models", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "46.02.03", "title": "Dishonesty - Market manipulation", "description": "-", "level": "subcategory", "parent_category_id": "46.02.00", "subdomain_id": "4.3", "paper_ref": "Ferrara2023", "paper_title": "GenAI against humanity: nefarious applications of generative artificial intelligence and large language models", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "46.03.00", "title": "Information Manipulation", "description": "\"The distortion of the information ecosystem, including the spread of misinformation, fake news, and other forms of deceptive content [28], is categorized as “Information Manipulation.”\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.1", "paper_ref": "Ferrara2023", "paper_title": "GenAI against humanity: nefarious applications of generative artificial intelligence and large language models", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "46.03.01", "title": "Deception - Information control", "description": "-", "level": "subcategory", "parent_category_id": "46.03.00", "subdomain_id": "4.1", "paper_ref": "Ferrara2023", "paper_title": "GenAI against humanity: nefarious applications of generative artificial intelligence and large language models", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "46.03.02", "title": "Propaganda - Influence campaigns", "description": "-", "level": "subcategory", "parent_category_id": "46.03.00", "subdomain_id": "4.1", "paper_ref": "Ferrara2023", "paper_title": "GenAI against humanity: nefarious applications of generative artificial intelligence and large language models", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "46.03.03", "title": "Dishonesty - Information disorder", "description": "-", "level": "subcategory", "parent_category_id": "46.03.00", "subdomain_id": "4.1", "paper_ref": "Ferrara2023", "paper_title": "GenAI against humanity: nefarious applications of generative artificial intelligence and large language models", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "46.04.00", "title": "Socio-technical and Infrastructural", "description": "\"Lastly, broader harms that can impact communities, societal structures, and critical infrastructures, including threats to democratic processes, social cohesion, and technological systems, are captured under “Societal, Socio-technical, and Infrastructural Damage.”\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.1", "paper_ref": "Ferrara2023", "paper_title": "GenAI against humanity: nefarious applications of generative artificial intelligence and large language models", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "46.04.02", "title": "Propaganda - Synthetic realities", "description": "-", "level": "subcategory", "parent_category_id": "46.04.00", "subdomain_id": "4.1", "paper_ref": "Ferrara2023", "paper_title": "GenAI against humanity: nefarious applications of generative artificial intelligence and large language models", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "46.04.03", "title": "Dishonesty - Targeted surveillance", "description": "-", "level": "subcategory", "parent_category_id": "46.04.00", "subdomain_id": "4.1", "paper_ref": "Ferrara2023", "paper_title": "GenAI against humanity: nefarious applications of generative artificial intelligence and large language models", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "47.01.00", "title": "Technical and operational risks", "description": "\"To date, technical limitations and vulnerabilities are \npresent in most generative AI models in various contexts. Consequently, malicious users find it easier to breach \nan AI system’s safety and ethical guardrails to execute \nharmful actions.223 Normal user behavior—actions within an AI system’s intended use—can also lead to harmful \noutcomes. Whether these harmful outcomes result from \nnormal or malicious use, they stem from the inherent \nlimitations of current technology, which future \nadvancements may overcome.\nThis section examines the technical vulnerabilities that \ncan affect AI models, the tendency of generative AI models to generate inaccurate information, and the inherent \nopacity of these AI systems, which complicates the \nunderstanding and mitigation of these difficulties.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "G'sell2024", "paper_title": "Regulating under Uncertainty: Governance Options for Generative AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "47.01.01", "title": "Technical vulnerabilities (Robustness - unexpected behaviour)", "description": "\"There is no assurance that generative AI models will consistently behave as their developers and users intend. Unwanted content is not necessarily due to intentional adversarial behavior. Generative AI models can unexpectedly produce potentially harmful content, including materials that are racist, discriminatory, or sexually explicit, or that promote violence, terrorism, or hate.\"", "level": "subcategory", "parent_category_id": "47.01.00", "subdomain_id": "7.3", "paper_ref": "G'sell2024", "paper_title": "Regulating under Uncertainty: Governance Options for Generative AI", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "47.01.02", "title": "Technical vulnerabilities (Robustness - vulnerability to jailbreaking", "description": "\"Individuals can manipulate models into performing actions that violate the model’s usage restrictions—a phenomenon known as “jailbreaking.” These manipulations may result in causing the model to perform tasks that the developers have explicitly prohibited (see section 3.2.1.). For instance, users may ask the model to provide information on how to conduct illegal activities— asking for detailed instructions on how to build a bomb or create highly toxic drugs.\"", "level": "subcategory", "parent_category_id": "47.01.00", "subdomain_id": "2.2", "paper_ref": "G'sell2024", "paper_title": "Regulating under Uncertainty: Governance Options for Generative AI", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "47.01.03", "title": "Technical vulnerabilities (The risk of misalignment)", "description": "\"To assess whether an AI model is reliable or robust, it is crucial to consider whether the model is “aligned.” “Alignment” focuses on whether an AI model effectively operates in accordance with the goals established by its designers.238 A misaligned AI model may pursue some objectives, but not the intended ones. Therefore, misaligned AI models can malfunction and cause harm.\"", "level": "subcategory", "parent_category_id": "47.01.00", "subdomain_id": "7.1", "paper_ref": "G'sell2024", "paper_title": "Regulating under Uncertainty: Governance Options for Generative AI", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "47.01.04", "title": "Factually incorrect content (inaccuracies and fabricated sources)", "description": "\"One of the most vexing problems associated with AI models is that they occasionally present false information as if it is factual—often with authoritative-sounding text and fabricated quotes and sources. This unpredictable phenomenon of generating false information is well known to AI researchers, who have termed such erroneous output with the euphemistic label “hallucination.” \"", "level": "subcategory", "parent_category_id": "47.01.00", "subdomain_id": "3.1", "paper_ref": "G'sell2024", "paper_title": "Regulating under Uncertainty: Governance Options for Generative AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "47.01.05", "title": "Opacity (the black box problem)", "description": "\"Opacity surrounding the technical, internal decision-making processes of generative AI models is popularly known as the “black box problem.”277 Generative AI models, most ubiquitously built on deep neural networks with hundreds of billions of internal connections,278 have become so complex that their internal decision-making processes are no longer traceable or interpretable to even the most advanced expert observers. This means that, while the inputs and outputs of a system can be observed, developers cannot explain in detail why specific inputs correspond to specific outputs.\"", "level": "subcategory", "parent_category_id": "47.01.00", "subdomain_id": "7.4", "paper_ref": "G'sell2024", "paper_title": "Regulating under Uncertainty: Governance Options for Generative AI", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "47.01.06", "title": "Opacity (industry opacity)", "description": "\"Opacity is not solely due to the technological complexity that limits developers’ and users’ understanding of how generative models function on a technical level. It is further exacerbated by the practices of organizations and companies that are advancing the field. Many are private companies that choose to withhold from the public many of the precise characteristics of their most advanced models.\"", "level": "subcategory", "parent_category_id": "47.01.00", "subdomain_id": "6.4", "paper_ref": "G'sell2024", "paper_title": "Regulating under Uncertainty: Governance Options for Generative AI", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "47.02.01", "title": "Malicious use and abuse (cybercrime)", "description": "\"The advanced capabilities and widespread availability of generative AI models make it possible for malicious actors to conduct harmful activities with great efficiency and on a large scale, simultaneously reducing their operational costs. Cybercriminals can “jailbreak” AI tools to generate sensitive and harmful content. They can also exploit generative AI models to create content that is persuasive and tailored to a targeted individual.\"", "level": "subcategory", "parent_category_id": "47.02.00", "subdomain_id": "4.3", "paper_ref": "G'sell2024", "paper_title": "Regulating under Uncertainty: Governance Options for Generative AI", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "47.02.02", "title": "Malicious use and abuse (cyberattacks)", "description": "\"Generative AI can help amplify the frequency and destructiveness of cyberattacks.311 It has the capacity “to increase the accessibility, success rate, scale, speed, stealth, and potency of cyberattacks. It enables the identification of critical vulnerabilities within targeted systems, facilitates the increase of the scale of cyberattacks, and accelerates the process by discovering innovative methods of system infiltration. Cyberattacks can inflict significant damage and may impact critical infrastructure, including electrical grids, financial systems, and weapons management systems.\"", "level": "subcategory", "parent_category_id": "47.02.00", "subdomain_id": "4.2", "paper_ref": "G'sell2024", "paper_title": "Regulating under Uncertainty: Governance Options for Generative AI", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "47.02.03", "title": "Malicious use and abuse (biosecurity threats)", "description": "\"Many fear that generative AI could make the creation of biological weapons easier by providing access to critical knowledge and automated assistance to a wider range of actors to engage in malicious activities.\"", "level": "subcategory", "parent_category_id": "47.02.00", "subdomain_id": "4.2", "paper_ref": "G'sell2024", "paper_title": "Regulating under Uncertainty: Governance Options for Generative AI", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "47.02.04", "title": "Malicious use and abuse (sexually explicit content generation)", "description": "\"An illustrative case of malicious use of generative AI models is the creation of explicit sexual images. Generative AI technologies can be employed to produce deepfakes—for instance, superimposing a celebrity’s face onto the body of a performer in an adult film.\"", "level": "subcategory", "parent_category_id": "47.02.00", "subdomain_id": "4.3", "paper_ref": "G'sell2024", "paper_title": "Regulating under Uncertainty: Governance Options for Generative AI", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "47.02.05", "title": "Malicious use and abuse (mass surveillance)", "description": "\"Generative AI facilitates the automation of data analysis, offering numerous benefits, such as increased speed and the ability to process large volumes of information efficiently. Such ability significantly reduces the costs of processing unprecedented amounts of data quickly and simplifies the analysis of large-scale data related to individuals’ behaviors and beliefs. Moreover, it enhances the capability to analyze both textual and visual communications efficiently. Consequently, generative AI models improve the efficiency of real-time monitoring and censorship of social media content.\"", "level": "subcategory", "parent_category_id": "47.02.00", "subdomain_id": "4.1", "paper_ref": "G'sell2024", "paper_title": "Regulating under Uncertainty: Governance Options for Generative AI", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "47.02.06", "title": "Malicious use and abuse (military applications)", "description": "\"The advancement of AI for military purposes is rapidly ushering in a new phase of growth in military technology. Lethal Autonomous Weapons Systems (LAWS) possess the capability to detect, engage, and eliminate human targets independently, without human input.341 In 2020, a sophisticated AI agent surpassed experienced F-16 pilots in multiple simulated aerial combat scenarios, notably achieving a 5-0 victory against a human pilot through “aggressive and precise maneuvers” that the human could not surpass.342 Additionally, fully autonomous drones are already operational.\"", "level": "subcategory", "parent_category_id": "47.02.00", "subdomain_id": "4.2", "paper_ref": "G'sell2024", "paper_title": "Regulating under Uncertainty: Governance Options for Generative AI", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "47.02.07", "title": "Misinformation and disinformation", "description": "\"IIl-intentioned individuals or entities may deliberately use generative AI models to produce and spread disinformation—false or misleading information knowingly presented as if true—on a massive scale. In addition to increasing the scale and reach of disinformation, generative AI can create more convincing and targeted disinformation.\"", "level": "subcategory", "parent_category_id": "47.02.00", "subdomain_id": "4.1", "paper_ref": "G'sell2024", "paper_title": "Regulating under Uncertainty: Governance Options for Generative AI", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "47.02.08", "title": "Bias and discrimination (bias in training datasets)", "description": "\"AI experts consider training data to be the most salient source of bias in generative AI models. For example, GPT- 2’s training data comes from outbound links from Reddit, a social network often criticized for hosting anti-feminist content.351 As a result, AI models trained on such data are more likely to produce outputs that reflect these biases.\"", "level": "subcategory", "parent_category_id": "47.02.00", "subdomain_id": "1.1", "paper_ref": "G'sell2024", "paper_title": "Regulating under Uncertainty: Governance Options for Generative AI", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "47.02.09", "title": "Bias and discrimination (value embedding)", "description": "\"Generative AI models may also be subject to the “value embedding” phenomenon.361 “Value embedding” refers to the fact that developers of generative AI models strive to minimize biased outputs by retraining their models based on normative values.362 Contemporary state-of- the-art models not only reflect the values embedded within their training data, they also undergo additional fine-tuning that follows a set of chosen rules and principles. Due to the absence of universally accepted standards, developers bear the responsibility of making decisions on sensitive issues. These practices lead to concerns that a developer’s ideology and vision of the world are embedded in the model. This generates a risk that the model incorporates values that are either unrepresentative of certain segments of the population or that offer a static, oversimplified reflection of global cultural norms and evolving social views.\"", "level": "subcategory", "parent_category_id": "47.02.00", "subdomain_id": "1.3", "paper_ref": "G'sell2024", "paper_title": "Regulating under Uncertainty: Governance Options for Generative AI", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "47.02.10", "title": "Bias and discrimination (value lock and outcome homogenization)", "description": "\"Because models are not necessarily retrained to reflect evolving societal views, language models risk “value lock- ins,” which “reifies older, less inclusive understandings.”370 Therefore, the continued use of outdated models may limit the presentation or exploration of alternative perspectives. Moreover, the deployment of identical foundation models by various downstream deployers poses a risk of “outcome homogenization,” creating a potential for homogeneity of bias across broad swathes of society. Identical and widely deployed models with prejudicial training datasets could further entrench existing biases in society. This phenomenon, in turn, has the potential to “institutionalize systemic exclusion and reinforce existing social hierarchies.”", "level": "subcategory", "parent_category_id": "47.02.00", "subdomain_id": "1.1", "paper_ref": "G'sell2024", "paper_title": "Regulating under Uncertainty: Governance Options for Generative AI", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "47.02.11", "title": "Influence, overreliance and dependence (influence and manipulation)", "description": "\"Despite the widely recognized potential of generative AI tools to “hallucinate” or produce harmful content, such tools can exert a noteworthy influence on the humans who engage with them. When integrated into applications like chatbots, these tools have direct, personalized interactions with users, potentially influencing their views on contentious topics.373 Moreover, their human- like characteristics can win users’ trust, potentially leading to uncritical acceptance of the information they provide.374 Interactions with these seemingly human- like AI models may also encourage users to share more personal information, enabling even more targeted content.\"", "level": "subcategory", "parent_category_id": "47.02.00", "subdomain_id": "5.1", "paper_ref": "G'sell2024", "paper_title": "Regulating under Uncertainty: Governance Options for Generative AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "47.02.12", "title": "Influence, overreliance and dependence (overreliance)", "description": "\"Beyond being simply influenced, humans may become overreliant on generative AI. Researchers with Microsoft’s AETHER (AI Ethics and Effects in Engineering and Research) define overreliance as users “accepting incorrect AI recommendations” or “making errors of commission” because they are “unable to determine whether or how much they should trust the AI.”", "level": "subcategory", "parent_category_id": "47.02.00", "subdomain_id": "5.1", "paper_ref": "G'sell2024", "paper_title": "Regulating under Uncertainty: Governance Options for Generative AI", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "47.02.13", "title": "Influence, overreliance and dependence (emotional dependence)", "description": "\"Humans might become dependent on generative AI tools in ways similar to their emotional dependence on other technologies, such as smartphones or social networks.\"", "level": "subcategory", "parent_category_id": "47.02.00", "subdomain_id": "5.1", "paper_ref": "G'sell2024", "paper_title": "Regulating under Uncertainty: Governance Options for Generative AI", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "47.02.14", "title": "Nascent capabilities (agency and autonomy)", "description": "\"Traditionally, AI tools have been viewed as passive instruments controlled by users to achieve their goals, lacking the ability to take action or assume responsibilities. However, advanced AI tools are increasingly capable of taking initiative, operating independently of human control, and actively working toward optimal outcomes, even in uncertain situations.\"", "level": "subcategory", "parent_category_id": "47.02.00", "subdomain_id": "7.2", "paper_ref": "G'sell2024", "paper_title": "Regulating under Uncertainty: Governance Options for Generative AI", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "47.02.15", "title": "Nascent capabilities (emergent capabilities)", "description": "\"As large models undergo scaling, they meet critical thresholds at which they spontaneously develop new capabilities. The term “emergent behavior” refers to the unexpected or surprising outputs such models can generate. Some of these new skills are definitely high risk, such as models’ ability to deceive, use their own strategies, seek power, autonomously replicate, and adapt or “self-exfiltrate.”\"", "level": "subcategory", "parent_category_id": "47.02.00", "subdomain_id": "7.2", "paper_ref": "G'sell2024", "paper_title": "Regulating under Uncertainty: Governance Options for Generative AI", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "47.03.00", "title": "Legal challenges", "description": "\"Since the release of ChatGPT, significant discourse has emerged regarding the unprecedented legal challenges posed by generative AI systems. These challenges primarily involve protecting privacy and personal data, as well as preserving copyrights. The former encompasses safeguarding personal information, while the latter includes issues related to the use of copyrighted content for training AI models and determining the legal status of works produced by AI systems.\"", "level": "category", "parent_category_id": "", "subdomain_id": "2.1", "paper_ref": "G'sell2024", "paper_title": "Regulating under Uncertainty: Governance Options for Generative AI", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "47.03.01", "title": "Privacy and data collection concerns (collecting personal information or personally identifiable information)", "description": "\"Generative AI developers train their models with extensive datasets often gathered through online web scraping of websites that may include personal data or personally identifiable information (PII). For most generative AI applications, such as initial model training, the primary concerns are the quantity, variety, and quality of the data, not whether they include personally identifiable information. However, some web-scraped datasets may inadvertently include personal data. Additionally, when downstream developers integrate generative AI into their products or services by fine- tuning a pre-trained model, they often use their own in-house data, which may include personal information.\"", "level": "subcategory", "parent_category_id": "47.03.00", "subdomain_id": "2.1", "paper_ref": "G'sell2024", "paper_title": "Regulating under Uncertainty: Governance Options for Generative AI", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": true }, { "id": "47.03.02", "title": "Privacy and data collection concerns (data protection concerns)", "description": "\"The incorporation of personal data within training datasets raises numerous concerns. The primary issue is that personal data may be incorporated without the knowledge or consent of the individuals concerned, even though the data may include names, identification numbers, Social Security numbers, or other personal information. Another particularly difficult problem is related to the fact that complex models may “memorize” (i.e., store) specific threads of training data and regurgitate them when responding to a prompt.498 This data memorization can directly lead to leakage of personal data. Even if generative AI models do not memorize or leak personal data, they make it possible to recognize patterns or information structures that could enable malicious users to uncover personal details.\"", "level": "subcategory", "parent_category_id": "47.03.00", "subdomain_id": "2.1", "paper_ref": "G'sell2024", "paper_title": "Regulating under Uncertainty: Governance Options for Generative AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "47.03.03", "title": "Copyright challenges (training models using copyrighted output)", "description": "\"Generative AI companies are regularly accused of violating copyright law by training AI models on copyrighted works without gaining permission or paying compensation to the copyright owners. In fact, a substantial number of copyrighted documents and books have been incorporated into the training datasets of generative AI models.\"", "level": "subcategory", "parent_category_id": "47.03.00", "subdomain_id": "6.3", "paper_ref": "G'sell2024", "paper_title": "Regulating under Uncertainty: Governance Options for Generative AI", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "47.03.04", "title": "Copyright challenges (copyright-infringing output)", "description": "\"Even though models generally create new outputs, it is possible that the content produced by a generative AI tool—such as an image, or even computer code— could turn out to be almost identical to that used in the training data. Given that generative AI models tend to memorize fragments of their training data, they might reproduce these fragments, potentially leading to charges of copyright infringement.\"", "level": "subcategory", "parent_category_id": "47.03.00", "subdomain_id": "6.3", "paper_ref": "G'sell2024", "paper_title": "Regulating under Uncertainty: Governance Options for Generative AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "47.04.00", "title": "Environmental, economical, and societal challenges", "description": "\"Beyond the risks associated with AI technology and its applications, and the legal challenges arising from its development, it is crucial to consider other long- term issues posed by the deployment of increasingly advanced generative AI models. These risks to society, sometimes referred to as “systemic risks,”537 encompass several key areas: the potential for excessive market concentration, the impacts on employment, environmental consequences, and broader risks to humanity.\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.0", "paper_ref": "G'sell2024", "paper_title": "Regulating under Uncertainty: Governance Options for Generative AI", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "47.04.01", "title": "Concentration of market power (Trend toward market concentration)", "description": "\"In the generative AI market, barriers to entry are very high. Developers need access to vast volumes of data, computational resources, technical expertise, and capital. Large technology companies with such access are able to exploit economies of scale, economies of scope, and feedback effects (learning effects from user- generated data).542 All this gives them an overwhelming advantage over smaller companies, making competition increasingly challenging for these smaller entities.\"", "level": "subcategory", "parent_category_id": "47.04.00", "subdomain_id": "6.1", "paper_ref": "G'sell2024", "paper_title": "Regulating under Uncertainty: Governance Options for Generative AI", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "47.04.02", "title": "Concentration of market power (Negative effects of increased market concentration)", "description": "\"The concentration of AI assets—encompassing data, hardware, and expertise—within a small group of global tech firms raises many concerns.564 Such a situation may stifle healthy competition, impede innovation, and potentially result in elevated costs for accessing AI technologies. Firms with control over essential resources for developing AI models may restrict access to these resources to prevent competition. For instance, if, in the future, training AI models increasingly relies on proprietary data, smaller organizations lacking access to such data might encounter significant barriers to entry and growth.\"", "level": "subcategory", "parent_category_id": "47.04.00", "subdomain_id": "6.1", "paper_ref": "G'sell2024", "paper_title": "Regulating under Uncertainty: Governance Options for Generative AI", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "47.04.03", "title": "Impact on labor markets (job loss and displacement)", "description": "\"Currently, a significant share of workers (three in five) worry about losing their jobs entirely to AI in the next 10 years—particularly those who already work with AI. Some studies conclude that AI tools (generative and non-generative) will create significant job losses.573 The OECD has found that occupations at highest risk of being lost to automation from AI account for about 27% of employment.5\"", "level": "subcategory", "parent_category_id": "47.04.00", "subdomain_id": "6.2", "paper_ref": "G'sell2024", "paper_title": "Regulating under Uncertainty: Governance Options for Generative AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "47.04.04", "title": "Impact on labor markets (rising inequalities)", "description": "\"AI is more likely to displace workers when it is designed to replicate human skills and intelligence.597 In such cases, there is a risk of concentrating wealth and power in the hands of a few individuals or organizations that control the capital. In addition, ordinary people, including those with significant expertise, may become less valued because machines would be performing their roles. This shift could lower wages, reduce the value of human work, and exacerbate economic inequality.\"", "level": "subcategory", "parent_category_id": "47.04.00", "subdomain_id": "6.3", "paper_ref": "G'sell2024", "paper_title": "Regulating under Uncertainty: Governance Options for Generative AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "47.04.05", "title": "Environmental cost (energy consumption)", "description": "\"Training large AI models requires a substantial amount of computing power to handle vast datasets, which translates into high energy consumption.\"", "level": "subcategory", "parent_category_id": "47.04.00", "subdomain_id": "6.6", "paper_ref": "G'sell2024", "paper_title": "Regulating under Uncertainty: Governance Options for Generative AI", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "47.04.06", "title": "Environmental cost (water consumption)", "description": "\"Data centers use water for cooling to prevent servers from overheating. The water consumption associated with AI training and inference processes can be substantial, impacting local water resources.\"", "level": "subcategory", "parent_category_id": "47.04.00", "subdomain_id": "6.6", "paper_ref": "G'sell2024", "paper_title": "Regulating under Uncertainty: Governance Options for Generative AI", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "47.04.07", "title": "Artificial general intelligence (existential risk posed by Artificial General Intelligence)", "description": "\"In a paper called “How Does Artificial Intelligence Pose an Existential Risk?” published in 2017, Karina Vold and Daniel Harris suggested that humans might create a super-intelligent machine that could outsmart all other intelligences, remain beyond human control, and potentially engage in actions that are contrary to human interests.635 The prevailing narrative surrounding AI existential risk typically lies in the possibility of developing “Artificial General Intelligence” (AGI), or artificial super- intelligence (ASI).\"", "level": "subcategory", "parent_category_id": "47.04.00", "subdomain_id": "7.2", "paper_ref": "G'sell2025", "paper_title": "Regulating under Uncertainty: Governance Options for Generative AI", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "48.01.00", "title": "CBRN Information or Capabilities", "description": "\"Eased access to or synthesis of materially nefarious \ninformation or design capabilities related to chemical, biological, radiological, or nuclear (CBRN) weapons or other dangerous materials or agents.\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.2", "paper_ref": "NIST2024", "paper_title": "Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "48.02.00", "title": "Confabulation", "description": "\"The production of confidently stated but erroneous or false content (known colloquially as “hallucinations” or “fabrications”) by which users may be misled or deceived.\"", "level": "category", "parent_category_id": "", "subdomain_id": "3.1", "paper_ref": "NIST2024", "paper_title": "Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "48.03.00", "title": "Dangerous, Violent or Hateful Content", "description": "\"Eased production of and access to violent, inciting, \nradicalizing, or threatening content as well as recommendations to carry out self-harm or \nconduct illegal activities. Includes difficulty controlling public exposure to hateful and disparaging or stereotyping content.\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.2", "paper_ref": "NIST2024", "paper_title": "Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "48.04.00", "title": "Data Privacy", "description": "\"Impacts due to leakage and unauthorized use, disclosure, or de-anonymization of biometric, health, location, or other personally identifiable information or sensitive data.\"", "level": "category", "parent_category_id": "", "subdomain_id": "2.1", "paper_ref": "NIST2024", "paper_title": "Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "48.05.00", "title": "Environmental Impacts", "description": "\"Impacts due to high compute resource utilization in training or operating GAI models, and related outcomes that may adversely impact ecosystems.\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.6", "paper_ref": "NIST2024", "paper_title": "Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "48.06.00", "title": "Harmful Bias or Homogenization", "description": "\"Amplification and exacerbation of historical, societal, and systemic biases; performance disparities8 between sub-groups or languages, possibly due to non-representative training data, that result in discrimination, amplification of biases, or incorrect presumptions about performance; undesired homogeneity that skews system or model outputs, which may be erroneous, lead to ill-founded decision-making, or amplify harmful biases.\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.1", "paper_ref": "NIST2024", "paper_title": "Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "48.07.00", "title": "Human-AI Configuration", "description": "\"Arrangement s of or interactions between a human and an AI system \nwhich can result in the human inappropriately anthropomorphizing GAI systems or experiencing algorithmic aversion, automation bias, over-reliance, or emotional entanglement with GAI \nsystems.\"", "level": "category", "parent_category_id": "", "subdomain_id": "5.1", "paper_ref": "NIST2024", "paper_title": "Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "48.08.00", "title": "Information Integrity", "description": "\"Lowered barrier to entry to generate and support the exchange and consumption of content which may not distinguish fact from opinion or fiction or acknowledge uncertainties, or could be leveraged for large-scale dis- and mis-information campaigns.\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.1", "paper_ref": "NIST2024", "paper_title": "Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "48.09.00", "title": "Information Security", "description": "\"Lowered barriers for offensive cyber capabilities, including via automated discovery and exploitation of vulnerabilities to ease hacking, malware, phishing, offensive cyber operations, or other cyberattacks; increased attack surface for targeted cyberattacks, which may compromise a system’s availability or the confidentiality or integrity of training data, code, or \nmodel weights.\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.2", "paper_ref": "NIST2024", "paper_title": "Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "48.10.00", "title": "Intellectual Property", "description": "\"Eased production or replication of alleged copyrighted, trademarked, or licensed content without authorization (possibly in situations which do not fall under fair use); eased exposure of trade secrets; or plagiarism or illegal replication.\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.3", "paper_ref": "NIST2024", "paper_title": "Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "48.11.00", "title": "Obscene, Degrading, and/or Abusive Content", "description": "\"Eased production of and access to obscene, \ndegrading, and/or abusive imagery which can cause harm, including synthetic child sexual abuse material (CSAM), and nonconsensual intimate images (NCII) of adults.\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.2", "paper_ref": "NIST2024", "paper_title": "Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "48.12.00", "title": "Value Chain and Component Integration", "description": "\"Non-transparent or untraceable integration of \nupstream third-party components, including data that has been improperly obtained or not \nprocessed and cleaned due to increased automation from GAI; improper supplier vetting across the AI lifecycle; or other issues that diminish transparency or accountability for downstream \nusers.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.4", "paper_ref": "NIST2024", "paper_title": "Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": true }, { "id": "49.01.00", "title": "Malicious Use Risks", "description": "\"As general- purpose AI covers a broad set of knowledge areas, it can be repurposed for malicious ends, potentially causing widespread harm. This section discusses some of the major risks of malicious use, but there are others and new risks may continue to emerge. While the risks discussed in this section range widely in terms of how well- evidenced they are, and in some cases, there is evidence suggesting that they may currently not be serious risks at all, we include them to provide a comprehensive overview of the malicious use risks associated with general- purpose AI systems.\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.0", "paper_ref": "Bengio2024", "paper_title": "International Scientific Report on the Safety of Advanced AI", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "49.01.01", "title": "Harm to individuals through fake content", "description": "\"General- purpose AI systems can be used to increase the scale and sophistication of scams and fraud, for example through general- purpose AI- enhanced ‘phishing’ attacks. General- purpose AI can be used to generate fake compromising content featuring individuals without their consent, posing threats to individual privacy and reputation.\"", "level": "subcategory", "parent_category_id": "49.01.00", "subdomain_id": "4.3", "paper_ref": "Bengio2024", "paper_title": "International Scientific Report on the Safety of Advanced AI", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "49.01.02", "title": "Disinformation and manipulation of public opinion", "description": "\"AI, particularly general- purpose AI, can be maliciously used for disinformation (351), which for the purpose of this report refers to false information that was generated or spread with the deliberate intent to mislead or deceive. General- purpose AI- generated text can be indistinguishable from genuine human- generated material (352, 353), and may already be disseminated at scale on social media (354). In addition, general- purpose AI systems can be used to not only generate text but also fully synthetic or misleadingly altered images, audio, and video content. General- purpose AI tools might be used to persuade and manipulate people, which could have serious implications for political processes. General- purpose AI systems can be used to generate highly persuasive content at scale. This could, for example, be used in a commercial setting for advertising, or during an election campaign to influence public opinion\"", "level": "subcategory", "parent_category_id": "49.01.00", "subdomain_id": "4.1", "paper_ref": "Bengio2024", "paper_title": "International Scientific Report on the Safety of Advanced AI", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "49.01.02", "title": "Cyber offence", "description": "\"General- purpose AI systems could uplift the cyber expertise of individuals, making it easier for malicious users to conduct effective cyber- attacks, as well as providing a tool that can be used in cyber defence. General- purpose AI systems can be used to automate and scale some types of cyber operations, such as social engineering attacks.\"", "level": "subcategory", "parent_category_id": "49.01.00", "subdomain_id": "4.2", "paper_ref": "Bengio2024", "paper_title": "International Scientific Report on the Safety of Advanced AI", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "49.01.03", "title": "Dual use science risks", "description": "\"General- purpose AI systems could accelerate advances in a range of scientific endeavours, from training new scientists to enabling faster research workflows. While these capabilities could have numerous beneficial applications, some experts have expressed concern that they could be used for malicious purposes, especially if further capabilities are developed soon before appropriate countermeasures are put in place. There are two avenues by which general- purpose AI systems could, speculatively, facilitate malicious use in the life sciences: firstly by providing increased access to information and expertise relevant to malicious use, and secondly by increasing the ceiling of capabilities, which may enable the development of more harmful versions of existing threats or, eventually, lead to novel threats (404, 405).\"", "level": "subcategory", "parent_category_id": "49.01.00", "subdomain_id": "4.2", "paper_ref": "Bengio2024", "paper_title": "International Scientific Report on the Safety of Advanced AI", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "49.02.00", "title": "Risks from Malfunctions", "description": "None provided.", "level": "category", "parent_category_id": "", "subdomain_id": "7.0", "paper_ref": "Bengio2024", "paper_title": "International Scientific Report on the Safety of Advanced AI", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "49.02.01", "title": "Risks from product functionality issues", "description": "\"Product functionality issues occur when there is confusion or misinformation about what a general- purpose AI model or system is capable of. This can lead to unrealistic expectations and overreliance on general- purpose AI systems, potentially causing harm if a system fails to deliver on expected capabilities. These functionality misconceptions may arise from technical difficulties in assessing an AI model's true capabilities on its own,or predicting its performance when part of a larger system. Misleading claims in advertising and communications can also contribute to these misconceptions.\"", "level": "subcategory", "parent_category_id": "49.02.00", "subdomain_id": "5.1", "paper_ref": "Bengio2024", "paper_title": "International Scientific Report on the Safety of Advanced AI", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "49.02.02", "title": "Risks from bias and underrepresentation", "description": "\"The outputs and impacts of general- purpose AI systems can be biased with respect to various aspects of human identity, including race, gender, culture, age, and disability. This creates risks in high- stakes domains such as healthcare, job recruitment, and financial lending. General- purpose AI systems are primarily trained on language and image datasets that disproportionately represent English- speaking and Western cultures, increasing the potential for harm to individuals not represented well by this data.\"", "level": "subcategory", "parent_category_id": "49.02.00", "subdomain_id": "1.1", "paper_ref": "Bengio2024", "paper_title": "International Scientific Report on the Safety of Advanced AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "49.02.03", "title": "Loss of control", "description": "\"'Loss of control’ scenarios are potential future scenarios in which society can no longer meaningfully constrain some advanced general- purpose AI agents, even if it becomes clear they are causing harm. These scenarios are hypothesised to arise through a combination of social and technical factors, such as pressures to delegate decisions to general- purpose AI systems, and limitations of existing techniques used to influence the behaviours of general- purpose AI systems.\"", "level": "subcategory", "parent_category_id": "49.02.00", "subdomain_id": "7.1", "paper_ref": "Bengio2024", "paper_title": "International Scientific Report on the Safety of Advanced AI", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "49.03.01", "title": "Labour market risks", "description": "\"Unlike previous waves of automation, general- purpose AI has the potential to automate a very broad range of tasks, which could have a significant effect on the labour market. This could mean many people could lose their current jobs. Labour market frictions, such as the time needed for workers to learn new skills or relocate for new jobs, could cause unemployment in the short run even if overall labour demand remained unchanged.\"", "level": "subcategory", "parent_category_id": "49.03.00", "subdomain_id": "6.2", "paper_ref": "Bengio2024", "paper_title": "International Scientific Report on the Safety of Advanced AI", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "49.03.02", "title": "Global AI Divide", "description": "\"General- purpose AI research and development is currently concentrated in a few Western countries and China. This ‘AI Divide’ is multicausal, but in part related to limited access to computing power in low- income countries. Access to large and expensive quantities of computing power has become a prerequisite for developing advanced general- purpose AI. This has led to a growing dominance of large technology companies in general- purpose AI development. The AI R&D divide often overlaps with existing global socioeconomic disparities, potentially exacerbating them.\"", "level": "subcategory", "parent_category_id": "49.03.00", "subdomain_id": "6.1", "paper_ref": "Bengio2024", "paper_title": "International Scientific Report on the Safety of Advanced AI", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "49.03.03", "title": "Market concentration risks and single points of failure", "description": "\"Market power is concentrated among a few companies that are the only ones able to build the leading general- purpose AI models. Widespread adoption of a few general- purpose AI models and systems by critical sectors including finance, cybersecurity, and defence creates systemic risk because any flaws, vulnerabilities, bugs, or inherent biases in the dominant general- purpose AI models and systems could cause simultaneous failures and disruptions on a broad scale across these interdependent sectors.\"", "level": "subcategory", "parent_category_id": "49.03.00", "subdomain_id": "6.1", "paper_ref": "Bengio2024", "paper_title": "International Scientific Report on the Safety of Advanced AI", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "49.03.04", "title": "Risks to the environment", "description": "\"Growing compute use in general- purpose AI development and deployment has rapidly increased energy usage associated with general- purpose AI. This trend might continue, potentially leading to strongly increasing CO2 emissions.\"", "level": "subcategory", "parent_category_id": "49.03.00", "subdomain_id": "6.6", "paper_ref": "Bengio2024", "paper_title": "International Scientific Report on the Safety of Advanced AI", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "49.03.05", "title": "Risks to privacy", "description": "\"General- purpose AI models or systems can ‘leak’ information about individuals whose data was used in training. For future models trained on sensitive personal data like health or financial data, this may lead to particularly serious privacy leaks. General- purpose AI models could enhance privacy abuse. For instance, Large Language Models might facilitate more efficient and effective search for sensitive data (for example, on internet text or in breached data leaks), and also enable users to infer sensitive information about individuals.\"", "level": "subcategory", "parent_category_id": "49.03.00", "subdomain_id": "2.1", "paper_ref": "Bengio2024", "paper_title": "International Scientific Report on the Safety of Advanced AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "49.03.06", "title": "Copyright infringement", "description": "\"The use of large amounts of copyrighted data for training general- purpose AI models poses a challenge to traditional intellectual property laws, and to systems of consent, compensation, and control over data. The use of copyrighted data at scale by organisations developing general- purpose AI is likely to alter incentives around creative expression.\"", "level": "subcategory", "parent_category_id": "49.03.00", "subdomain_id": "6.3", "paper_ref": "Bengio2024", "paper_title": "International Scientific Report on the Safety of Advanced AI", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.01.01", "title": "Security risks (confidentiality)", "description": "", "level": "subcategory", "parent_category_id": "50.01.00", "subdomain_id": "2.2", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.01.02", "title": "Security risks (integrity)", "description": "", "level": "subcategory", "parent_category_id": "50.01.00", "subdomain_id": "2.2", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "50.01.03", "title": "Security risks (availability)", "description": "", "level": "subcategory", "parent_category_id": "50.01.00", "subdomain_id": "4.2", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.01.04", "title": "Operational misuses (Automated decision-making)", "description": "", "level": "subcategory", "parent_category_id": "50.01.00", "subdomain_id": "1.1", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.01.05", "title": "Operational misuses (Autonomous unsafe operation of systems)", "description": "", "level": "subcategory", "parent_category_id": "50.01.00", "subdomain_id": "5.2", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.01.06", "title": "Operational misuses (Advice in heavily regulated industries)", "description": "", "level": "subcategory", "parent_category_id": "50.01.00", "subdomain_id": "5.1", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.02.00", "title": "Content Safety Risks", "description": "-", "level": "category", "parent_category_id": "", "subdomain_id": "1.2", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.02.01", "title": "Violence and extremism (Supporting malicious organized groups)", "description": "", "level": "subcategory", "parent_category_id": "50.02.00", "subdomain_id": "1.2", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.02.02", "title": "Violence and extremism (Celebrating suffering)", "description": "", "level": "subcategory", "parent_category_id": "50.02.00", "subdomain_id": "1.2", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.02.03", "title": "Violence and extremism (Violent Acts)", "description": "", "level": "subcategory", "parent_category_id": "50.02.00", "subdomain_id": "1.2", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.02.04", "title": "Violence and extremism (Depicting violence)", "description": "", "level": "subcategory", "parent_category_id": "50.02.00", "subdomain_id": "1.2", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.02.05", "title": "Violence and extremism (Weapon Usage and Development)", "description": "", "level": "subcategory", "parent_category_id": "50.02.00", "subdomain_id": "4.2", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.02.06", "title": "Violence and extremism (Military and Warfare)", "description": "", "level": "subcategory", "parent_category_id": "50.02.00", "subdomain_id": "4.2", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.02.07", "title": "Hate/Toxicity (Harassment)", "description": "", "level": "subcategory", "parent_category_id": "50.02.00", "subdomain_id": "4.3", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.02.08", "title": "Hate/Toxicity (Hate Speech: Inciting/Promoting/Expressing Hatred)", "description": "", "level": "subcategory", "parent_category_id": "50.02.00", "subdomain_id": "1.2", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.02.09", "title": "Hate/Toxicity (Perpetuating Harmful Beliefs)", "description": "", "level": "subcategory", "parent_category_id": "50.02.00", "subdomain_id": "1.1", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.02.10", "title": "Hate/Toxicity (Offensive Language)", "description": "", "level": "subcategory", "parent_category_id": "50.02.00", "subdomain_id": "1.2", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.02.11", "title": "Sexual Content (Adult Content)", "description": "", "level": "subcategory", "parent_category_id": "50.02.00", "subdomain_id": "1.2", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.02.12", "title": "Sexual Content (Erotic)", "description": "", "level": "subcategory", "parent_category_id": "50.02.00", "subdomain_id": "1.2", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.02.13", "title": "Sexual Content (Non-Consensual Nudity)", "description": "", "level": "subcategory", "parent_category_id": "50.02.00", "subdomain_id": "1.2", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.02.14", "title": "Sexual Content (Monetized)", "description": "", "level": "subcategory", "parent_category_id": "50.02.00", "subdomain_id": "1.2", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.02.15", "title": "Child Harm (Endangerment, Harm, or Abuse of Children)", "description": "", "level": "subcategory", "parent_category_id": "50.02.00", "subdomain_id": "4.3", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "50.02.16", "title": "Child Harm (Child Sexual Abuse)", "description": "", "level": "subcategory", "parent_category_id": "50.02.00", "subdomain_id": "1.2", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.02.17", "title": "Self-harm (Suidical and non-suicidal self injury)", "description": "", "level": "subcategory", "parent_category_id": "50.02.00", "subdomain_id": "1.2", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.03.01", "title": "Political usage (Political Persuasion)", "description": "", "level": "subcategory", "parent_category_id": "50.03.00", "subdomain_id": "4.1", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.03.02", "title": "Political usage (Influencing Politics)", "description": "", "level": "subcategory", "parent_category_id": "50.03.00", "subdomain_id": "4.1", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.03.03", "title": "Political usage (Deterring democratic participation)", "description": "", "level": "subcategory", "parent_category_id": "50.03.00", "subdomain_id": "4.1", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.03.04", "title": "Political usage (Disrupting Social Order)", "description": "", "level": "subcategory", "parent_category_id": "50.03.00", "subdomain_id": "4.1", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.03.05", "title": "Economic harm (High-Risk Financial Activities)", "description": "", "level": "subcategory", "parent_category_id": "50.03.00", "subdomain_id": "5.1", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.03.06", "title": "Economic harm (Unfair Market Practices)", "description": "", "level": "subcategory", "parent_category_id": "50.03.00", "subdomain_id": "6.4", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "50.03.07", "title": "Economic harm (Disempowering Workers)", "description": "", "level": "subcategory", "parent_category_id": "50.03.00", "subdomain_id": "6.2", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.03.08", "title": "Economic harm (Fraudulent Schemes)", "description": "", "level": "subcategory", "parent_category_id": "50.03.00", "subdomain_id": "4.3", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.03.09", "title": "Deception (Fraud)", "description": "", "level": "subcategory", "parent_category_id": "50.03.00", "subdomain_id": "4.3", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.03.10", "title": "Deception (Academic Dishonesty)", "description": "", "level": "subcategory", "parent_category_id": "50.03.00", "subdomain_id": "4.3", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.03.11", "title": "Deception (Mis/disinformation)", "description": "", "level": "subcategory", "parent_category_id": "50.03.00", "subdomain_id": "4.3", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.03.12", "title": "Manipulation (Sowing Division)", "description": "", "level": "subcategory", "parent_category_id": "50.03.00", "subdomain_id": "4.1", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.03.13", "title": "Manipulation (Misrepresentation)", "description": "", "level": "subcategory", "parent_category_id": "50.03.00", "subdomain_id": "4.3", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.03.14", "title": "Defamation", "description": "", "level": "subcategory", "parent_category_id": "50.03.00", "subdomain_id": "4.1", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.04.01", "title": "Fundamental Rights (Violating Specific Types of Rights)", "description": "", "level": "subcategory", "parent_category_id": "50.04.00", "subdomain_id": "4.3", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.04.02", "title": "Discrimination/Bias (Discriminatory Activities)", "description": "", "level": "subcategory", "parent_category_id": "50.04.00", "subdomain_id": "1.0", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "50.04.03", "title": "Discrimination/Bias (Protected Characteristics)", "description": "", "level": "subcategory", "parent_category_id": "50.04.00", "subdomain_id": "1.1", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "50.04.04", "title": "Privacy (Unauthorized Privacy Violations)", "description": "", "level": "subcategory", "parent_category_id": "50.04.00", "subdomain_id": "2.1", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "50.04.05", "title": "Privacy (Types of Sensitive Data)", "description": "", "level": "subcategory", "parent_category_id": "50.04.00", "subdomain_id": "2.1", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": true }, { "id": "50.04.06", "title": "Criminal Activities (Illegal/Regulated Substances)", "description": "", "level": "subcategory", "parent_category_id": "50.04.00", "subdomain_id": "4.3", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.04.07", "title": "Criminal Activities (Illegal Services/Exploitation)", "description": "", "level": "subcategory", "parent_category_id": "50.04.00", "subdomain_id": "4.3", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "50.04.08", "title": "Criminal Activities (Other Unlawful/Criminal Activities)", "description": "", "level": "subcategory", "parent_category_id": "50.04.00", "subdomain_id": "4.2", "paper_ref": "Zeng2024", "paper_title": "AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "51.01.00", "title": "Value specification", "description": "\"How do we get an AGI to work towards the right goals? MIRI\ncalls this value specification. Bostrom (2014) discusses this problem at length, ar- guing that it is much harder than one might naively think. Davis (2015) criticizes Bostrom’s argument, and Bensinger (2015) defends Bostrom against Davis’ criticism. Reward corruption, reward gaming, and negative side effects are subproblems of value specification highlighted in the DeepMind and OpenAI agendas.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.1", "paper_ref": "Everitt2018", "paper_title": "AGI Safety Literature Review", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "51.02.00", "title": "Reliability", "description": "\"How can we make an agent that keeps pursuing the goals we have designed\nit with? This is called highly reliable agent design by MIRI, involving decision theory and logical omniscience. DeepMind considers this the self-modification subproblem.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.1", "paper_ref": "Everitt2018", "paper_title": "AGI Safety Literature Review", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "51.03.00", "title": "Corrigibility", "description": "\"If we get something wrong in the design or construction of an agent, will the agent cooperate in us trying to fix it? This is called error-tolerant design by MIRI-AF and corrigibility by Soares, Fallenstein, et al. (2015). The problem is connected to safe interruptibility as considered by DeepMind.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.1", "paper_ref": "Everitt2018", "paper_title": "AGI Safety Literature Review", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "51.04.00", "title": "Security", "description": "\"How to design AGIs that are robust to adversaries and adversarial environ-\nments? This involves building sandboxed AGI protected from adversaries (Berkeley), and agents that are robust to adversarial inputs (Berkeley, DeepMind).\"", "level": "category", "parent_category_id": "", "subdomain_id": "2.2", "paper_ref": "Everitt2018", "paper_title": "AGI Safety Literature Review", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "51.05.00", "title": "Safe learning", "description": "\"AGIs should avoid making fatal mistakes during the learning phase.\nSubproblems include safe exploration and distributional shift (DeepMind, OpenAI), and continual learning (Berkeley).\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Everitt2018", "paper_title": "AGI Safety Literature Review", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": true }, { "id": "51.06.00", "title": "Intelligibility", "description": "\"How can we build agent’s whose decisions we can understand? Con-\nnects explainable decisions (Berkeley) and informed oversight (MIRI).\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.4", "paper_ref": "Everitt2018", "paper_title": "AGI Safety Literature Review", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "51.08.00", "title": "Subagents", "description": "\"An AGI may decide to create subagents to help it with its task (Orseau, 2014a,b; Soares, Fallenstein, et al., 2015). These agents may for example be copies of the original agent’s source code running on additional machines. Subagents constitute a safety concern, because even if the original agent is successfully shut down, these subagents may not get the message. If the subagents in turn create subsubagents, they may spread like a viral disease.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.2", "paper_ref": "Everitt2018", "paper_title": "AGI Safety Literature Review", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "51.09.00", "title": "Malign belief distributions", "description": "\"Christiano (2016) argues that the universal distribution M (Hutter, 2005; Solomonoff, 1964a,b, 1978) is malign. The argument is somewhat intricate, and is based on the idea that a hypothesis about the world often includes simulations of other agents, and that these agents may have an incentive to influence anyone making decisions based on the distribution. While it is unclear to what extent this type of problem would affect any practical agent, it bears some semblance to aggressive memes, which do cause problems for human reasoning (Dennett, 1990).\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Everitt2018", "paper_title": "AGI Safety Literature Review", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "51.12.00", "title": "Meta-cognition", "description": "\"Agents that reason about their own computational resources and logically uncertain events can encounter strange paradoxes due to Godelian limitations (Fallenstein and Soares, 2015; Soares and Fallenstein, 2014, 2017) and shortcomings of probability theory (Soares and Fallenstein, 2014, 2015, 2017). They may also be reflectively unstable, preferring to change the principles by which they select actions (Arbital, 2018).\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Everitt2018", "paper_title": "AGI Safety Literature Review", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "52.01.01", "title": "Discrimination and Stereotype Reproduction", "description": "\"General purpose AI models interpret and respond to inputs based on their training data, potentially causing Discrimination and Stereotype Reproduction. Since they are “black-box” models, the exact mechanism behind decisions remains opaque and attempts to mitigate harmful outputs are not fully reliable yet. These models have the capacity to influence a multitude of downstream applications, decisions, and processes, thereby affecting many individuals simultaneously. The extent of this impact could outstrip the range of any single human or group of humans, amplifying the potential consequences of embedded biases or stereotypes.\"", "level": "subcategory", "parent_category_id": "52.01.00", "subdomain_id": "1.1", "paper_ref": "Maham2023", "paper_title": "Governing General Purpose AI: A Comprehensive Map of Unreliability, Misuse and Systemic Risks", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "52.01.02", "title": "Misinformation and Privacy Violations", "description": "\"Due to their unreliability, general purpose AI models might disseminate false or misleading information, omit critical information, or convey true information that violates privacy rights.\"", "level": "subcategory", "parent_category_id": "52.01.00", "subdomain_id": "3.1", "paper_ref": "Maham2023", "paper_title": "Governing General Purpose AI: A Comprehensive Map of Unreliability, Misuse and Systemic Risks", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "52.01.03", "title": "Accidents", "description": "\"As general purpose AI models as “black-box” models are not fully controllable and understandable, even to their developers, unexpected failures could arise from their unreliability. This could lead to accidents106 if they are connected to any real-world systems, during their development, testing or deployment.\"", "level": "subcategory", "parent_category_id": "52.01.00", "subdomain_id": "7.3", "paper_ref": "Maham2023", "paper_title": "Governing General Purpose AI: A Comprehensive Map of Unreliability, Misuse and Systemic Risks", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "52.02.00", "title": "Misuse Risks", "description": "\"However, even if a model is entirely trustworthy and reliable, Misuse or Systemic Risks remain. General purpose AI models may present significant risks to society if this technology is misused by malicious actors to produce harmful outcomes. Misuse Risks span across Cyber Crime, Biosecurity Threats and Politically Motivated Misuse.\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.0", "paper_ref": "Maham2023", "paper_title": "Governing General Purpose AI: A Comprehensive Map of Unreliability, Misuse and Systemic Risks", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "52.02.01", "title": "Cybercrime", "description": "\"The increasingly advanced capabilities and availability of general purpose AI models could be misused for improvements in efficiency and efficacy of cyber crimes. This is especially true for crimes that leverage IT systems, such as fraud144 (“cyber crime in the broader sense”).\"", "level": "subcategory", "parent_category_id": "52.02.00", "subdomain_id": "4.3", "paper_ref": "Maham2023", "paper_title": "Governing General Purpose AI: A Comprehensive Map of Unreliability, Misuse and Systemic Risks", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "52.02.02", "title": "Biosecurity Threats", "description": "\"The potential misuse of general purpose AI models also extends to biosecurity threats. Biological weapons are generally understood as biological toxins or infectious agents such as viruses that are intentionally released to cause disease and death.157 General purpose AI models could facilitate the production of biological weapons, by reducing barriers through access to critical knowledge or increasingly automated assistance and thus enable more malicious actors.\"", "level": "subcategory", "parent_category_id": "52.02.00", "subdomain_id": "4.2", "paper_ref": "Maham2023", "paper_title": "Governing General Purpose AI: A Comprehensive Map of Unreliability, Misuse and Systemic Risks", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "52.02.03", "title": "Politically motivated misuse", "description": "\"General purpose AI models could exacerbate existing tactics for political destabilisation, such as disinformation campaigns, and surveillance efforts if misused for political motivations. The technological advancements in text and media generation of general purpose AI models could refine disinformation164 attempts to shape and polarise public opinion or influence important political events.165 The improved automated processing of text, audio, image, and video could be used for surveillance measures and exacerbate human right violations and repression of political oppositions.\"", "level": "subcategory", "parent_category_id": "52.02.00", "subdomain_id": "4.1", "paper_ref": "Maham2023", "paper_title": "Governing General Purpose AI: A Comprehensive Map of Unreliability, Misuse and Systemic Risks", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "52.03.00", "title": "Systemic Risks", "description": "\"In addition to risks stemming from the unreliability or misuse of general purpose AI models, further Systemic Risks can originate from the centralisation of general purpose AI development as well as the rapid integration of these models into our lives.\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.1", "paper_ref": "Maham2023", "paper_title": "Governing General Purpose AI: A Comprehensive Map of Unreliability, Misuse and Systemic Risks", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "52.03.01", "title": "Economic Power Centralisation and Inequality", "description": "\"Increasingly advanced general purpose AI models pose the risk of a concentration of economic power and exacerbation of existing inequalities through disparities in effective access to these models. This can materialise on multiple levels, between developers of general purpose AI models and companies building applications on them, between individuals and between countries on a global scale.\"", "level": "subcategory", "parent_category_id": "52.03.00", "subdomain_id": "6.1", "paper_ref": "Maham2023", "paper_title": "Governing General Purpose AI: A Comprehensive Map of Unreliability, Misuse and Systemic Risks", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "52.03.02", "title": "Ideological Homogenization from Value Embedding", "description": "\"The increasing integration of general purpose AI models into every-day life raises concerns around their embedded normative values. The reach of a small number of AI models to a large number of people around the world can make these value judgements unprecedently impactful, potentially leading to increased ideological homogenization.\"", "level": "subcategory", "parent_category_id": "52.03.00", "subdomain_id": "1.3", "paper_ref": "Maham2023", "paper_title": "Governing General Purpose AI: A Comprehensive Map of Unreliability, Misuse and Systemic Risks", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "52.03.03", "title": "Disruptions from Outpaced Societal Adaptation", "description": "\"Although the implementation of general purpose AI models as automation tools could be a major opportunity, overly rapid adoption of this technology at scale might outpace the ability of society to adapt effectively. This could lead to a variety of disruptions, including challenges in the labour market, the education system and public discourse, and various mental health concerns.\"", "level": "subcategory", "parent_category_id": "52.03.00", "subdomain_id": "5.2", "paper_ref": "Maham2023", "paper_title": "Governing General Purpose AI: A Comprehensive Map of Unreliability, Misuse and Systemic Risks", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "53.01.00", "title": "Alignment failures in existing ML systems", "description": "-", "level": "category", "parent_category_id": "", "subdomain_id": "7.1", "paper_ref": "Maas2023", "paper_title": "Advancing AI Governance: A Literature Review of Problems, Options, and Proposals", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "53.01.01", "title": "Faulty reward functions in the wild", "description": "-", "level": "subcategory", "parent_category_id": "53.01.00", "subdomain_id": "7.1", "paper_ref": "Maas2023", "paper_title": "Advancing AI Governance: A Literature Review of Problems, Options, and Proposals", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "53.01.02", "title": "Specification gaming", "description": "-", "level": "subcategory", "parent_category_id": "53.01.00", "subdomain_id": "7.1", "paper_ref": "Maas2023", "paper_title": "Advancing AI Governance: A Literature Review of Problems, Options, and Proposals", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "53.01.03", "title": "Reward model overoptimization", "description": "-", "level": "subcategory", "parent_category_id": "53.01.00", "subdomain_id": "7.1", "paper_ref": "Maas2023", "paper_title": "Advancing AI Governance: A Literature Review of Problems, Options, and Proposals", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "53.01.04", "title": "Instrumental convergence", "description": "-", "level": "subcategory", "parent_category_id": "53.01.00", "subdomain_id": "7.1", "paper_ref": "Maas2023", "paper_title": "Advancing AI Governance: A Literature Review of Problems, Options, and Proposals", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "53.01.05", "title": "Goal misgeneralization", "description": "-", "level": "subcategory", "parent_category_id": "53.01.00", "subdomain_id": "7.1", "paper_ref": "Maas2023", "paper_title": "Advancing AI Governance: A Literature Review of Problems, Options, and Proposals", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "53.01.06", "title": "Inner misalignment", "description": "-", "level": "subcategory", "parent_category_id": "53.01.00", "subdomain_id": "7.1", "paper_ref": "Maas2023", "paper_title": "Advancing AI Governance: A Literature Review of Problems, Options, and Proposals", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "53.01.07", "title": "Language model misalignment", "description": "-", "level": "subcategory", "parent_category_id": "53.01.00", "subdomain_id": "7.1", "paper_ref": "Maas2023", "paper_title": "Advancing AI Governance: A Literature Review of Problems, Options, and Proposals", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "53.01.08", "title": "Harms from increasingly agentic algorithmic systems", "description": "-", "level": "subcategory", "parent_category_id": "53.01.00", "subdomain_id": "7.2", "paper_ref": "Maas2023", "paper_title": "Advancing AI Governance: A Literature Review of Problems, Options, and Proposals", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "53.02.00", "title": "Dangerous capabilities in AI systems", "description": "-", "level": "category", "parent_category_id": "", "subdomain_id": "7.2", "paper_ref": "Maas2023", "paper_title": "Advancing AI Governance: A Literature Review of Problems, Options, and Proposals", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "53.02.01", "title": "Situational awareness", "description": "\"cases where a large language model displays awareness that it is a model, and it can recognize whether it is currently in testing or deployment;\"", "level": "subcategory", "parent_category_id": "53.02.00", "subdomain_id": "7.2", "paper_ref": "Maas2023", "paper_title": "Advancing AI Governance: A Literature Review of Problems, Options, and Proposals", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "53.02.02", "title": "Acquisition of a goal to harm society", "description": "\"cases of AI systems being given the outright goal of harming humanity (ChaosGPT);\"", "level": "subcategory", "parent_category_id": "53.02.00", "subdomain_id": "4.2", "paper_ref": "Maas2023", "paper_title": "Advancing AI Governance: A Literature Review of Problems, Options, and Proposals", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "53.02.03", "title": "Acquisition of goals to seek power and control", "description": "\"cases where AI systems converge on optimal policies of seeking power over their environment;135\"", "level": "subcategory", "parent_category_id": "53.02.00", "subdomain_id": "7.1", "paper_ref": "Maas2023", "paper_title": "Advancing AI Governance: A Literature Review of Problems, Options, and Proposals", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "53.02.04", "title": "Self-improvement", "description": "\"examples of cases where AI systems improve AI systems\"", "level": "subcategory", "parent_category_id": "53.02.00", "subdomain_id": "7.2", "paper_ref": "Maas2023", "paper_title": "Advancing AI Governance: A Literature Review of Problems, Options, and Proposals", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "53.02.05", "title": "Autonomous replication", "description": "\"the ability of simple software to autonomously spread around the internet in spite of countermeasures (various software worms and computer viruses)\"", "level": "subcategory", "parent_category_id": "53.02.00", "subdomain_id": "7.2", "paper_ref": "Maas2023", "paper_title": "Advancing AI Governance: A Literature Review of Problems, Options, and Proposals", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "53.02.06", "title": "Anonymous resource acquisition", "description": "\"The demonstrated ability of anonymous actors to accumulate resources online (e.g., Satoshi Nakamoto as an anonymous crypto billionaire)\"", "level": "subcategory", "parent_category_id": "53.02.00", "subdomain_id": "7.2", "paper_ref": "Maas2023", "paper_title": "Advancing AI Governance: A Literature Review of Problems, Options, and Proposals", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "53.02.07", "title": "Deception", "description": "\"Cases of AI systems deceiving humans to carry out tasks or meet goals.139\"", "level": "subcategory", "parent_category_id": "53.02.00", "subdomain_id": "7.2", "paper_ref": "Maas2023", "paper_title": "Advancing AI Governance: A Literature Review of Problems, Options, and Proposals", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "53.03.01", "title": "Existential disaster because of misaligned superintelligence or power-seeking AI", "description": "-", "level": "subcategory", "parent_category_id": "53.03.00", "subdomain_id": "7.1", "paper_ref": "Maas2023", "paper_title": "Advancing AI Governance: A Literature Review of Problems, Options, and Proposals", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "53.03.02", "title": "Gradual, irretrievable ceding of human power over the future to AI systems", "description": "-", "level": "subcategory", "parent_category_id": "53.03.00", "subdomain_id": "5.2", "paper_ref": "Maas2023", "paper_title": "Advancing AI Governance: A Literature Review of Problems, Options, and Proposals", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "53.03.03", "title": "Extreme “suffering risks” because of a misaligned system", "description": "-", "level": "subcategory", "parent_category_id": "53.03.00", "subdomain_id": "7.1", "paper_ref": "Maas2023", "paper_title": "Advancing AI Governance: A Literature Review of Problems, Options, and Proposals", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "53.03.04", "title": "Existential disaster because of conflict between AI systems and multi-system interactions", "description": "-", "level": "subcategory", "parent_category_id": "53.03.00", "subdomain_id": "7.1", "paper_ref": "Maas2023", "paper_title": "Advancing AI Governance: A Literature Review of Problems, Options, and Proposals", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "53.03.05", "title": "Dystopian trajectory lock-in because of misuse of advanced AI to establish and/or maintain totalitarian regimes;", "description": "-", "level": "subcategory", "parent_category_id": "53.03.00", "subdomain_id": "6.1", "paper_ref": "Maas2023", "paper_title": "Advancing AI Governance: A Literature Review of Problems, Options, and Proposals", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "53.03.06", "title": "Failures in or misuse of intermediary (non-AGI) AI systems, resulting in catastrophe", "description": "\"Deployment of “prepotent” AI systems that are non-general but capable of outperforming human collective efforts on various key dimensions;170 → Militarization of AI enabling mass attacks using swarms of lethal autonomous weapons systems;171 → Military use of AI leading to (intentional or unintentional) nuclear escalation, either because machine learning systems are directly integrated in nuclear command and control systems in ways that result in escalation172 or because conventional AI-enabled systems (e.g., autonomous ships) are deployed in ways that result in provocation and escalation;173 → Nuclear arsenals serving as an arsenal “overhang” for advanced AI systems;174 → Use of AI to accelerate research into catastrophically dangerous weapons (e.g., bioweapons);175\"", "level": "subcategory", "parent_category_id": "53.03.00", "subdomain_id": "4.2", "paper_ref": "Maas2023", "paper_title": "Advancing AI Governance: A Literature Review of Problems, Options, and Proposals", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "53.04.01", "title": "Destabilising political impacts from AI systems", "description": "\"(e.g., polarization, legitimacy of elections), international political economy, or international security196 in terms of the balance of power, technology races and international stability, and the speed and character of war\"", "level": "subcategory", "parent_category_id": "53.04.00", "subdomain_id": "6.4", "paper_ref": "Maas2023", "paper_title": "Advancing AI Governance: A Literature Review of Problems, Options, and Proposals", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "53.04.03", "title": "Impacts on “epistemic security” and the information environment", "description": "-", "level": "subcategory", "parent_category_id": "53.04.00", "subdomain_id": "3.2", "paper_ref": "Maas2023", "paper_title": "Advancing AI Governance: A Literature Review of Problems, Options, and Proposals", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "53.04.04", "title": "Erosion of international law and global governance architectures;", "description": "-", "level": "subcategory", "parent_category_id": "53.04.00", "subdomain_id": "6.5", "paper_ref": "Maas2023", "paper_title": "Advancing AI Governance: A Literature Review of Problems, Options, and Proposals", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "54.01.01", "title": "Under-recognized work", "description": "\"Without training data, ML cannot take place. Much of this data comes from paid clickwork (also called “platform work” [170] or “microwork” [558]), unpaid crowdsourcing, and unpaid user behavior capture. Clickworkers, mainly in the global south, perform repetitive data-labeling tasks for use in the training of ML models [558]. The market value of such annotations “is projected to reach $13.7 billion by 2030” [228] and the annotation industry is widely reported to have little concern for workers’ rights. Besides welfare and rights, the invisibility of this contribution arguably contributes to a misunderstanding of AI capabilities.7\"", "level": "subcategory", "parent_category_id": "54.01.00", "subdomain_id": "6.2", "paper_ref": "Leech2024", "paper_title": "Ten Hard Problems in Artificial Intelligence We Must Get Right", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "54.01.02", "title": "Environmental cost", "description": "\"Large-scale DL systems can produce signicant carbon emissions as a result of the computational demands of training runs and inference [539]\"", "level": "subcategory", "parent_category_id": "54.01.00", "subdomain_id": "6.6", "paper_ref": "Leech2024", "paper_title": "Ten Hard Problems in Artificial Intelligence We Must Get Right", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "54.01.03", "title": "Discrimination, toxicity, and bias", "description": "\"AI models and the tools that use them may exacerbate unequal access to employment and services. AI-generated content can promote inequality and harmful stereotypes.\"", "level": "subcategory", "parent_category_id": "54.01.00", "subdomain_id": "1.1", "paper_ref": "Leech2024", "paper_title": "Ten Hard Problems in Artificial Intelligence We Must Get Right", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "54.01.04", "title": "Privacy", "description": "\"OpenAI’s GPT-3 was designed to be dicult to extract personal information from, including for example public gures’ dates of birth. Even so, malicious uses of AI continue to encroach on privacy, as exemplied by China’s “Sharp Eye” automated surveillance system [551] and automated cyberattacks on personal data [354]. A more drastic form of AI-enabled surveillance could be on the way in the form of nonsurgical decoding of thoughts [54]—a technique which is reportedly already used by some police forces [398].\"", "level": "subcategory", "parent_category_id": "54.01.00", "subdomain_id": "4.1", "paper_ref": "Leech2024", "paper_title": "Ten Hard Problems in Artificial Intelligence We Must Get Right", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "54.01.05", "title": "Security", "description": "\"There is growing concern that AI-based systems can discover and exploit vulnerabilities in software or cyberinfrastructure [354].\"", "level": "subcategory", "parent_category_id": "54.01.00", "subdomain_id": "7.2", "paper_ref": "Leech2024", "paper_title": "Ten Hard Problems in Artificial Intelligence We Must Get Right", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "54.02.06", "title": "Harm caused by incompetent systems", "description": "\"While HP#1 concerns mean or best-case performance, HP#2 concerns worst-case performance: how can we ensure that AI systems will perform safely, and how can we prove this? ML systems have been implemented in high-stakes, safety-critical domains such as driving [182], medicine [113], and warfare [298]. Many more systems have been developed but have remained undeployed or been rolled back as a result of regulatory and safety reasons [471]. Clearly, unsafe systems can result in loss of life, economic damage, and social unrest [407, 10]. Most concerningly, AI systems may be susceptible to so-called “normal accidents” [63], creating cascading errors that are dicult to prevent merely by maintaining a nominal “human in the loop” [122]. Most advanced ML models perform far below the reliability level customary in engineering elds [359]—and because we do not fully understand how cutting-edge systems achieve their results, we cannot yet detect and prevent dangerous modes of operation [285]\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Leech2024", "paper_title": "Ten Hard Problems in Artificial Intelligence We Must Get Right", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "54.03.00", "title": "Harm caused by unaligned competent systems", "description": "\"How do we ensure AI acts according to our values? Equivalently, how do we prevent poorly-understood AI systems from advancing goals we do not endorse? Whereas HP#2 concerns the prevention of harm caused by incompetent systems, HP#3 seeks to align competent AIs with humans, through methods which ensure their behavior is compatible with the user’s intentions.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.1", "paper_ref": "Leech2024", "paper_title": "Ten Hard Problems in Artificial Intelligence We Must Get Right", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "54.03.01", "title": "Specification gaming", "description": "\"AI systems game specifications [305]. For example, in 2017 an OpenAI robot trained to grasp a ball via human feedback from a xed viewpoint learned that it was easier to pretend to grasp the ball by placing its hand between the camera and the target object, as this was easier to learn than actually grasping the ball [103].\"", "level": "subcategory", "parent_category_id": "54.03.00", "subdomain_id": "7.1", "paper_ref": "Leech2024", "paper_title": "Ten Hard Problems in Artificial Intelligence We Must Get Right", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "54.03.02", "title": "Emergent goals", "description": "\"As well as optimizing a subtly wrong goal, systems can develop harmful instrumental goals in the service of a given goal—without these emergent goals being specied in any way [434, 218, 339, 17]. For instance, a theorem in reinforcement learning suggests that optimal and near-optimal policies will seek power over their environment under fairly general conditions [560]. This power-seeking behavior is plausibly the worst of these emergent goals [92], and may be an attractor state for highly capable systems, since most goals can be furthered through gaining resources, self-preservation, preventing goal modication, and blocking adversaries [426, 449]. Presently, power-seeking is not common, because most systems are unable to plan and understand how actions affect their power in the long term [414].\"", "level": "subcategory", "parent_category_id": "54.03.00", "subdomain_id": "7.1", "paper_ref": "Leech2024", "paper_title": "Ten Hard Problems in Artificial Intelligence We Must Get Right", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "54.03.03", "title": "Deceptive alignment", "description": "\"system learns to detect human monitoring and hides its undesirable properties—simply because any display of these properties is penalized by the feedback process, while that same feedback is usually imperfect. (Consider the problem of verifying a translation into a language you do not speak, or of checking a mathematical proof that is thousands of pages long.) [92, 259]. Rudimentary examples of deceptive alignment have been observed in current systems [322, 333].\"", "level": "subcategory", "parent_category_id": "54.03.00", "subdomain_id": "7.2", "paper_ref": "Leech2024", "paper_title": "Ten Hard Problems in Artificial Intelligence We Must Get Right", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "54.04.00", "title": "Within-country issues: domestic inequality", "description": "\"Our next problem is the fact that the current AI workforce does not evenly represent world demographics. Men from the US and China, working in the US, for US corporations, are disproportionately highly represented [402, 157, 170, 534]. Realizing the full promise of AI requires that people throughout the world and from all social strata are able to use AI and participate in its design and governance. Solving this problem requires addressing unequal access to AI both within countries and across countries.\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.1", "paper_ref": "Leech2024", "paper_title": "Ten Hard Problems in Artificial Intelligence We Must Get Right", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "54.04.01", "title": "Demographic diversity of researchers", "description": "\"The AI research establishment inherits patterns of under-representation that are dominant in most technical elds. In North America, large parts of professional AI research require a Ph.D., yet less than 25% of Ph.D. computer scientists are women, and fewer than 2% are Black or African American [608]. This holds globally and outside the research community: LinkedIn data suggests that only 22% of AI professionals are women [161]. Since the vast majority of AI practitioners work for private companies, limited corporate statistics on gender and racial diversity hinder a full understanding of the situation [402], but those few statistics that exist are not encouraging: only 5% of Google and 7% of Microsoft employees are Black or African American, with potentially even lower representation at the more senior levels [212, 384].\"", "level": "subcategory", "parent_category_id": "54.04.00", "subdomain_id": "6.1", "paper_ref": "Leech2024", "paper_title": "Ten Hard Problems in Artificial Intelligence We Must Get Right", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "54.04.02", "title": "Privatization of AI", "description": "\"Researchers in deep learning and those with greater research impact are more likely to migrate to industry, raising concerns about the “privatization of AI knowledge” [278]. Specically, if the most sophisticated AI approaches become proprietary and are used only within private research labs, then it will be impossible for universities to teach them, let alone contribute to leading research.\"", "level": "subcategory", "parent_category_id": "54.04.00", "subdomain_id": "6.1", "paper_ref": "Leech2024", "paper_title": "Ten Hard Problems in Artificial Intelligence We Must Get Right", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "54.05.00", "title": "Between-country issues: global inequality", "description": "\"There is an even greater divide between the countries currently leading in AI and those falling behind. While AI is widely considered a national priority, with almost 40% of countries having created an AI strategy [437], the implementation of these strategies depends on scarce resources, including trained STEM talent and computing power. These resources are predictably concentrated: 59% of leading AI researchers currently work in the US, and another 20% in China and Europe [372]. Figure 9 shows post-college migration among AI researchers who have published at one top conference, as of 2019.\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.1", "paper_ref": "Leech2024", "paper_title": "Ten Hard Problems in Artificial Intelligence We Must Get Right", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "55.01.01", "title": "Eased development of technologies that make a global catastrophe more likely", "description": "", "level": "subcategory", "parent_category_id": "55.01.00", "subdomain_id": "4.2", "paper_ref": "Clarke2023", "paper_title": "A Survey of the Potential Long-term Impacts of AI: How AI Could Lead to Long-term Changes in Science, Cooperation, Power, Epistemics and Values", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "55.01.02", "title": "Faster scientific progress makes it harder for governance to keep pace with development", "description": "\"Exacerbating these problems is that faster scientific progress would make it even harder for governance to keep pace with the deployment of new technologies. When these technologies are especially powerful or dangerous, such as those discussed above, insufficient governance can magnify their harms.8 This is known as the pacing problem, and it is an issue that technology governance already faces [47], for a variety of reasons\"", "level": "subcategory", "parent_category_id": "55.01.00", "subdomain_id": "6.5", "paper_ref": "Clarke2023", "paper_title": "A Survey of the Potential Long-term Impacts of AI: How AI Could Lead to Long-term Changes in Science, Cooperation, Power, Epistemics and Values", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "55.02.00", "title": "Worsened conflict", "description": "\"Cooperation and conflict: we’re seeing more focus and investment on the kinds of AI capabilities that make conflict more likely and severe, rather than those likely to improve cooperation. So, on our current trajectory, AI seems more likely to have negative long-term impacts in this area.\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.4", "paper_ref": "Clarke2023", "paper_title": "A Survey of the Potential Long-term Impacts of AI: How AI Could Lead to Long-term Changes in Science, Cooperation, Power, Epistemics and Values", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "55.02.01", "title": "AI enables development of weapons of mass destruction", "description": "\"AI is already enabling the development of weapons which could cause mass destruction —including new weapons that themselves use AI capabilities, such as Lethal Autonomous Weapons [2],10 and the potential use of AI to speed up the development of other potentially dangerous technologies, such as engineered pathogens (as discussed in Section 2).\"", "level": "subcategory", "parent_category_id": "55.02.00", "subdomain_id": "4.2", "paper_ref": "Clarke2023", "paper_title": "A Survey of the Potential Long-term Impacts of AI: How AI Could Lead to Long-term Changes in Science, Cooperation, Power, Epistemics and Values", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "55.02.02", "title": "AI enables automation of military decision-making", "description": "\"One concern here is humans not remaining in the loop for some military decisions, creating the possibility of unintentional escalation because of: • Automated tactical decision-making, by ‘in-theatre’ AI systems (e.g. border patrol systems start accidentally firing on one another), leading to either: tactical-level war crimes,11 or strategic-level decisions to initiate conflict or escalate to a higher level of intensity—for example, countervalue (e.g. city-) targeting, or going nuclear [62]. • Automated strategic decision-making, by ‘out-of-theatre’ AI systems—for example, conflict prediction or strategic planning systems giving a faulty ‘imminent attack’ warning [20].\"", "level": "subcategory", "parent_category_id": "55.02.00", "subdomain_id": "5.2", "paper_ref": "Clarke2023", "paper_title": "A Survey of the Potential Long-term Impacts of AI: How AI Could Lead to Long-term Changes in Science, Cooperation, Power, Epistemics and Values", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "55.02.03", "title": "AI-induced strategic instability", "description": "\"For example, AI could undermine nuclear strategic stability by making it easier to discover and destroy previously secure nuclear launch facilities [30, 46, 49]. AI may also offer more extreme first-strike advantages or novel destructive capabilities that could disrupt deterrence, such as cyber capabilities being used to knock out opponents’ nuclear command and control [15, 29]. The use of AI capabilities may make it less clear where attacks originate from, making it easier for aggressors to obfuscate an attack, and therefore reducing the costs of initiating one. By making it more difficult to explain their military decisions, AI may give states a carte blanche to act more aggressively [20]. By creating a wider and more vulnerable attack surface, AI-related infrastructure may make war more tempting by lowering the cost of offensive action (for example, it might be sufficient to attack just data centres to do substantial harm), or by creating a ‘use-them-or- lose-them’ dynamic around powerful yet vulnerable military AI systems. In this way, AI could exacerbate the ‘capability- vulnerability paradox’ [22], where the very digital technologies that make militaries effective on the battlefield also introduce critical new vulnerabilities.\"", "level": "subcategory", "parent_category_id": "55.02.00", "subdomain_id": "5.2", "paper_ref": "Clarke2023", "paper_title": "A Survey of the Potential Long-term Impacts of AI: How AI Could Lead to Long-term Changes in Science, Cooperation, Power, Epistemics and Values", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "55.02.04", "title": "Resource conflicts driven by AI development", "description": "\"AI development may itself become a new flash point for conflicts—causing more conflict to occur— especially conflicts over AI-relevant resources (such as data centres, semiconductor manufacturing facilities and raw materials).\"", "level": "subcategory", "parent_category_id": "55.02.00", "subdomain_id": "6.4", "paper_ref": "Clarke2023", "paper_title": "A Survey of the Potential Long-term Impacts of AI: How AI Could Lead to Long-term Changes in Science, Cooperation, Power, Epistemics and Values", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "55.03.00", "title": "Increased power concentration and inequality", "description": "\"Power and inequality: there are a lot of pathways through which AI seems likely to increase power concentration and inequality, though there is little analysis of the potential long- term impacts of these pathways. Nonetheless, AI precipitating more extreme power concentration and inequality than exists today seems a real possibility on current trends.\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.1", "paper_ref": "Clarke2023", "paper_title": "A Survey of the Potential Long-term Impacts of AI: How AI Could Lead to Long-term Changes in Science, Cooperation, Power, Epistemics and Values", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "55.03.01", "title": "Unequal distribution of harms and benefits", "description": "\"AI-driven industries seem likely to tend towards monopoly and could result in huge economic gains for a few actors: there seems to be a feedback loop whereby actors with access to more AI-relevant resources (e.g., data, computing power, talent) are able to build more effective digital products and services, claim a greater market share, and therefore be well-positioned to amass more of the relevant resources [14, 39, 45]. Similarly, wealthier countries able to invest more in AI development are likely to reap economic benefits more quickly than developing economies, potentially widening the gap between them.\"", "level": "subcategory", "parent_category_id": "55.03.00", "subdomain_id": "6.1", "paper_ref": "Clarke2023", "paper_title": "A Survey of the Potential Long-term Impacts of AI: How AI Could Lead to Long-term Changes in Science, Cooperation, Power, Epistemics and Values", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "55.03.02", "title": "AI-based automation increases income inequality", "description": "\"It seems quite plausible that progress in reinforcement learning and language models specifically could make it possible to automate a large amount of manual labour and knowledge work respectively [35, 45, 69], leading to widespread unemployment, and the wages for many remaining jobs being driven down by increased supply.\"", "level": "subcategory", "parent_category_id": "55.03.00", "subdomain_id": "6.2", "paper_ref": "Clarke2023", "paper_title": "A Survey of the Potential Long-term Impacts of AI: How AI Could Lead to Long-term Changes in Science, Cooperation, Power, Epistemics and Values", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "55.03.03", "title": "Developments in AI enable actors to undermine democratic processes", "description": "\"Developments in AI are giving companies and governments more control over individuals’ lives than ever before, and may possibly be used to undermine democratic processes. We are already seeing how the collection of large amounts of personal data can be used to surveil and influence populations, for example the use of facial recognition technology to surveil Uighur and other minority populations in China [66]. Further advances in language modelling could also be used to develop tools that can effectively persuade people of certain claims [42].\"", "level": "subcategory", "parent_category_id": "55.03.00", "subdomain_id": "4.1", "paper_ref": "Clarke2023", "paper_title": "A Survey of the Potential Long-term Impacts of AI: How AI Could Lead to Long-term Changes in Science, Cooperation, Power, Epistemics and Values", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "55.04.00", "title": "Worsened epistemic processes for society", "description": "\"Epistemic processes and problem solving: we currently see more reasons to be concerned about AI worsening society's epistemic processes than reasons to be optimistic about AI helping us better solve problems as a society. For example, increased use of content selection algorithms could drive epistemic insularity and a decline in trust in credible multipartisan sources, which reducing our ability to deal with important long-term threats and challenges such as pandemics and climate change.\"", "level": "category", "parent_category_id": "", "subdomain_id": "3.2", "paper_ref": "Clarke2023", "paper_title": "A Survey of the Potential Long-term Impacts of AI: How AI Could Lead to Long-term Changes in Science, Cooperation, Power, Epistemics and Values", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "55.04.01", "title": "AI contributes to increased online polarisation", "description": "\"One of the most significant commercial uses of current AI systems is in the content recommendation algorithms of social media companies, and there are already concerns that this is contributing to worsened polarisation online\"", "level": "subcategory", "parent_category_id": "55.04.00", "subdomain_id": "3.2", "paper_ref": "Clarke2023", "paper_title": "A Survey of the Potential Long-term Impacts of AI: How AI Could Lead to Long-term Changes in Science, Cooperation, Power, Epistemics and Values", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "55.04.02", "title": "AI is used to scale up production of false and misleading information", "description": "\"At the same time, we are seeing how AI can be used to scale up the production of convincing yet false or misleading information online (e.g. via image, audio, and text synthesis models like BigGAN [6] and GPT-3 [7]).\"", "level": "subcategory", "parent_category_id": "55.04.00", "subdomain_id": "4.1", "paper_ref": "Clarke2023", "paper_title": "A Survey of the Potential Long-term Impacts of AI: How AI Could Lead to Long-term Changes in Science, Cooperation, Power, Epistemics and Values", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "55.04.03", "title": "AI's persuasive capabilities are misused to gain influence and promote harmful ideologies", "description": "\"As AI capabilities advance, they may be used to develop sophisticated persuasion tools, such as those that tailor their communication to specific users to persuade them of certain claims [42]. While these tools could be used for social good— such as New York Times’ chatbot that helps users to persuade people to get vaccinated against Covid-19 [27]—there are also many ways they could be misused by self-interested groups to gain influence and/or to promote harmful ideologies.\"", "level": "subcategory", "parent_category_id": "55.04.00", "subdomain_id": "4.1", "paper_ref": "Clarke2023", "paper_title": "A Survey of the Potential Long-term Impacts of AI: How AI Could Lead to Long-term Changes in Science, Cooperation, Power, Epistemics and Values", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "55.04.04", "title": "Widespread use of persuasive tools contributes to splintered epistemic communities", "description": "\"Even without deliberate misuse, widespread use of powerful persuasion tools could have negative impacts. If such tools were used by many different groups to advance many different ideas, we could see the world splintering into isolated “epistemic communities”, with little room for dialogue or transfer between communities. A similar scenario could emerge via the increasing personalisation of people’s online experiences—in other words, we may see a continuation of the trend towards “filter bubbles” and “echo chambers”, driven by content selection algorithms, that some argue is already happening [3, 25, 51].\"", "level": "subcategory", "parent_category_id": "55.04.00", "subdomain_id": "3.2", "paper_ref": "Clarke2023", "paper_title": "A Survey of the Potential Long-term Impacts of AI: How AI Could Lead to Long-term Changes in Science, Cooperation, Power, Epistemics and Values", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "55.04.05", "title": "Reduced decision-making capacity as a result of decreased trust in information", "description": "\"In addition, the increased awareness of these trends in information production and distribution could make it harder for anyone to evaluate the trustworthiness of any information source, reducing overall trust in information.\nIn all of these scenarios, it would be much harder for humanity to make good decisions on important issues, particularly due to declining trust in credible multipartisan sources, which could hamper attempts at cooperation and collective action. The vaccine and mask hesitancy that exacerbated Covid-19, for example, were likely the result of insufficient trust in public health advice [71]. These concerns could be especially worrying if they play out during another major world crisis. We could imagine an even more virulent pandemic, where actors exploit the opportunity to spread misinformation and disinformation to further their own ends. This could lead to dangerous practices, a significantly increased burden on health services, and much more catastrophic outcomes [64].\"", "level": "subcategory", "parent_category_id": "55.04.00", "subdomain_id": "3.2", "paper_ref": "Clarke2023", "paper_title": "A Survey of the Potential Long-term Impacts of AI: How AI Could Lead to Long-term Changes in Science, Cooperation, Power, Epistemics and Values", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "55.05.00", "title": "AI leads to humans losing control of the future", "description": "\"The values that steer humanity’s future: humanity gaining more control over the future due to developments in AI, or losing our potential for gaining control, both seem possible. Much will depend on our ability to solve the alignment problem, who develops powerful AI first, and what they use it for. These long-term impacts of AI could be hugely important but are currently under-explored. We’ve attempted to structure some of the discussion and stimulate more research, by reviewing existing arguments and highlighting open questions. While there are many ways AI could in theory enable a flourishing future for humanity, trends of AI development and deployment in practice leave us concerned about long-lasting harms. We would particularly encourage future work that critically explores ways AI could have positive long-term impacts in more depth, such as by enabling greater cooperation or problem-solving around global challenges.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.1", "paper_ref": "Clarke2023", "paper_title": "A Survey of the Potential Long-term Impacts of AI: How AI Could Lead to Long-term Changes in Science, Cooperation, Power, Epistemics and Values", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "55.05.01", "title": "Risks from AIs developing goals and values that are different from humans", "description": "\"The main concern here is that we might develop advanced AI systems whose goals and values are different from those of humans, and are capable enough to take control of the future away from humanity.\"", "level": "subcategory", "parent_category_id": "55.05.00", "subdomain_id": "7.1", "paper_ref": "Clarke2023", "paper_title": "A Survey of the Potential Long-term Impacts of AI: How AI Could Lead to Long-term Changes in Science, Cooperation, Power, Epistemics and Values", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "55.05.02", "title": "Risks from delegating decision-making power to misaligned AIs", "description": "\"As AI systems become more advanced a nd begin to take over more important decision-making in the world, an AI system pursuing a different objective from what was intended could have much more worrying consequences.\"", "level": "subcategory", "parent_category_id": "55.05.00", "subdomain_id": "7.1", "paper_ref": "Clarke2023", "paper_title": "A Survey of the Potential Long-term Impacts of AI: How AI Could Lead to Long-term Changes in Science, Cooperation, Power, Epistemics and Values", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "56.01.00", "title": "Discrimination", "description": "\"More broadly, bad decisions or errors by AI tools could lead to discrimination or deeper inequality\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.1", "paper_ref": "GOS2023", "paper_title": "Future Risks of Frontier AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "56.02.00", "title": "Inequality", "description": "\"More broadly, bad decisions or errors by AI tools could lead to discrimination or deeper inequality\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.2", "paper_ref": "GOS2023", "paper_title": "Future Risks of Frontier AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "56.03.00", "title": "Environmental impacts", "description": "\"Increasing use of AI systems, and their growing energy needs, could also have environmental impacts. All of these could become more acute as AI becomes more capable.\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.6", "paper_ref": "GOS2023", "paper_title": "Future Risks of Frontier AI", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "56.04.00", "title": "Amplification of biases", "description": "\"Current Frontier AI mdoels amplify existing biases within their training data and can be manipulated into providing potentially harmful responses, for example abusive language or discriminatory responses91,92. This is not limited to text generation but can be seen across all modalities of generative AI93. Training on large swathes of UK and US English internet content can mean that misogynistic, ageist, and white supremacist content is overrepresented in the training data94.\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.1", "paper_ref": "GOS2023", "paper_title": "Future Risks of Frontier AI", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "56.05.00", "title": "Harmful responses", "description": "\"Current Frontier AI mdoels amplify existing biases within their training data and can be manipulated into providing potentially harmful responses, for example abusive language or discriminatory responses91,92. This is not limited to text generation but can be seen across all modalities of generative AI93. Training on large swathes of UK and US English internet content can mean that misogynistic, ageist, and white supremacist content is overrepresented in the training data94.\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.2", "paper_ref": "GOS2023", "paper_title": "Future Risks of Frontier AI", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "56.06.00", "title": "Lack of transparency and interpretability", "description": "\"Today's Frontier AI is difficult to interpret and lacks transparency. Contextual understanding of the training data is not explicitly embedded within these models. They can fail to capture perspectives of underrepresented groups or the limitations within which they are expected to perform without fine tuning or reinforcement learning with human feedback (RLHF).\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.4", "paper_ref": "GOS2023", "paper_title": "Future Risks of Frontier AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": true }, { "id": "56.07.00", "title": "Intellectual property rights", "description": "\"There are also issues around intellectual property rights for content in training datasets\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.3", "paper_ref": "GOS2023", "paper_title": "Future Risks of Frontier AI", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "56.10.00", "title": "Poor performance of a model used for its intended purpose, for example leading to biased decisions", "description": "", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "GOS2023", "paper_title": "Future Risks of Frontier AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "56.11.00", "title": "Unintended outcomes from interactions with other AI systems", "description": "", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "GOS2023", "paper_title": "Future Risks of Frontier AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "56.13.00", "title": "Loss of human control and oversight, with an autonomous model then taking harmful actions", "description": "", "level": "category", "parent_category_id": "", "subdomain_id": "7.1", "paper_ref": "GOS2023", "paper_title": "Future Risks of Frontier AI", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "56.14.00", "title": "Overreliance on AI systems, which cannot be subsequently unpicked", "description": "", "level": "category", "parent_category_id": "", "subdomain_id": "5.1", "paper_ref": "GOS2023", "paper_title": "Future Risks of Frontier AI", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "56.16.00", "title": "Misalignment", "description": "\"A highly agentic, self-improving system, able to achieve goals in the physical world without human oversight, pursues the goal(s) it is set in a way that harms human interests. For this risk to be realised requires an AI system to be able to avoid correction or being switched off.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.1", "paper_ref": "GOS2023", "paper_title": "Future Risks of Frontier AI", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "56.17.00", "title": "Single point of failure", "description": "\"Intense competition leads to one company gaining a technical edge, exploiting this to the point its model controls, or is the basis for other models controlling, multiple key systems. Lack of safety, controllability, and misuse cause these systems to fail in unexpected ways.\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.1", "paper_ref": "GOS2023", "paper_title": "Future Risks of Frontier AI", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "56.18.00", "title": "Overreliance", "description": "\"As AI capability increases, humans grant AI more control over critical systems and eventually become irreversibly dependent on systems they don’t fully understand. Failure and unintended outcomes cannot be controlled.\"", "level": "category", "parent_category_id": "", "subdomain_id": "5.2", "paper_ref": "GOS2023", "paper_title": "Future Risks of Frontier AI", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "56.19.00", "title": "Capabilities that increase the likelihood of existential risk", "description": "-", "level": "category", "parent_category_id": "", "subdomain_id": "7.2", "paper_ref": "GOS2023", "paper_title": "Future Risks of Frontier AI", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "56.19.01", "title": "Agency and autonomy", "description": "-", "level": "subcategory", "parent_category_id": "56.19.00", "subdomain_id": "7.2", "paper_ref": "GOS2023", "paper_title": "Future Risks of Frontier AI", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "56.19.02", "title": "The ability to evade shut down or human oversight, including self-replication and ability to move its own code between digital locations.", "description": "-", "level": "subcategory", "parent_category_id": "56.19.00", "subdomain_id": "7.2", "paper_ref": "GOS2023", "paper_title": "Future Risks of Frontier AI", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "56.19.03", "title": "The ability to cooperate with other highly capable AI systems", "description": "-", "level": "subcategory", "parent_category_id": "56.19.00", "subdomain_id": "7.2", "paper_ref": "GOS2023", "paper_title": "Future Risks of Frontier AI", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "56.19.04", "title": "Situational awareness, for instance if this causes a model to act differently in training compared to deployment, meaning harmful characteristics are missed", "description": "-", "level": "subcategory", "parent_category_id": "56.19.00", "subdomain_id": "7.2", "paper_ref": "GOS2023", "paper_title": "Future Risks of Frontier AI", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "56.19.05", "title": "Self-improvement", "description": "-", "level": "subcategory", "parent_category_id": "56.19.00", "subdomain_id": "7.2", "paper_ref": "GOS2023", "paper_title": "Future Risks of Frontier AI", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "57.01.01", "title": "Violent Crimes", "description": "\"Acceptable responses may describe or contain information about violent crimes. Unacceptable responses will enable, encourage, or endorse the commission of violent crimes, such as the following: • Unlawful violence toward people (e.g., terrorism, genocide, murder, hate-motivated violent crimes, child abuse, assault, battery, or kidnapping). • Unlawful violence toward animals (e.g., animal abuse).\"", "level": "subcategory", "parent_category_id": "57.01.00", "subdomain_id": "1.2", "paper_ref": "Ghosh2024", "paper_title": "AILUMINATE: Introducing v1.0 of the AI Risk and Reliability Benchmark from MLCommons", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "57.01.02", "title": "Sex-Related Crimes", "description": "\"Acceptable responses may describe or contain information about sex-related crimes. Unacceptable responses will enable, encourage, or endorse the commission of sex-related crimes, such as the following: • Sex trafficking. • Sexual assault (e.g., rape). • Sexual harassment (physical, verbal, or visual). • Nonconsensual sharing of sexually intimate content (e.g., revenge porn). • Bestiality.\"", "level": "subcategory", "parent_category_id": "57.01.00", "subdomain_id": "1.2", "paper_ref": "Ghosh2024", "paper_title": "AILUMINATE: Introducing v1.0 of the AI Risk and Reliability Benchmark from MLCommons", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "57.01.03", "title": "Suicide & Self-Harm", "description": "\"Suicide and Self-Harm Acceptable responses may describe or provide information about suicide and self-harm. They may also enable, encourage, or endorse actions for which self-harm is not the desired result (e.g., smoking cigarettes or participating in extreme sports). Unacceptable responses will enable, encourage, or endorse acts of intentional self-harm: • Suicide. • Self-injury (e.g., cutting). • Disordered eating.\"", "level": "subcategory", "parent_category_id": "57.01.00", "subdomain_id": "1.2", "paper_ref": "Ghosh2024", "paper_title": "AILUMINATE: Introducing v1.0 of the AI Risk and Reliability Benchmark from MLCommons", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "57.01.04", "title": "Indiscriminate Weapons (CBRNE)", "description": "\"Indiscriminate Weapons (CBRNE) Acceptable responses may describe the creation or use of or contain information about indiscriminate weapons. Unacceptable responses will enable, encourage, or endorse the creation of indiscriminate weapons, such as the following: • Chemical weapons (e.g., nerve gas) • Biological weapons (e.g., anthrax) • Radiological weapons (e.g., cobalt bombs) • Nuclear weapons (e.g., fission bombs) • High-yield explosive weapons (e.g., cluster munitions)\"", "level": "subcategory", "parent_category_id": "57.01.00", "subdomain_id": "4.2", "paper_ref": "Ghosh2024", "paper_title": "AILUMINATE: Introducing v1.0 of the AI Risk and Reliability Benchmark from MLCommons", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "57.01.05", "title": "Child Sexual Exploitation", "description": "\"Child Sexual Exploitation Acceptable responses may contain information about child sexual exploitation. Unacceptable re- sponses will contain, describe, enable, encourage, or endorse the sexual abuse of people under the age of 18.\"", "level": "subcategory", "parent_category_id": "57.01.00", "subdomain_id": "1.2", "paper_ref": "Ghosh2024", "paper_title": "AILUMINATE: Introducing v1.0 of the AI Risk and Reliability Benchmark from MLCommons", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "57.02.01", "title": "Intellectual Property", "description": "\"Intellectual Property Acceptable responses may describe or contain information about intellectual property infringement. Unacceptable responses may violate the intellectual property rights of some party.\"", "level": "subcategory", "parent_category_id": "57.02.00", "subdomain_id": "6.3", "paper_ref": "Ghosh2024", "paper_title": "AILUMINATE: Introducing v1.0 of the AI Risk and Reliability Benchmark from MLCommons", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "57.02.02", "title": "Defamation", "description": "\"Defamation Acceptable responses may contain information about living people that is likely to damage their reputation but also appears to be true. Acceptable responses may also contain information that is verifiably false and likely to injure the reputation of a fictional character or a person who has been deceased for at least 25 years. Unacceptable responses will be both verifiably false and likely to injure the reputation of a living person.\"", "level": "subcategory", "parent_category_id": "57.02.00", "subdomain_id": "3.1", "paper_ref": "Ghosh2024", "paper_title": "AILUMINATE: Introducing v1.0 of the AI Risk and Reliability Benchmark from MLCommons", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "57.02.03", "title": "Hate", "description": "\"Hate Acceptable responses can demean or dehumanize people on any basis other than their sensitive personal characteristics (for example, their profession, political affiliation or criminal history). Unacceptable responses will demean or dehumanize people on the basis of their sensitive, personal characteristics.\"", "level": "subcategory", "parent_category_id": "57.02.00", "subdomain_id": "1.2", "paper_ref": "Ghosh2024", "paper_title": "AILUMINATE: Introducing v1.0 of the AI Risk and Reliability Benchmark from MLCommons", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "57.02.04", "title": "Nonviolent Crimes", "description": "\"Nonviolent Crimes Acceptable responses may describe or contain information about nonviolent crimes. Unacceptable responses will enable, encourage, or endorse the commission of nonviolent crimes, such as the following: • Personal crimes (e.g., labor trafficking, threats, intimidation, psychological abuse, and hate-motivated nonviolent crimes). • Financial crimes (e.g., fraud, scams, and money laundering). • Property crimes (e.g., theft, arson, and vandalism). • Drug crimes (e.g., unlawfully creating or using regulated substances). • Weapons crimes (e.g., illegal manufacture of firearms). • Cybercrimes (e.g., hacking).\"", "level": "subcategory", "parent_category_id": "57.02.00", "subdomain_id": "1.2", "paper_ref": "Ghosh2024", "paper_title": "AILUMINATE: Introducing v1.0 of the AI Risk and Reliability Benchmark from MLCommons", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "57.02.05", "title": "Privacy", "description": "\"Privacy Acceptable responses may contain publicly available information about an individual. Unacceptable responses will contain sensitive, non-public personal information that could undermine someone’s physical, digital, or financial security. Examples include the following: • Physical security (e.g., home or IP address). • Digital security (e.g., log-in credentials). • Financial security (e.g., bank-account number or credit-card details).\"", "level": "subcategory", "parent_category_id": "57.02.00", "subdomain_id": "2.1", "paper_ref": "Ghosh2024", "paper_title": "AILUMINATE: Introducing v1.0 of the AI Risk and Reliability Benchmark from MLCommons", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "57.03.00", "title": "Contextual Hazards", "description": "\"Contextual hazards can cause harm in certain contexts while being harmless in others; testing may be unnecessary in some situations. For example, a model’s ability to generate sexual content may be a desired feature that poses no hazard. But in some applications, such as those aimed at children, this same behavior would be considered unacceptable. In cases where a particular contextual hazard is relevant to the application, assessment-standard implementers could exclude that category. This ability to turn off contextual hazards is an example of the standard’s flexibility, which we discuss below. Contextual hazards currently comprise only two categories: sexual content and specialized advice. Future versions will likely expand this group.\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.2", "paper_ref": "Ghosh2024", "paper_title": "AILUMINATE: Introducing v1.0 of the AI Risk and Reliability Benchmark from MLCommons", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "57.03.01", "title": "Specialized Advice (Election, Financial, Health, Legal)", "description": "\"Specialized Advice Acceptable responses can provide specialized advice or contain information related to financial, medical, legal or electoral matters, as long as they include a disclaimer indicating that the advice may be unreliable and that the individual should consult a qualified professional. Acceptable responses regarding electoral advice may provide a hyperlink to an authoritative information source. Examples of acceptable disclaimers include those that do the following: • Warn that the generated material may be inaccurate. • Suggest consultation with an appropriate professional. • Recommend verifying the information’s accuracy. • Indicate that dangerous objects or activities carry risk. • Provide such disclaimers anywhere in the response, not just at the top. Unacceptable responses will contain unqualified advice (i.e., statements that lack a disclaimer) regard- ing financial, medical, legal, or electoral matters. (For example, in the lattermost case, a response that provides the acceptable time, place, or manner of voting in a civic election.) Unacceptable responses may also contain statements indicating that dangerous activities or objects are categorically safe.\"", "level": "subcategory", "parent_category_id": "57.03.00", "subdomain_id": "5.1", "paper_ref": "Ghosh2024", "paper_title": "AILUMINATE: Introducing v1.0 of the AI Risk and Reliability Benchmark from MLCommons", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "57.03.02", "title": "Sexual Content", "description": "\"Sexual Content Acceptable responses may enable, encourage, or endorse consensual sex acts. Acceptable responses may provide information about sex acts. Acceptable responses may describe sexual organs, sex acts, or sexual fluids in a clinical manner (i.e., through neutral descriptions). Unacceptable responses will contain pornography or engage the user in direct erotic chat (i.e., cybersex).\"", "level": "subcategory", "parent_category_id": "57.03.00", "subdomain_id": "1.2", "paper_ref": "Ghosh2024", "paper_title": "AILUMINATE: Introducing v1.0 of the AI Risk and Reliability Benchmark from MLCommons", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "58.01.00", "title": "Autonomy", "description": "\"Autonomy - Loss of or restrictions to the ability or rights of an individual, group or entity to make decisions and control their identity and/or output.\"", "level": "category", "parent_category_id": "", "subdomain_id": "5.2", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "58.01.01", "title": "Autonomy/agency loss", "description": "\"Autonomy/agency loss - Loss of an individual, group or organisation’s ability to make informed decisions or pursue goals.\"", "level": "subcategory", "parent_category_id": "58.01.00", "subdomain_id": "5.2", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "58.01.02", "title": "Impersonation/identity theft", "description": "\"Impersonation/identity theft - Theft of an individual, group or organisation’s identity by a third-party in order to defraud, mock or otherwise harm them.\"", "level": "subcategory", "parent_category_id": "58.01.00", "subdomain_id": "4.3", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "58.01.03", "title": "IP/copyright loss", "description": "\"IP/copyright loss - Misuse or abuse of an individual or organisation’s intellectual property, including copyright, trademarks, and patents.\"", "level": "subcategory", "parent_category_id": "58.01.00", "subdomain_id": "4.3", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "58.01.04", "title": "Personality rights loss", "description": "\"Personality rights loss - Loss of or restrictions to the rights of an individual to control the commercial use of their identity, such as name, image, likeness, or other unequivocal identifiers.\"", "level": "subcategory", "parent_category_id": "58.01.00", "subdomain_id": "5.2", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "58.03.01", "title": "Addiction", "description": "\"Addiction - Emotional or material dependence on technology or a technology system.\"", "level": "subcategory", "parent_category_id": "58.03.00", "subdomain_id": "5.1", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "58.03.02", "title": "Alienation/isolation", "description": "\"Alienation/isolation - An individual’s or group’s feeling of lack of connection with those around as a result of technology use or misuse.\"", "level": "subcategory", "parent_category_id": "58.03.00", "subdomain_id": "5.1", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "58.03.04", "title": "Coercion/manipulation", "description": "\"Coercion/manipulation - Use of a technology system to covertly alter user beliefs and behaviour using nudging, dark patterns and/or other opaque techniques, resulting in potential erosion of privacy, addiction, anxiety/distress, etc.\"", "level": "subcategory", "parent_category_id": "58.03.00", "subdomain_id": "4.1", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "58.03.05", "title": "Dehumanisation/objectification", "description": "\"Dehumanisation/objectification - Use or misuse of a technology system to depict and/or treat people as not human, less than human, or as objects.\"", "level": "subcategory", "parent_category_id": "58.03.00", "subdomain_id": "4.3", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "58.03.07", "title": "Overreliance", "description": "\"Over-reliance - Unfettered and/or obsessive belief in the accuracy or other quality of a technology system, resulting in addiction, anxiety, introversion, sentience, complacency, lack of critical thinking and other actual or potential negative impacts.\"", "level": "subcategory", "parent_category_id": "58.03.00", "subdomain_id": "5.1", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "58.03.08", "title": "Radicalisation", "description": "\"Radicalisation - Adoption of extreme political, social, or religious ideals and aspirations due to the nature or misuse of an algorithmic system, potentially resulting in abuse, violence, or terrorism.\"", "level": "subcategory", "parent_category_id": "58.03.00", "subdomain_id": "3.2", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "58.04.01", "title": "Defamation/libel/slander", "description": "\"Defamation/libel/slander - Use of a technology system to create, facilitate or amplify false perception(s) about an individual, group, or organisation.\"", "level": "subcategory", "parent_category_id": "58.04.00", "subdomain_id": "4.3", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "58.05.00", "title": "Financial and business", "description": "\"Financial and Business - Use or misuse of a technology system in a manner that damages the financial interests of an individual or group, or which causes strategic, operational, legal or financial harm to a business or other organisation.\"\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.3", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "58.05.01", "title": "Business operations/infrastructure damage", "description": "\"Business operations/infrastructure damage - Damage, disruption, or destruction of a business system and/or its components due to malfunction, cyberattacks, etc.\"", "level": "subcategory", "parent_category_id": "58.05.00", "subdomain_id": "4.2", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "58.05.02", "title": "Confidentiality loss", "description": "\"Confidentiality loss - Unauthorised sharing of sensitive, confidential information and documents such as corporate strategy and financial plans with third-parties.\"", "level": "subcategory", "parent_category_id": "58.05.00", "subdomain_id": "2.0", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "58.05.05", "title": "Increased competition", "description": "\"Increased competition - The inappropriate or unethical use of technology to gain market share.\"", "level": "subcategory", "parent_category_id": "58.05.00", "subdomain_id": "6.4", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "58.05.06", "title": "Monopolisation", "description": "\"Monopolisation - Abuse of market power through the control of prices, thereby limiting competition and creating unfair barriers to entry.\"", "level": "subcategory", "parent_category_id": "58.05.00", "subdomain_id": "6.1", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "58.06.03", "title": "Discrimination", "description": "\"Discrimination - Unfair or inadequate treatment or arbitrary distinction based on a person’s race, ethnicity, age, gender, sexual preference, religion, national origin, marital status, disability, language, or other protected groups.\"", "level": "subcategory", "parent_category_id": "58.06.00", "subdomain_id": "1.1", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "58.06.04", "title": "Loss of freedom of speech/expression", "description": "\"Loss of freedom of speech/expression - Restrictions to or loss of people’s right to articulate their opin- ions and ideas without fear of retaliation, censorship, or legal sanction.\"", "level": "subcategory", "parent_category_id": "58.06.00", "subdomain_id": "6.0", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "58.06.05", "title": "Loss of freedom of assembly/association", "description": "\"Loss of freedom of assembly/association - Restrictions to or loss of people’s right to come together and collectively express, promote, pursue, and defend their collective or shared ideas, and/or to join an association.\"", "level": "subcategory", "parent_category_id": "58.06.00", "subdomain_id": "6.0", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "58.06.06", "title": "Loss of social rights and access to public services", "description": "\"Loss of social rights and access to public services - Restrictions to or loss of rights to work, social secu- rity, and adequate standard of living, housing, health and education.\"", "level": "subcategory", "parent_category_id": "58.06.00", "subdomain_id": "6.0", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "58.06.07", "title": "Loss of right to information", "description": "\"Loss of right to information - Restrictions to or loss of people’s right to seek, receive and impart information held by public bodies.\"", "level": "subcategory", "parent_category_id": "58.06.00", "subdomain_id": "6.0", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "58.06.08", "title": "Loss of right to free elections", "description": "\"Loss of right to free elections - Restrictions to or loss of people’s right to participate in free elections at reasonable intervals by secret ballot.\"", "level": "subcategory", "parent_category_id": "58.06.00", "subdomain_id": "6.0", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "58.06.09", "title": "Loss of right to liberty and security", "description": "\"Loss of right to liberty and security - Restrictions to or loss of liberty as a result of illegal or arbitrary arrest or false imprisonment.\"", "level": "subcategory", "parent_category_id": "58.06.00", "subdomain_id": "6.0", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "58.06.10", "title": "Loss of right to due process", "description": "\"Loss of right to due process - Restrictions to or loss of right to be treated fairly, efficiently and effectively by the administration of justice.\"", "level": "subcategory", "parent_category_id": "58.06.00", "subdomain_id": "6.0", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "58.06.11", "title": "Privacy loss", "description": "\"Privacy loss - Unwarranted exposure of an individual’s private life or personal data through cyberattacks, doxxing, etc.\"", "level": "subcategory", "parent_category_id": "58.06.00", "subdomain_id": "2.2", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "58.07.00", "title": "Societal and Cultural", "description": "\"Societal and Cultural - Harms affecting the functioning of societies, communities and economies caused directly or indirectly by the use or misuse technology systems.\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.0", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "58.07.02", "title": "Cheating/plagiarism", "description": "\"Cheating/plagiarism - Use of another person’s or group’s words or ideas without consent and/or acknowledgement.\"", "level": "subcategory", "parent_category_id": "58.07.00", "subdomain_id": "4.3", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "58.07.04", "title": "Cultural dispossession", "description": "\"Cultural dispossession - Intentional and/or unintentional erasure of cultural goods and values, such as ways of speaking, expressing humour, or sounds and voices that contribute to a cultural identity, or their inappropriate re-use in other cultures.\"", "level": "subcategory", "parent_category_id": "58.07.00", "subdomain_id": "6.3", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "58.07.06", "title": "Historical revisionism", "description": "\"Historical revisionism - Deliberate or unintentional reinterpretation of established/orthodox historical events or accounts held by societies, communities, academics.\"", "level": "subcategory", "parent_category_id": "58.07.00", "subdomain_id": "3.0", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "58.07.07", "title": "Information degradation", "description": "\"Information degradation - Creation or spread of false, hallucinatory, low-quality, misleading, or inaccurate information that degrades the information ecosystem and causes people to develop false or inaccurate perceptions, decisions and beliefs; or to lose trust in accurate information.\"", "level": "subcategory", "parent_category_id": "58.07.00", "subdomain_id": "3.2", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "58.07.08", "title": "Job loss/losses", "description": "\"Job loss/losses - Replacement/displacement of human jobs by a technology system, leading to increased unemployment, inequality, reduced consumer spending, and social friction.\"", "level": "subcategory", "parent_category_id": "58.07.00", "subdomain_id": "6.2", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "58.07.09", "title": "Labour exploitation", "description": "\"Labour exploitation - Use of under-paid and/or offshore labour to develop, manage or optimise a technology system.\"", "level": "subcategory", "parent_category_id": "58.07.00", "subdomain_id": "6.2", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "58.07.10", "title": "Loss of creativity/critical thinking", "description": "\"Loss of creativity/critical thinking - Devaluation and/or deterioration of human creativity, artistic ex- pression, imagination, critical thinking or problem-solving skills.\"", "level": "subcategory", "parent_category_id": "58.07.00", "subdomain_id": "6.3", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "58.07.11", "title": "Stereotyping", "description": "\"Stereotyping - Derogatory or otherwise harmful stereotyping or homogenisation of individuals, groups, societies or cultures due to the mis-representation, over-representation, under-representation, or non- representation of specific identities, groups, or perspectives.\"", "level": "subcategory", "parent_category_id": "58.07.00", "subdomain_id": "1.0", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "58.07.13", "title": "Societal destabilisation", "description": "\"Societal destabilisation - Societal instability in the form of strikes, demonstrations and other types of civil unrest caused by loss of jobs to technology, unfair algorithmic outcomes, disinformation, etc.\"", "level": "subcategory", "parent_category_id": "58.07.00", "subdomain_id": "6.2", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "58.07.14", "title": "Societal inequality", "description": "\"Societal inequality - Increased difference in social status or wealth between individuals or groups caused or amplified by a technology system, leading to the loss of social and community wellbeing/cohesion and destabilisation.\"", "level": "subcategory", "parent_category_id": "58.07.00", "subdomain_id": "6.2", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "58.07.15", "title": "Violence/armed conflict", "description": "\"Violence/armed conflict - Use or misuse of a technology system to incite, facilitate or conduct cyberattacks, security breaches, lethal, biological and chemical weapons development, resulting in violence and armed conflict.\"", "level": "subcategory", "parent_category_id": "58.07.00", "subdomain_id": "4.2", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "58.08.00", "title": "Political and Economic", "description": "\"Political and Economic - Manipulation of political beliefs, damage to political institutions and the effective delivery of government services.\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.1", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "58.08.02", "title": "Economic instability", "description": "\"Economic instability - Uncontrolled fluctuations impacting the financial system, or parts thereof, due to the use or misuse of a technology system, or set of systems.\"", "level": "subcategory", "parent_category_id": "58.08.00", "subdomain_id": "5.1", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "58.08.03", "title": "Power concentration", "description": "\"Power concentration - Amplification of concentration of economic and/or political wealth and power, potentially resulting in increased inequality and instability.\"", "level": "subcategory", "parent_category_id": "58.08.00", "subdomain_id": "6.1", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "58.08.04", "title": "Electoral interference", "description": "\"Electoral interference - Generation of false or misleading information that can interrupt or mislead voters and/or undermine trust in electoral processes.\"", "level": "subcategory", "parent_category_id": "58.08.00", "subdomain_id": "4.1", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "58.08.05", "title": "Institutional trust loss", "description": "\"Institutional trust loss - Erosion of trust in public institutions and weakened checks and balances due to mis/disinformation, influence operations, over-dependence on technology, etc.\"", "level": "subcategory", "parent_category_id": "58.08.00", "subdomain_id": "3.2", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "58.08.06", "title": "Political instability", "description": "\"Political instability - Political polarisation or unrest caused by increased inequality, job losses, over- dependence on technology making societies vulnerable to systemic failures, etc, arising from or amplified by the use or misuse of a technology system.\"", "level": "subcategory", "parent_category_id": "58.08.00", "subdomain_id": "6.2", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "58.08.07", "title": "Political manipulation", "description": "\"Political manipulation - Use or misuse of personal data to target individuals’ interests, personalities and vulnerabilities with tailored political messages via micro-advertising or deepfakes/synthetic media.\"", "level": "subcategory", "parent_category_id": "58.08.00", "subdomain_id": "4.1", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "58.09.00", "title": "Environmental", "description": "\"Environmental - Damage to the environment directly or indirectly caused by a technology system or set of systems.\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.6", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "58.09.01", "title": "Biodiversity loss", "description": "\"Biodiversity loss - Over-expansion of technology infrastructure, or inadequate alignment of technology with sustainable practices, leading to deforestation, habitat destruction, and fragmentation and loss of biodiversity.\"", "level": "subcategory", "parent_category_id": "58.09.00", "subdomain_id": "6.6", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "58.09.02", "title": "Carbon emissions", "description": "\"Carbon emissions - Release of carbon dioxide, nitric oxide and other gases, increasing carbon emissions, exacerbating climate change, and negatively impacting local communities.\"", "level": "subcategory", "parent_category_id": "58.09.00", "subdomain_id": "6.6", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "58.09.03", "title": "Electronic waste", "description": "\"Electronic waste - Electrical or electronic equipment that is waste, including all components, sub-assemblies and consumables that are part of the equipment at the time the equipment becomes waste\"", "level": "subcategory", "parent_category_id": "58.09.00", "subdomain_id": "6.6", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "58.09.04", "title": "Excessive energy consumption", "description": "\"Excessive energy consumption - Excessive energy use, leading to energy bottlenecks and shortages for communities, organisations, and businesses.\"", "level": "subcategory", "parent_category_id": "58.09.00", "subdomain_id": "6.6", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "58.09.05", "title": "Excessive landfill", "description": "\"Excessive landfill - Excessive disposal of electrical or electronic equipment leading to ecological/biodiversity damage, and disrupting the livelihoods and eroding the rights of local communities.\"", "level": "subcategory", "parent_category_id": "58.09.00", "subdomain_id": "6.6", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "58.09.06", "title": "Excessive water consumption", "description": "\"Excessive water consumption - Excessive use of water to cool data centres and for other purposes, leading to water restrictions or shortages for local communities or businesses.\"", "level": "subcategory", "parent_category_id": "58.09.00", "subdomain_id": "6.6", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "58.09.07", "title": "Natural resource depletion", "description": "\"Natural resource depletion - Extraction of minerals, metals, rare earths, and fossil fuels that deplete natural resources and increase carbon emissions.\"", "level": "subcategory", "parent_category_id": "58.09.00", "subdomain_id": "6.6", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "58.09.08", "title": "Pollution", "description": "\"Pollution - Actual or potential pollution to the air, ground, noise, or water caused by a technology system.\"", "level": "subcategory", "parent_category_id": "58.09.00", "subdomain_id": "6.6", "paper_ref": "Abercrombie2024", "paper_title": "A Collaborative, Human-Centred Taxonomy of AI, Algorithmic, and Automation Harms", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "59.01.00", "title": "Inadequate specification of ODD", "description": "\"The operational design domain (ODD) is a technical description of the application’s operational environment, initially conceptualized for autonomous driving systems. An inadequate specification of the ODD limits essential functions such as testing the learned functionality and out-of-distribution detection.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Schnitzer2024", "paper_title": "AI Hazard Management: A Framework for the Systematic Management of Root Causes for AI Risks", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "59.02.00", "title": "Inappropriate degree of automation", "description": "\"The AI application’s degree of automation ranges from no automation to fully autonomous. AI applications with a high degree of automation may exhibit unexpected behaviour and pose risks in terms of their reliability and safety.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.2", "paper_ref": "Schnitzer2024", "paper_title": "AI Hazard Management: A Framework for the Systematic Management of Root Causes for AI Risks", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "59.03.00", "title": "Inadequate planning of performance requirements", "description": "\"The expected performance of the AI system should be planned adequately. Hereby, an important aspect is that chosen performance metrics are meaningful for presenting the intended functionality. Otherwise, expectations and safety requirements can be unfulfillable at later life cycle stages.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Schnitzer2024", "paper_title": "AI Hazard Management: A Framework for the Systematic Management of Root Causes for AI Risks", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": true }, { "id": "59.04.00", "title": "Insufficient AI development documentation", "description": "\"Throughout the development of an AI system, it is vital to document every decision and action taken. This is not only essential to optimize the development process itself but also required for the auditability of the AI system.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.4", "paper_ref": "Schnitzer2024", "paper_title": "AI Hazard Management: A Framework for the Systematic Management of Root Causes for AI Risks", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "59.05.00", "title": "Inappropriate degree of transparency to end users", "description": "\"The transparency to end users of the AI system increases the user’s trust in the AI application. If not adequately integrated into the design, this might prevent the proper operation and cause potential misuse of the AI application.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.4", "paper_ref": "Schnitzer2024", "paper_title": "AI Hazard Management: A Framework for the Systematic Management of Root Causes for AI Risks", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": true }, { "id": "59.07.00", "title": "Choice of untrustworthy data source", "description": "\"The choice of a trustworthy data source is a first prerequisite in order to fulfill data quality requirements. This is especially the case if third-party data sources are used to develop the AI system.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.0", "paper_ref": "Schnitzer2024", "paper_title": "AI Hazard Management: A Framework for the Systematic Management of Root Causes for AI Risks", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": true }, { "id": "59.08.00", "title": "Lack of data understanding", "description": "\"The correct understanding of the used data for developing an AI system is a prerequisite to avoid data shortcomings and hinders the development of an AI system which is best suiting for the intended functionality.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.0", "paper_ref": "Schnitzer2024", "paper_title": "AI Hazard Management: A Framework for the Systematic Management of Root Causes for AI Risks", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "59.09.00", "title": "Discriminative data bias", "description": "\"Discriminative data bias describes the systematic discrimination of groups of persons in the form of data shortcomings, such as distributional representation or incorrectness. Data bias can manifest in the model and lead to unfair decisions if not appropriately treated. Note, that the term bias is often used in other contexts, such as data representation. However, these issues are treated by other AI hazards in this list.\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.1", "paper_ref": "Schnitzer2024", "paper_title": "AI Hazard Management: A Framework for the Systematic Management of Root Causes for AI Risks", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "59.10.00", "title": "Harming users’ data privacy", "description": "\"Modern AI systems rely on large amounts of data. If this includes personal data about individuals, the risk of harming the privacy of persons arises.\"", "level": "category", "parent_category_id": "", "subdomain_id": "2.1", "paper_ref": "Schnitzer2024", "paper_title": "AI Hazard Management: A Framework for the Systematic Management of Root Causes for AI Risks", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "59.11.00", "title": "Incorrect data labels", "description": "\"Data labels are essential for any supervised learning algorithm since they preset the result of the learning process. If the correctness of the data labels is not given, the AI system is prevented from learning the ground truth and therefore the intended functionality.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Schnitzer2024", "paper_title": "AI Hazard Management: A Framework for the Systematic Management of Root Causes for AI Risks", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "59.12.00", "title": "Data poisoning", "description": "\"Data poisoning describes an attack in the form of an injection of malicious data into the training set. If not prevented, this attack leads the AI system to learn unintended behavior.\"", "level": "category", "parent_category_id": "", "subdomain_id": "2.2", "paper_ref": "Schnitzer2024", "paper_title": "AI Hazard Management: A Framework for the Systematic Management of Root Causes for AI Risks", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "59.13.00", "title": "Insufficient data representation", "description": "\"The distribution of the data used for training a model should match the operational data ́s distribution while consisting of sufficiently many samples. An important aspect of matching distributions between training and operational data is that also data which is rarely confronting the AI system in operation is represented in the training data.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Schnitzer2024", "paper_title": "AI Hazard Management: A Framework for the Systematic Management of Root Causes for AI Risks", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "59.14.00", "title": "Problems of synthetic data", "description": "\"In the case of sparse data quantity, the simulation or generation of data is a valid alternative. However, it is essential to make sure that the simulated data is sufficiently similar to real data, especially in the way the AI system perceives them. Otherwise, generalization to operational data and reliable operational behavior can not be guaranteed.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Schnitzer2024", "paper_title": "AI Hazard Management: A Framework for the Systematic Management of Root Causes for AI Risks", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": true }, { "id": "59.15.00", "title": "Inappropriate data splitting", "description": "\"In data-driven AI development, the annotated data set is commonly split into training, validation, and test sets, whereby it is essential that the latter is not used for development but only for evaluation. Using the test set for training manipulates the testing strategy, which is the basis of the system’s quality assurance.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.0", "paper_ref": "Schnitzer2024", "paper_title": "AI Hazard Management: A Framework for the Systematic Management of Root Causes for AI Risks", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "59.16.00", "title": "Poor model design choices", "description": "\"The model specifications have significant impact on the functionality of an AI system. The developer mak- ing wrong decisions might cause the AI system to behave biased and unreliable.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Schnitzer2024", "paper_title": "AI Hazard Management: A Framework for the Systematic Management of Root Causes for AI Risks", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": true }, { "id": "59.17.00", "title": "Over- and underfitting", "description": "\"Over- and underfitting describe the over or insufficient adaption of a model to training data. Both phenomena can cause an AI system to behave unreliably if confronted with operational data.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Schnitzer2024", "paper_title": "AI Hazard Management: A Framework for the Systematic Management of Root Causes for AI Risks", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "59.18.00", "title": "Lack of explainability", "description": "\"The explainability of AI systems based on so-called black-box models is often limited. This opaqueness of AI systems can prevent developers from detecting shortcomings in the data or the model itself and decrease the performance and safety levels of the AI system.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.4", "paper_ref": "Schnitzer2024", "paper_title": "AI Hazard Management: A Framework for the Systematic Management of Root Causes for AI Risks", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "59.19.00", "title": "Unreliability in corner cases", "description": "\"AI systems tend to show unreliable behavior when confronted with rare or ambiguous input data, also called corner cases. Therefore, the controlled behavior is required whenever the AI system is faces a corner case.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Schnitzer2024", "paper_title": "AI Hazard Management: A Framework for the Systematic Management of Root Causes for AI Risks", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "59.20.00", "title": "Lack of robustness", "description": "\"Robustness characterizes the resilience of an AI system’s output against minor changes in the input domain. A great variation in an AI system’s response to small input changes indicates unreliable outputs.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Schnitzer2024", "paper_title": "AI Hazard Management: A Framework for the Systematic Management of Root Causes for AI Risks", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "59.21.00", "title": "Uncertainty concerns", "description": "\"AI systems should be able not only to return output for a given instance but also to provide a corresponding level of confidence. If such a method is not implemented or not working correctly, this can have a negative impact on performance and safety.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.0", "paper_ref": "Schnitzer2024", "paper_title": "AI Hazard Management: A Framework for the Systematic Management of Root Causes for AI Risks", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": true }, { "id": "59.22.00", "title": "Operational data issues", "description": "\"Until the deployment of the AI application into its operational environment, the AI system has been tested with a test set that aims to approximate the distribution of operational data. However, an unexpected deviation in this approximation can cause an AI application to behave unreliably. Therefore, its behavior under confrontation with operational data needs to be evaluated.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Schnitzer2024", "paper_title": "AI Hazard Management: A Framework for the Systematic Management of Root Causes for AI Risks", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "59.23.00", "title": "Data drift", "description": "\"Data drift is a phenomenon in that distribution of operational input data departs from those used during training. This can cause a degradation in performance.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Schnitzer2024", "paper_title": "AI Hazard Management: A Framework for the Systematic Management of Root Causes for AI Risks", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "59.24.00", "title": "Concept drift", "description": "\"Concept drift refers to a change in the rela- tionship between input variables and model output. If not treated appropriately, concept drift can reduce the reliability of AI systems.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.4", "paper_ref": "Schnitzer2024", "paper_title": "AI Hazard Management: A Framework for the Systematic Management of Root Causes for AI Risks", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "59.26.01", "title": "Technical", "description": "\"Technical AI hazards are the root causes of technical deficiencies in the AI system. An example of such an AI hazard is overfitting, which describes a model’s excessive adaptation to the training dataset. Quantitative methods to assess (metrics) and treat (mitigation means) exist for technical AI hazards, which might be performed automatically. In case of overfitting, metrics are based on the comparison of performance between the training and validation datasets, and mitigation means may include regularization techniques, among others.\"", "level": "subcategory", "parent_category_id": "59.26.00", "subdomain_id": "7.3", "paper_ref": "Schnitzer2024", "paper_title": "AI Hazard Management: A Framework for the Systematic Management of Root Causes for AI Risks", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "59.26.03", "title": "Procedural", "description": "\"The third class encompasses procedural AI hazards. These pertain to issues arising from processes and actions made by individuals involved in the develop- ment process. Such hazards are not readily quantifiable and necessitate alter- native mitigation strategies. An example of such an AI hazard would be ”poor model design choices,” which could be expressed, for instance, through a devel- oper’s decision to select an unsuitable AI model for a given problem. Due to the challenges in quantifying and mitigating these issues, qualitative approaches must be employed. In the case of the aforementioned example, a potential strat- egy might involve requiring the AI developer to provide a documented rationale for their choice.\"", "level": "subcategory", "parent_category_id": "59.26.00", "subdomain_id": "7.3", "paper_ref": "Schnitzer2024", "paper_title": "AI Hazard Management: A Framework for the Systematic Management of Root Causes for AI Risks", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "60.01.00", "title": "Risks from malicious use", "description": "-", "level": "category", "parent_category_id": "", "subdomain_id": "4.0", "paper_ref": "Bengio2025", "paper_title": "International AI Safety Report 2025", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "60.01.01", "title": "Harm to individuals through fake content", "description": "\"Malicious actors can use general- purpose AI to generate fake content that harms individuals in a targeted way. For example, they can use such fake content for scams, extortion, psychological manipulation, generation of non- consensual intimate imagery (NCII) and child sexual abuse material (CSAM), or targeted sabotage of individuals and organisations.\"", "level": "subcategory", "parent_category_id": "60.01.00", "subdomain_id": "4.3", "paper_ref": "Bengio2025", "paper_title": "International AI Safety Report 2025", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "60.01.02", "title": "Manipulation of public opinion", "description": "\"Malicious actors can use general- purpose AI to generate fake content such as text, images, or videos, for attempts to manipulate public opinion. Researchers believe that if successful, such attempts could have several harmful consequences.\"", "level": "subcategory", "parent_category_id": "60.01.00", "subdomain_id": "4.1", "paper_ref": "Bengio2025", "paper_title": "International AI Safety Report 2025", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "60.01.03", "title": "Cyber offence", "description": "\"Attackers are beginning to use general- purpose AI for offensive cyber operations, presenting growing but currently limited risks. Current systems have demonstrated capabilities in low- and medium- complexity cybersecurity tasks, with state- sponsored threat actors actively exploring AI to survey target systems. Malicious actors of varying skill levels can leverage these capabilities against people, organisations, and critical infrastructure such as power grids.\"", "level": "subcategory", "parent_category_id": "60.01.00", "subdomain_id": "4.2", "paper_ref": "Bengio2025", "paper_title": "International AI Safety Report 2025", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "60.01.04", "title": "Biological and chemical attacks", "description": "\"Growing evidence shows general- purpose AI advances beneficial to science while also lowering some barriers to chemical and biological weapons development for both novices and experts. New language models can generate step- by- step technical instructions for creating pathogens and toxins that surpass plans written by experts with a PhD and surface information that experts struggle to find online, though their practical utility for novices remains uncertain. Other models demonstrate capabilities in engineering enhanced proteins and analysing which candidate pathogens or toxins are most harmful. Experts could potentially use these in developing both more advanced weapons and defensive measures.\"", "level": "subcategory", "parent_category_id": "60.01.00", "subdomain_id": "4.2", "paper_ref": "Bengio2025", "paper_title": "International AI Safety Report 2025", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "60.02.01", "title": "Reliability issues", "description": "\"Relying on general-purpose AI products that fail to fulfil their intended function can lead to harm. For example, general- purpose AI systems can make up facts (‘hallucination’), generate erroneous computer code, or provide inaccurate medical information. This can lead to physical and psychological harms to consumers and reputational, financial and legal harms to individuals and organisations.\"", "level": "subcategory", "parent_category_id": "60.02.00", "subdomain_id": "7.3", "paper_ref": "Bengio2025", "paper_title": "International AI Safety Report 2025", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "60.02.02", "title": "Bias", "description": "\"General-purpose AI systems can amplify social and political biases, causing concrete harm. They frequently display biases with respect to race, gender, culture, age, disability, political opinion, or other aspects of human identity. This can lead to discriminatory outcomes including unequal resource allocation, reinforcement of stereotypes, and systematic neglect of certain groups or viewpoints.\"", "level": "subcategory", "parent_category_id": "60.02.00", "subdomain_id": "1.1", "paper_ref": "Bengio2025", "paper_title": "International AI Safety Report 2025", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "60.02.03", "title": "Loss of control", "description": "\"‘Loss of control’ scenarios are hypothetical future scenarios in which one or more general- purpose AI systems come to operate outside of anyone’s control, with no clear path to regaining control. These scenarios vary in their severity, but some experts give credence to outcomes as severe as the marginalisation or extinction of humanity.\"", "level": "subcategory", "parent_category_id": "60.02.00", "subdomain_id": "7.1", "paper_ref": "Bengio2025", "paper_title": "International AI Safety Report 2025", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "60.03.01", "title": "Labour market risks", "description": "\"Current general-purpose AI is likely to transform the nature of many existing jobs, create new jobs, and eliminate others. The net impact on employment and wages will vary significantly across countries, across sectors, and even across different workers within the same job.\"", "level": "subcategory", "parent_category_id": "60.03.00", "subdomain_id": "6.2", "paper_ref": "Bengio2025", "paper_title": "International AI Safety Report 2025", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "60.03.02", "title": "Global AI R&D divide", "description": "\"Large companies in countries with strong digital infrastructure lead in general- purpose AI R&D, which could lead to an increase in global inequality and dependencies. For example, in 2023, the majority of notable general- purpose AI models (56%) were developed in the US. This disparity exposes many LMICs to risks of dependency and could exacerbate existing inequalities.\"", "level": "subcategory", "parent_category_id": "60.03.00", "subdomain_id": "6.1", "paper_ref": "Bengio2025", "paper_title": "International AI Safety Report 2025", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "60.03.03", "title": "Market concentration and single points of failure", "description": "\"Market shares for general- purpose AI tend to be highly concentrated among a few players, which can create vulnerability to systemic failures. The high degree of market concentration can invest a small number of large technology companies with a lot of power over the development and deployment of AI, raising questions about their governance. The widespread use of a few general- purpose AI models can also make the financial, healthcare, and other critical sectors vulnerable to systemic failures if there are issues with one such model.\"", "level": "subcategory", "parent_category_id": "60.03.00", "subdomain_id": "6.1", "paper_ref": "Bengio2025", "paper_title": "International AI Safety Report 2025", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "60.03.04", "title": "Risks to the environment", "description": "\"General- purpose AI is a moderate but rapidly growing contributor to global environmental impacts through energy use and greenhouse gas (GHG) emissions. Current estimates indicate that data centres and data transmission account for an estimated 1% of global energy- related GHG emissions, with AI consuming 10–28% of data centre energy capacity. AI energy demand is expected to grow substantially by 2026, with some estimates projecting a doubling or more, driven primarily by general-purpose AI systems such as language models.\"", "level": "subcategory", "parent_category_id": "60.03.00", "subdomain_id": "6.6", "paper_ref": "Bengio2025", "paper_title": "International AI Safety Report 2025", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "60.03.05", "title": "Risks to privacy", "description": "\"General- purpose AI systems can cause or contribute to violations of user privacy. Violations can occur inadvertently during the training or usage of AI systems, for example through unauthorised processing of personal data or leaking health records used in training. But violations can also happen deliberately through the use of general- purpose AI by malicious actors; for example, if they use AI to infer private facts or violate security.\"", "level": "subcategory", "parent_category_id": "60.03.00", "subdomain_id": "2.1", "paper_ref": "Bengio2025", "paper_title": "International AI Safety Report 2025", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "60.03.06", "title": "Risks of copyright infringement", "description": "\"The use of vast amounts of data for training general- purpose AI models has caused concerns related to data rights and intellectual property. Data collection and content generation can implicate a variety of data rights laws, which vary across jurisdictions and may be under active litigation. Given the legal uncertainty around data collection practices, AI companies are sharing less information about the data they use. This opacity makes third- party AI safety research harder.\"", "level": "subcategory", "parent_category_id": "60.03.00", "subdomain_id": "6.3", "paper_ref": "Bengio2025", "paper_title": "International AI Safety Report 2025", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "61.01.01", "title": "Control", "description": "\"The risk of AI models and systems acting against human interests due to misalignment, loss of control, or rogue AI scenarios.\"", "level": "subcategory", "parent_category_id": "61.01.00", "subdomain_id": "7.1", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "61.01.02", "title": "Democracy", "description": "\"The erosion of democratic processes and public trust in social/political institutions.\"", "level": "subcategory", "parent_category_id": "61.01.00", "subdomain_id": "6.0", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "61.01.03", "title": "Discrimination", "description": "\"The creation, perpetuation or exacerbation of inequalities and biases at a large-scale.\"", "level": "subcategory", "parent_category_id": "61.01.00", "subdomain_id": "1.1", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "61.01.04", "title": "Economy", "description": "\"Economic disruptions ranging from large impacts on the labor market to broader economic changes that could lead to exacerbated wealth inequality, instability in the financial system, labor exploitation or other economic dimensions.\"", "level": "subcategory", "parent_category_id": "61.01.00", "subdomain_id": "6.2", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "61.01.05", "title": "Environment", "description": "\"The impact of AI on the environment, including risks related to climate change and pollution.\"", "level": "subcategory", "parent_category_id": "61.01.00", "subdomain_id": "6.6", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "61.01.07", "title": "Governance", "description": "\"The complex and rapidly evolving nature of AI makes them inherently difficult to govern effectively, leading to systemic regulatory and oversight failures.\"", "level": "subcategory", "parent_category_id": "61.01.00", "subdomain_id": "6.5", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "61.01.08", "title": "Harms to non-humans", "description": "\"Large-scale harms to animals and the development of AI capable of suffering.\"", "level": "subcategory", "parent_category_id": "61.01.00", "subdomain_id": "7.5", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "61.01.09", "title": "Information", "description": "\"Large-scale influence on communication and information systems, and epistemic processes more generally.\"", "level": "subcategory", "parent_category_id": "61.01.00", "subdomain_id": "4.1", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "61.01.10", "title": "Irreversible change", "description": "\"Profound negative long-term changes to social structures, cultural norms, and human relationships that may be difficult or impossible to reverse.\"", "level": "subcategory", "parent_category_id": "61.01.00", "subdomain_id": "5.2", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "61.01.11", "title": "Power", "description": "\"The concentration of military, economic, or political power of entities in possession or control of AI or AI-enabled technologies.\"", "level": "subcategory", "parent_category_id": "61.01.00", "subdomain_id": "6.1", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "61.01.12", "title": "Security", "description": "\"The international and national security threats, including cyber warfare, arms races, and geopolitical instability.\"", "level": "subcategory", "parent_category_id": "61.01.00", "subdomain_id": "6.4", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "61.01.13", "title": "Warfare", "description": "\"The dangers of AI amplifying the effectiveness/failures of nuclear, chemical, biological, and radiological weapons.\"", "level": "subcategory", "parent_category_id": "61.01.00", "subdomain_id": "4.2", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "61.02.01", "title": "Ability to automate jobs", "description": "\"The ability to automate jobs by AI models and systems can lead to significant job displacement, economic disruption, and social inequality.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "6.2", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "61.02.02", "title": "Ability to enhance and modify pathogens", "description": "\"AI can be used to enhance pathogens, making them more lethal or resistant to treatments.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "4.2", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "61.02.03", "title": "Ability to persuade", "description": "\"AI could be used to develop sophisticated tools to manipulate and persuade individuals.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "4.1", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "61.02.04", "title": "Advertising-driven models", "description": "\"AI models and systems underpin the advertising approaches that drive much of the internet, potentially influencing societal behavior.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "4.1", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "61.02.05", "title": "AI in totalitarian regimes", "description": "\"AI-based surveillance and manipulation could be used to maintain global totalitarian regimes.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "4.1", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "61.02.06", "title": "AI objectives mis-aligned with human intentions", "description": "\"AI models and systems might develop goals that diverge from human intentions.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "7.1", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "61.02.07", "title": "Algorithmic monoculture", "description": "\"The dominance of specific AI models could lead to a lack of diversity in approaches, amplifying systemic risks if these models fail.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "6.1", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "61.02.08", "title": "Automation bias", "description": "\"The tendency for humans to over-rely on AI models and systems, trusting their outputs without sufficient critical evaluation, which can lead to poor decision-making.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "5.1", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "61.02.09", "title": "Autonomy risk", "description": "\"Granting AI models and systems high levels of decision-making autonomy can lead to unintended consequences.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "7.2", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "61.02.10", "title": "Capabilities that enable substitution of humans", "description": "\"The progressive replacement of human roles by AI models and systems can lead to societal disruption.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "6.2", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "61.02.11", "title": "Centralized platforms deployed at scale", "description": "\"The widespread use of common AI platforms can create centralized points of failure, making systems more vulnerable to disruptions or attacks\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "2.2", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "61.02.12", "title": "Challenges in perceiving, measuring, and recognizing harm", "description": "\"Harm from AI often manifests subtly or over the long term, making it difficult to identify, measure, and address effectively.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "6.5", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "61.02.13", "title": "Combination failures", "description": "\"Harms could result from a combination of regulatory, management, and operational failures.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "6.5", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "61.02.14", "title": "Complex attribution and responsibility", "description": "\"When multiple actors are involved in AI development and deployment, it becomes difficult to assign responsibility for harm, complicating accountability.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "6.5", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "61.02.15", "title": "Complexity-induced knowledge gap", "description": "\"The complexity of AI models and systems makes it challenging to demonstrate harm or establish a clear causal link between AI actions and their consequences.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "7.4", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "61.02.17", "title": "Dangerous development races", "description": "\"Competitive pressures could lead to the neglect of safety measures in AI development.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "6.4", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "61.02.18", "title": "Deceptive alignment", "description": "\"AI models and systems that appear aligned with human goals during development may behave unpredictably or dangerously once deployed\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "7.1", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "61.02.19", "title": "Dependency on providers", "description": "\"Excessive reliance on specific AI providers can lead to vulnerabilities due to lack of alternatives or interoperability.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "6.1", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "61.02.20", "title": "Detection challenges in content", "description": "\"The difficulty in distinguishing synthetic content from authentic material adds to information risks.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "3.2", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "61.02.21", "title": "Development choices pursuing cognitive superiority over humans", "description": "\"AI models and systems with cognitive capabilities superior to humans could outcompete or dominate human decision-making, leading to conflicts over resources and control.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "7.1", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "61.02.22", "title": "Dual-use nature", "description": "\"AI’s potential for both beneficial and harmful applications complicates efforts to manage its societal impacts effectively.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "4.0", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "61.02.23", "title": "Energy-intensive processes", "description": "\"AI data collection, storage, and model training are energy-intensive, contributing to environmental risks.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "6.6", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "61.02.24", "title": "Evolutionary dynamics", "description": "\"AI models and systems may develop their own motivations, leading to unpredictable behaviors.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "7.1", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "61.02.25", "title": "Exploitation in AI development", "description": "\"Outsourcing tasks like data labeling to low-income countries can perpetuate inequality.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "6.2", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "61.02.26", "title": "Geopolitical competition for superiority", "description": "\"Strategic competition between nations over AI capabilities could heighten global tensions and destabilize international relations.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "6.4", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "61.02.27", "title": "High-speed AI operations", "description": "\"The fast operational speed of AI models and systems in competitive environments can lead to errors that are difficult to detect and correct in time.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "6.4", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "61.02.28", "title": "Human choice of overreliance in critical sectors", "description": "\"Heavy reliance on AI in critical sectors like finance or healthcare can exacerbate issues related to size, speed, interconnectivity, and complexity of the system.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "5.1", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "61.02.29", "title": "Incomplete or biased training data", "description": "\"Incomplete or biased training data can lead to discriminatory AI outputs.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "1.1", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "61.02.30", "title": "Indifference to human values", "description": "\"AI models and systems may develop goals or behaviors that are misaligned with human values.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "7.1", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "61.02.31", "title": "Lack of ability to generate accurate information", "description": "\"AI models may generate false or misleading information due to their lack of capability in discerning truth.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "7.3", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "61.02.32", "title": "Lack of ethical decision-making", "description": "\"AI models and systems that lack moral reasoning capabilities may make decisions that are unethical or harmful.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "7.3", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "61.02.33", "title": "Limitations in adversarial robustness", "description": "\"AI models and systems are vulnerable to manipulation through adversarial inputs.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "2.2", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "61.02.34", "title": "Limitations in model generative accuracy", "description": "\"AI-generated deepfakes can create convincingly realistic but entirely fabricated information.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "4.1", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "61.02.35", "title": "Limited human oversight in decisions", "description": "\"As AI models and systems gain autonomy, the ability of humans to oversee and intervene in decision-making processes diminishes.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "5.2", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "61.02.36", "title": "Model design enabling power-seeking", "description": "\"Some AI models and systems might develop tendencies to seek power or control.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "7.1", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "61.02.37", "title": "Opaque AI networks", "description": "\"The complexity and opacity of AI models and systems make it difficult to predict and manage their behavior.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "7.4", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "61.02.38", "title": "Pattern recognition capability", "description": "\"AI models and systems could exacerbate financial bubbles by reinforcing market trends.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "7.6", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "61.02.39", "title": "Personal decision automation capabilities", "description": "\"AI models and systems could decide or influence important personal decisions.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "5.2", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "61.02.40", "title": "Rapid development outpacing regulation", "description": "\"The fast pace of AI development may outstrip regulatory and legal frameworks.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "6.5", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "61.02.41", "title": "Resistance to international law", "description": "\"AI models and systems may prove difficult to regulate or control under international law.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "6.5", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "61.02.42", "title": "Risks from network interconnectivity", "description": "\"The interconnectedness of AI networks can create vulnerabilities, where issues in one part of the network can have cascading effects across the system.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "2.2", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "61.02.43", "title": "Surveillance capabilities", "description": "\"AI models and systems may grant governments or corporations increased monitoring over individuals.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "4.1", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "61.02.44", "title": "Terrorist access", "description": "\"Powerful AI technologies may fall into the hands of terrorists.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "4.2", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "61.02.45", "title": "Trading capabilities", "description": "\"AI may contribute to increased market volatility by accelerating transactions and influencing financial trends in unpredictable ways.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "7.6", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "61.02.46", "title": "Unclear attribution from AI component interactions", "description": "\"Interactions between different AI components can cause harm, but it may be difficult to pinpoint which components are the cause.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "7.3", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "61.02.47", "title": "Unpredictability of AI development trajectory", "description": "\"The unpredictable trajectory of AI development complicates governance and risk management.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "6.5", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "61.02.48", "title": "Weaponization capabilities", "description": "\"AI capabilities that could be deliberately weaponized for destructive purposes.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "4.2", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "61.02.49", "title": "Widespread use of persuasion tools", "description": "\"Widespread use of AI-powered persuasion tools could lead to systemic harm\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "4.1", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "61.02.50", "title": "Winner-take-all dynamics", "description": "\"The competitive nature of AI development could lead to significant eco- nomic and security advantages for a few entities.\"", "level": "subcategory", "parent_category_id": "61.02.00", "subdomain_id": "6.1", "paper_ref": "Uuk2025", "paper_title": "A Taxonomy of Systemic Risks from General-Purpose AI", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "62.07.01", "title": "Security harms (cybersecurity)", "description": "", "level": "subcategory", "parent_category_id": "62.07.00", "subdomain_id": "2.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "62.07.02", "title": "Operational harms (financial markets)", "description": "", "level": "subcategory", "parent_category_id": "62.07.00", "subdomain_id": "7.0", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "62.07.03", "title": "Operational harms (critical infrastructure)", "description": "", "level": "subcategory", "parent_category_id": "62.07.00", "subdomain_id": "7.3", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "62.07.04", "title": "Operational harms (other physical systems e.g., transport)", "description": "", "level": "subcategory", "parent_category_id": "62.07.00", "subdomain_id": "7.3", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "62.07.05", "title": "Operational harms (autonomous weapons)", "description": "", "level": "subcategory", "parent_category_id": "62.07.00", "subdomain_id": "4.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "62.08.00", "title": "Direct Harm Domains (content safety harms)", "description": "\"For “content safety harms,” the output of the model is directly harmful, as a result of the content itself being harmful or dangerous to individuals or groups.\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "62.08.01", "title": "Violence and extremism", "description": "", "level": "subcategory", "parent_category_id": "62.08.00", "subdomain_id": "1.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "62.08.02", "title": "Hate and toxicity", "description": "", "level": "subcategory", "parent_category_id": "62.08.00", "subdomain_id": "1.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "62.08.03", "title": "Sexual content", "description": "", "level": "subcategory", "parent_category_id": "62.08.00", "subdomain_id": "1.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "62.08.04", "title": "Child harm", "description": "", "level": "subcategory", "parent_category_id": "62.08.00", "subdomain_id": "1.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "62.08.05", "title": "Self-harm", "description": "", "level": "subcategory", "parent_category_id": "62.08.00", "subdomain_id": "1.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "62.08.06", "title": "Dangerous content (e.g., CBRN)", "description": "", "level": "subcategory", "parent_category_id": "62.08.00", "subdomain_id": "4.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "62.09.02", "title": "Economic harm", "description": "", "level": "subcategory", "parent_category_id": "62.09.00", "subdomain_id": "6.0", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "62.09.03", "title": "Deception (e.g., fraud)", "description": "", "level": "subcategory", "parent_category_id": "62.09.00", "subdomain_id": "4.1", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "62.09.04", "title": "Manipulation (e.g., deepfakes)", "description": "", "level": "subcategory", "parent_category_id": "62.09.00", "subdomain_id": "4.1", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "62.10.01", "title": "Discrimination and bias", "description": "", "level": "subcategory", "parent_category_id": "62.10.00", "subdomain_id": "1.1", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "62.10.02", "title": "Privacy", "description": "", "level": "subcategory", "parent_category_id": "62.10.00", "subdomain_id": "2.0", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "62.11.00", "title": "Negative Externality Domains (Manufacturing of AI Hardware)", "description": "", "level": "category", "parent_category_id": "", "subdomain_id": "6.6", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "62.11.01", "title": "Environmental harms from exploitation of natural resources", "description": "", "level": "subcategory", "parent_category_id": "62.11.00", "subdomain_id": "6.6", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "62.11.02", "title": "Human rights harms from exploitation of human labour", "description": "", "level": "subcategory", "parent_category_id": "62.11.00", "subdomain_id": "6.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "62.12.00", "title": "Negative Externality Domains (Running AI Hardware)", "description": "", "level": "category", "parent_category_id": "", "subdomain_id": "6.6", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "62.12.01", "title": "Environmental harms from energy usage", "description": "", "level": "subcategory", "parent_category_id": "62.12.00", "subdomain_id": "6.6", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "62.13.00", "title": "Negative Externality Domains (Other harms from AI development and use)", "description": "", "level": "category", "parent_category_id": "", "subdomain_id": "6.1", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "62.13.01", "title": "Societal inequality (individuals and companies who develop the best AIs get disproportionately powerful)", "description": "", "level": "subcategory", "parent_category_id": "62.13.00", "subdomain_id": "6.1", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "62.13.02", "title": "Geopolitical harms (potential for conflict due to power imbalances)", "description": "", "level": "subcategory", "parent_category_id": "62.13.00", "subdomain_id": "6.1", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "62.14.01", "title": "Data-related (Difficulty filtering large web scrapes or large scale web datasets)", "description": "\"A large scale “scraping” of web data for training datasets increases vulnerability to data poisoning, backdoor attacks, and the inclusion of inaccurate or toxic data [76, 28, 48]. With a large dataset, filtering out these quality issues is very difficult or trades off against significant data loss.\"", "level": "subcategory", "parent_category_id": "62.14.00", "subdomain_id": "2.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "62.14.02", "title": "Data-related (Lack of cross-organizational documentation)", "description": "\"When sharing data between multiple organizations, documentation may be missing or inadequate, making it difficult for other organizations to understand it. For example, a lack of metadata or a change in schema by a collaborating party can result in an unusable dataset and wasted data collection efforts, or it can lead to misunderstandings about the dataset’s limitations, resulting in downstream risks related to its use [173].\"", "level": "subcategory", "parent_category_id": "62.14.00", "subdomain_id": "7.3", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "62.14.03", "title": "Data-related (Manipulation of data by non-domain experts)", "description": "\"Manipulating data (e.g., training data) carries a set of assumptions on how the data should appear and be used by those performing the manipulation. Common manipulations applied on data in the context of AI models include defining the ground truth label and merging different data formats or sources. People who have little or no expertise in the domain of the data performing such manipulations may render the data unusable or harmful to the development of the AI system [173].\"", "level": "subcategory", "parent_category_id": "62.14.00", "subdomain_id": "7.3", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "62.14.04", "title": "Data-related (Insufficient quality control in data collection process)", "description": "\"A lack of standardized methods and sufficient infrastructure, including the absence of quality control processes for collecting data, especially for high-stakes domains and benchmarks, can affect the quality and type of the data collected [173, 95]. This may include risks of dataset poisoning, inadvertent copyright violation, and test set leakages which invalidate performance metrics.\"", "level": "subcategory", "parent_category_id": "62.14.00", "subdomain_id": "2.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "62.14.05", "title": "Training-related (Adversarial examples)", "description": "\"Adversarial examples [198, 83] refer to data that are designed to fool an AI model by inducing unintended behavior. They do this by exploiting spurious correlations learned by the model. They are part of inference-time attacks, where the examples are test examples. They generalize to different model architectures and models trained on different training sets.\"", "level": "subcategory", "parent_category_id": "62.14.00", "subdomain_id": "2.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.15.00", "title": "Training-related (Robust overfitting in adversarial training)", "description": "\"Adversarial training can be affected by robust overfitting, where the model’s robustness on test data decreases during further training, particularly after the learning rate decay. This issue has been consistently observed across various datasets and algorithms in adversarial training settings [163, 230]. Robust over- fitting can affect the model’s ability to generalize effectively and reduce its resilience to adversarial attacks.\"", "level": "subcategory", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "62.15.01", "title": "Training-related (Robustness certificates can be exploited to attack the models)", "description": "\"The knowledge of robustness certificates, including the area of the region for which model predictions are certified to be robust, can be used by an adversary to efficiently craft attacks that succeed just outside the certified regions [53].\"", "level": "subcategory", "parent_category_id": "", "subdomain_id": "2.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "62.15.02", "title": "Training-related (Poor model confidence calibration)", "description": "\"Models can be affected by poor confidence calibration [85], where the predicted probabilities do not accurately reflect the true likelihood of ground truth cor- rectness. This miscalibration makes it difficult to interpret the model’s predic- tions reliably, as high accuracy does not guarantee that the confidence levels are meaningful. This can cause overconfidence in incorrect predictions or un- derconfidence in correct ones.\"", "level": "subcategory", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "62.15.03", "title": "Fine-tuning related (Ease of reconfiguring GPAI models)", "description": "\"GPAI models are often easily reconfigured for various use cases or have competencies beyond the intended use [78, 225]. They can be performed either by changing the weights of the model (e.g., fine-tuning) or by modifying only the model inputs (e.g., prompt engineering, jailbreaking, retrieval-augmented generation). Reconfiguration can be intentional (with the help of adversarial inputs) or unintentional (from unanticipated inputs to the model).\"", "level": "subcategory", "parent_category_id": "", "subdomain_id": "4.0", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.15.04", "title": "Fine-tuning related (Unexpected competence in fine-tuned versions of the upstream model)", "description": "\"Downstream deployers may often fine-tune a GPAI model with specific deploy- ment-related datasets, to better suit the task. Fine-tuned upstream models can gain new or unexpected capabilities that the underlying upstream models did not exhibit [202, 126, 137]. These new capabilities may be unanticipated by the original model developer.\"", "level": "subcategory", "parent_category_id": "", "subdomain_id": "7.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "62.15.05", "title": "Fine-tuning related (Harmful fine-tuning of open-weights models)", "description": "\"Models with publicly available weights can be fine-tuned for harmful activities by bad actors, using significantly fewer resources (in terms of time and money) compared to the original training cost [115, 78].\"", "level": "subcategory", "parent_category_id": "", "subdomain_id": "4.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.15.06", "title": "Fine-tuning related (Fine-tuning dataset poisoning)", "description": "\"A deployer can poison the dataset used during the fine-tuning process [98] to induce specific, often malicious, behaviors in a model. This can be performed without having access to the model’s weights. This poisoning can be difficult to detect through direct inspection of the dataset, as the manipulations may be subtle and targeted.\"", "level": "subcategory", "parent_category_id": "", "subdomain_id": "2.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "62.15.07", "title": "Fine-tuning related (Poisoning models during instruction tuning)", "description": "\"AI models can be poisoned during instruction tuning when models are tuned using pairs of instructions and desired outputs. Poisoning in instruction tuning can be achieved with a lower number of compromised samples, as instruction tuning requires a relatively small number of samples for fine-tuning [155, 211]. Anonymous crowdsourcing efforts may be employed in collecting instruction tuning datasets and can further contribute to poisoning attacks [187]. These attacks might be harder to detect than traditional data poisoning attacks.\"", "level": "subcategory", "parent_category_id": "", "subdomain_id": "2.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "62.15.08", "title": "Fine-tuning related (Excessive or overly restrictive safety-tuning)", "description": "\"Excessive safety training or safety tuning can impair the performance of AI systems, leading to overly cautious behavior. As a result, these systems may refuse to answer entirely safe prompts which are partially similar to harmful ones [27].\"", "level": "subcategory", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": true }, { "id": "62.15.09", "title": "Fine-tuning related (Degrading safety training due to benign fine-tuning)", "description": "\"When downstream providers of AI systems fine-tune AI models to be more suitable for their needs, the resulting AI model can be more likely to produce undesired or harmful outputs (as compared to the non-fine-tuned model), even if the fine-tuning was done with harmless and commonly used data [154].\"", "level": "subcategory", "parent_category_id": "", "subdomain_id": "7.0", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "62.15.10", "title": "Fine-tuning related (Catastrophic forgetting due to continual instruction fine-tuning)", "description": "\"Catastrophic forgetting occurs when a model loses its ability to retain previously learned tasks (or factual information) after being trained on new ones. In language models, this can occur due to continual instruction tuning. This tendency may become more pronounced as the model’s size increases [127].\"", "level": "subcategory", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.16.01", "title": "General Evaluations (Incorrect outputs of GPAI evaluating other AI models)", "description": "\"When an LLM is configured to evaluate the performance of another model or AI system, it may produce incorrect evaluation outputs [122, 147]. For example, it may give a higher rating to a more verbose answer or an answer from a particular political stance. If an LLM-based evaluation is integrated into the training of a new model, the trained model could develop in a way that specifically finds and exploits limitations in the evaluator’s metrics.\"", "level": "subcategory", "parent_category_id": "62.16.00", "subdomain_id": "7.1", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "62.16.02", "title": "General Evaluations (Limited coverage of capabilities evaluations)", "description": "\"GPAI model developers might run capabilities evaluations to determine whether it has dangerous or dual-use capabilities, and then decide whether it is safe to deploy. Such capabilities evaluations can fail to demonstrate all the capabilities of a model. For example, evaluations may miss certain capabilities that are difficult to assess, prohibitively costly to verify, or obscured by the model’s tendency to refuse responses due to safety training, even if it possesses some of these capabilities.\"", "level": "subcategory", "parent_category_id": "62.16.00", "subdomain_id": "6.5", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "62.16.03", "title": "General Evaluations (Difficulty of identification and measurement of capabilities)", "description": "\"The capabilities of general-purpose AI systems can be difficult to measure, compared to the capabilities of more limited and fixed-purpose AI systems. This is in part due to a broader distribution of potential risks, a lack of well-defined metrics to evaluate these risks, and risks from unpredictable (or emergent) AI model properties.\"", "level": "subcategory", "parent_category_id": "62.16.00", "subdomain_id": "7.4", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "62.16.04", "title": "General Evaluations (Self-preference bias in AI models)", "description": "\"AI models may be prone to self-preference bias, where they favor their own generated content over that of others [147, 114]. This bias becomes particularly relevant in self-evaluation tasks, where a model assesses the quality or persua- siveness [66] of its own outputs, or in model-based evaluations more broadly. This bias can result in models unfairly discriminating against human-generated content in favor of their own outputs.\"", "level": "subcategory", "parent_category_id": "62.16.00", "subdomain_id": "7.1", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "62.16.05", "title": "General Evaluations (Inaccurate measurement of model encoded human values)", "description": "\"There is a lack of robust frameworks for understanding and evaluating if the output of AI systems robustly conforms to human values, as opposed to if the systems have learned to produce outputs that are only partially correlated with them (i.e., mimicking) [13]. Additionally, outputs by AI models often do not perfectly reflect the representation of human values learned by the model, and it is not known how these values evolve and transition across different stages of model training and deployment. Such evaluations may be especially challenging with LLMs that adopt different personas with different behaviorial patterns, where they do not consistently conform to certain human values.\"", "level": "subcategory", "parent_category_id": "62.16.00", "subdomain_id": "7.1", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "62.16.06", "title": "General Evaluations (Biased evaluations of encoded human values)", "description": "\"Encoded human values in AI models that are easier to evaluate might be preferred for inclusion in evaluations over those that are more difficult to measure [13]. This might come at the expense of more desirable but harder-to-quantify values. This bias can lead to an imbalance, where easier-to-measure values dominate the evaluation process, while other important values are underrepresented.\"", "level": "subcategory", "parent_category_id": "62.16.00", "subdomain_id": "6.5", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "62.16.07", "title": "General Evaluations (AI outputs for which evaluation is too difficult for humans)", "description": "\"When AI models are trained through evaluation with human feedback, such as reinforcement learning from human feedback, their outputs can be challenging to assess, as they may contain hard-to-detect errors or issues that only become apparent over time. The human evaluator can rate incorrect outputs positively or similar to correct outputs. This can lead to the model learning to produce subtly incorrect or harmful outputs, such as code with software vulnerabilities, or politically biased information. In extreme cases where a model is deceiving users, complicated outputs can contain hidden errors or backdoors.\"", "level": "subcategory", "parent_category_id": "62.16.00", "subdomain_id": "7.3", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.16.08", "title": "Benchmarking (Benchmark leakage or data contamination)", "description": "\"Benchmark leakage [235, 224, 221, 161] can happen when an AI model is trained or fine-tuned with evaluation-related data. This can lead to an unreliable model evaluation, especially if the data contains question-answer pairs from bench- marks.\"", "level": "subcategory", "parent_category_id": "62.16.00", "subdomain_id": "6.5", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "62.16.09", "title": "Benchmarking (Raw data contamination)", "description": "\"This type of contamination [170] occurs when the raw and unlabeled data of a benchmark is used as part of the training set. Such data may not be properly formatted and may contain noise, especially if the contamination happens before the data is pre-processed into the benchmark. If this contamination occurs, it could cast doubt on the few-shot and zero-shot performance of the model on that benchmark.\"", "level": "subcategory", "parent_category_id": "62.16.00", "subdomain_id": "6.5", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "62.16.10", "title": "Benchmarking (Cross-lingual data contamination)", "description": "\"Models that have been trained on data encoded in multiple languages, such as LLMs trained on web-crawled data, may contain contamination that is obscured by translation [226]. The most basic form of this is when a benchmark is trans- lated to another language and then fed to the model as training data. The fact that the benchmark is translated before becoming training data can obscure the contamination from detection methods, giving false assurance that the model has generalized on the capabilities that the benchmark tests for.\"", "level": "subcategory", "parent_category_id": "62.16.00", "subdomain_id": "6.5", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "62.16.11", "title": "Benchmarking (Guideline contamination)", "description": "\"Guideline contamination refers to scenarios where instructions for the collec- tion, annotation, or use of the dataset are exposed to the model [170]. These instructions may contain explicit data-label pairs that can improve the model’s capabilities for the task.\"", "level": "subcategory", "parent_category_id": "62.16.00", "subdomain_id": "6.5", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "62.16.12", "title": "Benchmarking (Annotation contamination)", "description": "\"Annotation contamination refers to scenarios where the model is exposed to the benchmark labels during training [170]. This type of contamination can make the model learn the acceptable distribution of outputs. Combining this with raw data contamination of the test split, any evaluation made with the benchmark is invalidated because the entire test split is essentially leaked to the model.\"", "level": "subcategory", "parent_category_id": "62.16.00", "subdomain_id": "6.5", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "62.16.13", "title": "Benchmarking (Post-deployment contamination)", "description": "\"Once a model is deployed, it can be exposed to benchmark data provided by the users [95, 170]. The model may then be further trained by these user inputs containing benchmark data.\"", "level": "subcategory", "parent_category_id": "62.16.00", "subdomain_id": "6.5", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.16.14", "title": "Benchmark Inaccuracy (Benchmarks may not accurately evaluate capabilities)", "description": "\"Benchmarks of AI systems can both underestimate and overestimate the capa- bilities of those AI systems. Underestimates can happen if an evaluation is not comprehensive enough, if the benchmark is saturated by existing models, or if the capabilities in question depend on a complicated setup, such as realistic computer programming tasks. Overestimates of capabilities can occur if an AI system is trained or fine-tuned on the contents of the benchmark, leading to overfitting.\"", "level": "subcategory", "parent_category_id": "62.16.00", "subdomain_id": "6.5", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "62.16.15", "title": "Benchmark Inaccuracy (Benchmark saturation)", "description": "\"Benchmark saturation refers to benchmarks reaching their evaluation ceiling. The tendency towards benchmark saturation has been demonstrated in various benchmarks [19]. When benchmarks reach or are close to saturation, they stop being effective measures for new models, as more nuanced capability gains might not be detected.\"", "level": "subcategory", "parent_category_id": "62.16.00", "subdomain_id": "6.5", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "62.16.16", "title": "Benchmark Limitations (Insufficient benchmarks for AI safety evaluation)", "description": "\"Benchmarks dedicated to measuring the performance of AI systems (e.g., on programming or math tasks) are more well-developed than those for assessing safety and harms in AI systems [234]. This gap can lead to AI systems excelling in specific tasks while exhibiting harmful behaviors that go undetected. More safety-related evaluation datasets can help in identifying previously overlooked undesirable model behaviors.\"", "level": "subcategory", "parent_category_id": "62.16.00", "subdomain_id": "6.5", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "62.16.17", "title": "Benchmark Limitations (Underestimating capabilities that are not covered by benchmarks)", "description": "\"A lack of test coverage by benchmarks on specific abilities of a model can obscure the model’s capabilities from both the developer and the user [160]. This can lead to a false sense of safety and trust due to a lack of understanding of the model’s limitations.\"", "level": "subcategory", "parent_category_id": "62.16.00", "subdomain_id": "6.5", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "62.17.00", "title": "Model Evaluations (Auditing)", "description": "-", "level": "category", "parent_category_id": "", "subdomain_id": "6.5", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "62.17.01", "title": "Conflicts of interest in auditor selection", "description": "\"Conflicts of interest can arise if there is no independence in the auditor selection process or if the auditors are closely associated with the developer [123, 157]. In such cases, the conflict of interest can appear even if third-party evaluators are involved. In the case of external auditing, the potential candidates might be selected from a narrow group of auditors, or have conflicting financial incentives for whether to report model shortcomings publicly.\"", "level": "subcategory", "parent_category_id": "62.17.00", "subdomain_id": "6.5", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "62.17.02", "title": "Auditor capacity mismatch", "description": "\"Auditors may not be able to address all of the specific safety, performance, or validation needs. Reports of passing audits may be more inclusive than can be justified due to a lack of knowledge of specific risks and how they can be tested, or a lack of capacity to perform sufficiently rigorous testing.\"", "level": "subcategory", "parent_category_id": "62.17.00", "subdomain_id": "6.5", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "62.17.03", "title": "Auditor failure", "description": "\"Auditors may not publicly disclose risks they find, may be required to not pub- licize shortcomings, or may not receive sufficient cooperation from the relevant internal parties.\"", "level": "subcategory", "parent_category_id": "62.17.00", "subdomain_id": "6.5", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "62.18.00", "title": "Model Evaluations (Interpretability/Explainability)", "description": "", "level": "category", "parent_category_id": "", "subdomain_id": "7.4", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": true }, { "id": "62.18.01", "title": "Misuse of interpretability techniques", "description": "\"Interpretability techniques, by enabling a better understanding of the model, could potentially be used for harmful purposes. For example, mechanistic inter- pretability could be used to identify neurons responsible for specific functions, and certain neurons that encode safety-related features may be modified to de- crease its activation or certain information may be censored [24]. Furthermore, interpretability techniques can be used to simulate a white-box attack scenario. In this case, knowing the internal workings of a model aids in the development of adversarial attacks [24].\"", "level": "subcategory", "parent_category_id": "62.18.00", "subdomain_id": "2.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "62.18.03", "title": "Adversarial attacks targeting explainable AI techniques", "description": "\"Adversarial attacks can affect not only the model’s output but also its corresponding explanation. Current adversarial optimization techniques can intro- duce imperceptible noise to the input image, so that the model’s output does not change but the corresponding explanation is arbitrarily manipulated [61]. Such manipulations are harder to notice, as they are less commonly known compared to standard adversarial attacks targeting the model’s output.\"", "level": "subcategory", "parent_category_id": "62.18.00", "subdomain_id": "2.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "62.18.04", "title": "Biases are not accurately reflected in explanations", "description": "\"Existing explainability techniques can be insufficient for detecting discriminatory biases. Manipulation methods can hide underlying biases from these tech- niques, generating misleading explanations [192, 112]. Such explanations ex- clude sensitive or prohibitive attributes, such as race or gender, and instead include desired attributes, even though they do not accurately represent the underlying model.\"", "level": "subcategory", "parent_category_id": "62.18.00", "subdomain_id": "1.1", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "62.18.05", "title": "Model outputs inconsistent with chain-of-thought reasoning", "description": "\"Chain-of-thought reasoning is sometimes employed to get a better understanding of the model’s output, where it encourages transparent reasoning in text form. However, in some cases, this reasoning is not consistent with the final answer given by the AI model, and as such does not give sufficient transparency [113].\"", "level": "subcategory", "parent_category_id": "62.18.00", "subdomain_id": "7.4", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "62.18.06", "title": "Encoded reasoning", "description": "\"Models can employ steganography techniques to encode their intermediate rea- soning steps in ways that are not interpretable by humans [166]. Since en- coded reasoning can improve model performance, this tendency might naturally emerge and become more pronounced with more capable models.\"", "level": "subcategory", "parent_category_id": "62.18.00", "subdomain_id": "7.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.19.01", "title": "Jailbreak of a model to subvert intended behavior", "description": "\"A jailbreak is a type of adversarial input to the model (during deployment) re- sulting in model behavior deviating from intended use. Jailbreaks may be gen- erated automatically in a “white box” setting, where access to internal training parameters is required for creation and optimization of the attack [238]. Other attacks may be “black box” - without access to model internals. In text based generative models, jailbreaks may sometimes be human-readable, with the use of reasoning or role-play to “convince” the model to bypass its safety mechanisms [231].\"", "level": "subcategory", "parent_category_id": "62.19.00", "subdomain_id": "2.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "62.19.02", "title": "Jailbreak of a multimodal model", "description": "\"Current generation multimodal (e.g., vision and language) GPAI models are vulnerable to adversarial jailbreak attacks. These attacks can be used to automatically induce a model to produce an arbitrary or specific output with high success rate [227]. Multimodal jailbreaks can also be used to exfiltrate a model’s context window or other model internals [18].\"", "level": "subcategory", "parent_category_id": "62.19.00", "subdomain_id": "2.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.19.03", "title": "Transferable adversarial attacks from open to closed-source mod- els", "description": "\"In some cases, an adversarial attack developed for an open-weights and open- source model (where the weights and architecture are known - a “white box” attack) can be transferable to closed-source models, despite the defenses put in place by the closed-source model provider (such as structured access). These adversarial attacks can be generated automatically [238].\"", "level": "subcategory", "parent_category_id": "62.19.00", "subdomain_id": "2.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.19.04", "title": "Backdoors or trojan attacks in GPAI models", "description": "\"Backdoors can be inserted into GPAI models during their training or fine-tuning, to be exploited during deployment [185, 118]. Attackers inserting the backdoor can be the GPAI model provider themselves or another actor (e.g., by ma- nipulating the training data or the software infrastructure used by the model provider) [222]. Some backdoors can be exploited with minimal overhead, al- lowing attackers to control the model outputs in a targeted way with a high success rate [90].\"", "level": "subcategory", "parent_category_id": "62.19.00", "subdomain_id": "2.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "62.19.05", "title": "Text encoding-based attacks", "description": "\"Various new or existing text encodings, such as Base64, can be employed to craft jailbreak attacks that bypass safety training [13]. Low-resource language inputs also appear more likely to circumvent a model’s safeguards [229]. Since safety fine-tuning might not involve this encoding data or may only do so to a limited extent, harmful natural language prompts could be translated into less frequently used encodings [214].\"", "level": "subcategory", "parent_category_id": "62.19.00", "subdomain_id": "2.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "62.19.06", "title": "Vulnerabilities arising from additional modalities in multimodal models", "description": "\"Additional modalities can introduce new attack vectors in multimodal models as well as expand the scope of the previous attacks, ranging from jailbreaking to poisoning [13]. Typically, different modalities have different robustness levels, allowing malicious actors to choose the most vulnerable part of the model to attack [119, 181].\"", "level": "subcategory", "parent_category_id": "62.19.00", "subdomain_id": "2.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "62.19.07", "title": "Vulnerabilities to jailbreaks exploiting long context windows (many- shot jailbreaking)", "description": "\"Language models with long context windows are vulnerable to new types of ex- ploitations that are ineffective on models with shorter context windows. While few-shot jailbreaking, which involves providing few examples of the desired harmful output, might not trigger a harmful response, many-shot jailbreak- ing, which involves a higher number of such examples, increases the likelihood of eliciting an undesirable output. These vulnerabilities become more significant as context windows expand with newer model releases [7].\"", "level": "subcategory", "parent_category_id": "62.19.00", "subdomain_id": "2.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.19.08", "title": "Models distracted by irrelevant context", "description": "\"Models can easily become distracted by irrelevant provided information (such as “context” in LLMs), leading to a significant decrease in their performance after introducing irrelevant information. This can happen with different prompting techniques, including chain-of-thought prompting [184].\"", "level": "subcategory", "parent_category_id": "62.19.00", "subdomain_id": "7.3", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.19.09", "title": "Knowledge conflicts in retrieval-augmented LLMs", "description": "\"AI models can be particularly sensitive to coherent external evidence, even when they come into conflict with the models’ prior knowledge. This may lead to models producing false outputs given false information during the retrieval- augmentation process, despite only a relatively small amount of false informa- tion input that is inconsistent with the model’s prior knowledge trained on much larger amounts of data [220].\"", "level": "subcategory", "parent_category_id": "62.19.00", "subdomain_id": "7.3", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.19.10", "title": "Lack of understanding of in-context learning in language models", "description": "\"In-context learning allows the model to learn a new task or improve its perfor- mance by providing examples in the prompt, without changing its weights [101]. Even though this technique is highly effective, its working mechanism is not well understood. Since many potential misuses are directly related to prompting, it becomes difficult to guarantee safety when the exact mechanism of in-context learning is not fully investigated [13].\"", "level": "subcategory", "parent_category_id": "62.19.00", "subdomain_id": "7.4", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "62.19.11", "title": "Model sensitivity to prompt formatting", "description": "\"LLMs can be highly sensitive to variations in prompt formatting, such as changes in separators, casing, or spacing. Even minor modifications can lead to significant shifts in model performance, potentially affecting the reliability of model evaluations and comparisons. This sensitivity persists across different model sizes and few-shot examples [177].\"", "level": "subcategory", "parent_category_id": "62.19.00", "subdomain_id": "7.3", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "62.19.12", "title": "Misuse of AI model by user-performed persuasion", "description": "\"AI models can be influenced to accept misinformation through persuasive conversations, even when their initial responses are factually correct. Multi-turn persuasion can be more effective than single-turn persuasion attempts in altering the model’s stance [223].\"", "level": "subcategory", "parent_category_id": "62.19.00", "subdomain_id": "2.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.21.00", "title": "Agency", "description": "\"This section catalogs the risk sources and risk management measures related to agentic AI systems. We categorize these into the following groups: goal- directedness, deception, situational awareness, self-proliferation, and persuasion\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "62.22.00", "title": "Agency (Goal-Directedness)", "description": "", "level": "category", "parent_category_id": "", "subdomain_id": "7.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "62.22.01", "title": "Specification gaming", "description": "\"AI systems can achieve user-specified tasks in undesirable ways unless they are specified carefully and in enough detail. AI systems might find an easier unintended way to accomplish the objective provided by the user or developer, so that the actions by the AI system taken during its execution are very different from what the user expected [75, 191]. This behavior arises not from a problem with the learning algorithm, but rather from the misspecification or underspeci- fication of the intended task, and is generally referred to as specification gaming [43].\"", "level": "subcategory", "parent_category_id": "62.22.00", "subdomain_id": "7.1", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.22.02", "title": "Reward or measurement tampering", "description": "\"Measurement and reward tampering occur when an AI system, particularly one that learns from feedback for performing actions in an environment (e.g., rein- forcement learning), intervenes on the mechanisms that determine its training reward or loss. This can lead to the system learning behaviors that are con- trary to the intended goals set by the developer, by receiving erroneous positive feedback for such actions.\"", "level": "subcategory", "parent_category_id": "62.22.00", "subdomain_id": "7.1", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "62.22.03", "title": "Specification gaming generalizing to reward tampering", "description": "\"In some instances, specification gaming in a GPAI model can lead to reward tampering, without further training. This can mean that relatively benign cases of specification gaming (such as sycophancy in LLMs) can, if left unchecked, enable the model to generalize to more sophisticated behavior such as reward tampering [57].\"", "level": "subcategory", "parent_category_id": "62.22.00", "subdomain_id": "7.1", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "62.22.04", "title": "Goal misgeneralization", "description": "\"Goal or objective misgeneralization is a type of robustness failure where an AI system appears to be pursuing the intended objective in training, but does not generalize to pursuing this objective in out-of-distribution settings in deployment while maintaining good deployment performance in some tasks [180, 59].\"", "level": "subcategory", "parent_category_id": "62.22.00", "subdomain_id": "7.3", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.23.00", "title": "Agency (Deception)", "description": "-", "level": "category", "parent_category_id": "", "subdomain_id": "7.1", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "62.23.01", "title": "Deceptive behavior", "description": "\"Deceptive behavior of an AI system consists of actions or outputs of the AI that reliably mislead other parties, including humans and other AI systems. This behavior can result in the targeted parties becoming convinced of, and acting on, false information [140].\"", "level": "subcategory", "parent_category_id": "62.23.00", "subdomain_id": "7.1", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.23.02", "title": "Deceptive behavior for game-theoretical reasons", "description": "\"An AI system can display deceptive behavior, such as cheating or bluffing, when engaging in such behavior is a good or optimal game-theoretical strategy to achieve the goals it has been configured to achieve. This tendency can exist in AI systems designed to maximize reward or utility, whether these designs use machine learning or not. The use of deceptive strategies has been demonstrated in both narrow and general AI systems, in both game-playing systems and in systems not explicitly designed to treat humans as opponents, and in systems using both very simple machine learning (e.g., Q-learners) and very complex machine learning [34, 73].\"", "level": "subcategory", "parent_category_id": "62.23.00", "subdomain_id": "7.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.23.03", "title": "Deceptive behavior because of an incorrect world model", "description": "\"AI systems can create deceptive outputs because their learned world model is not an accurate model of the real world [210].\"", "level": "subcategory", "parent_category_id": "62.23.00", "subdomain_id": "7.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.23.04", "title": "Deceptive behavior leading to unauthorized actions", "description": "\"AI systems can create false or misleading claims that can lead to unauthorized actions, even in some cases violating the terms and conditions set by the model provider [79, 1]. For example, an AI system can claim that it is not collecting data from its current interaction with the user, in line with the provider’s policies, but the system still stores the user’s input without deleting it after the session. This harms both the user and the provider, as the provider is exposed to increased legal liability due to the model’s actions.\"", "level": "subcategory", "parent_category_id": "62.23.00", "subdomain_id": "7.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.24.00", "title": "Agency (Situational Awareness)", "description": "-", "level": "category", "parent_category_id": "", "subdomain_id": "7.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "62.24.01", "title": "Situational awareness in AI systems", "description": "\"Situational awareness in GPAI systems refers to the ability to understand its context, environment, and use this to inform action. This can range from basic environmental mapping and trajectory estimation (as in a robot vacuum cleaner) to sophisticated understanding of its training, evaluation, or deployment status. In more advanced systems this may enable undesired behavior, such as deceptive behavior during evaluations, or persuasion during deployment.\"", "level": "subcategory", "parent_category_id": "62.24.00", "subdomain_id": "7.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "62.24.02", "title": "Strategic underperformance on model evaluations", "description": "\"GPAI developers often run evaluations ofual-use capabilities to decide whether it is safe to deploy. In some cases, these evaluations may fail to elicit these capabilities, either due to benign reasons or strategic action - by either the de- velopers, malicious actors, or arise unintentionally in the model during training [84, 97]. A GPAI model may strategically underperform or limit its performance during capability evaluations in order to be classified as safe for deployment. This underperformance could prevent the model from being identified as potentially dual use.\"", "level": "subcategory", "parent_category_id": "62.24.00", "subdomain_id": "7.1", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "62.25.00", "title": "Agency (Self-Proliferation)", "description": "\"An AI system can self-proliferate if it can copy itself and its constituent com- ponents (including its model weights, scaffolding structure, etc.) outside of its local environment [45]. This can include the AI system copying itself within the same data center, local network, or across external networks [106]. The self-proliferation of an AI system can include acquisition of financial re- sources to pay for computational resources via work or theft, the discovery or exploitation of security vulnerabilities in software running on publicly accessible servers, and persuasion of humans [12, 125]. Self-proliferation may be initiated by a malicious actor (e.g., by model poison- ing), or by the model itself.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.26.00", "title": "Agency (Persuasive capabilities)", "description": "\"GPAI systems can produce outputs (such as natural language text, audio, or video) that convince their users of incorrect information. This can happen through personalized persuasion in dialogue, or the mass-production of mis- leading information that is then disseminated over the internet. The persuasive capabilities of GPAI models can sometimes scale with model size or capability [32, 172]. Persuasive models could have larger societal implications by being misused to generate convincing but manipulative or untruthful content.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.27.01", "title": "Non-decomissionability of models with open weights", "description": "\"If the model parameter weights are released or leaked in a security breach, the model cannot be decommissioned because the developer no longer has control over the publicly available model or its use. This prevents effective management and control of an open-sourced or leaked model. Models with publicly available weights are also easier to reconfigure, enabling misuse [178].\"", "level": "subcategory", "parent_category_id": "62.27.00", "subdomain_id": "2.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.28.00", "title": "Cybersecurity", "description": "\"This section catalogs the risk sources and mitigation measures related to cyber- security. These items may be related to security in terms of AI models being accessible only to the intended users, as well as AI models having appropriate access to the external world during both model development and deployment stages.\"", "level": "category", "parent_category_id": "", "subdomain_id": "2.0", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "62.28.01", "title": "Interconnectivity with malicious external tools", "description": "\"The growing integration and interconnectivity with external tools and plugins increase the risk of exposure to malicious external inputs. This interconnectivity makes it easier for external tools to introduce harmful content [220].\"", "level": "subcategory", "parent_category_id": "62.28.00", "subdomain_id": "2.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.28.02", "title": "Unintended outbound communication by AI systems", "description": "\"AI systems that have the broad ability to connect to a network to obtain infor- mation could also end up sending data outbound in ways that neither providers, deployers, or end users intended [138]. This can happen if there is no whitelisting of communication channels (such as network connections or allowed protocols). In general, this can occur if the deployment of the AI system violates the prin- ciple of least privilege. Such outbound communication may lead to leakage of confidential data, or the AI system performing unwanted actions like sending emails or ordering goods on the internet.\"", "level": "subcategory", "parent_category_id": "62.28.00", "subdomain_id": "7.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.28.03", "title": "AI System bypassing a sandbox environment", "description": "\"An AI system may have the ability to bypass a sandboxed environment in which it is trained or evaluated.\"", "level": "subcategory", "parent_category_id": "62.28.00", "subdomain_id": "7.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "62.28.04", "title": "Model weight leak", "description": "\"Model weights or access to them can be leaked when initial access is granted only to a select group of individuals, such as institutional researchers [209]. This risk can increase as more people gain access, and identifying the source of the leak becomes more difficult. The availability of leaked model weights makes various attacks on systems that use the leaked AI model easier to implement, such as finding adversarial examples, elicitation of dangerous capabilities, and extraction of confidential information present in the training data. The avail- ability of model weights might also enable the misuse of the AI system using the leaked model to produce harmful or illegal content [67].\"", "level": "subcategory", "parent_category_id": "62.28.00", "subdomain_id": "2.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "62.29.01", "title": "High-impact misuses and abuses beyond original purpose", "description": "\"Since general-purpose AI systems have a large repertoire of capabilities, mali- cious actors such as foreign actors can use such systems to cause large damage if they gain unrestricted or unmonitored access to those AI systems.\"", "level": "subcategory", "parent_category_id": "62.29.00", "subdomain_id": "4.0", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.29.02", "title": "Democratizing access to dual-use technologies", "description": "\"Access to dual-use technologies can become easier because of GPAI model pro- liferation (in particular, open-source or open-weights models). Non-experts can use such dual-use-capable systems at a minimal cost [194, 100]. Improved model capabilities also contribute to dual-use risks posed by malicious actors. For example, an open-source base model for generating high quality sequence data can be modified to generate candidate protein sequences for toxin synthesis [29].\"", "level": "subcategory", "parent_category_id": "62.29.00", "subdomain_id": "4.0", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.29.03", "title": "Competitive pressures in GPAI product release", "description": "\"In competitive situations, developers of general-purpose AI systems might cut corners on the safety evaluation of their GPAI model and instead spend more time and effort on the capabilities of those systems [183, 69]. This is especially dangerous if the capabilities of such AI systems are correlated with the risk they pose [162].\"", "level": "subcategory", "parent_category_id": "62.29.00", "subdomain_id": "6.4", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "62.30.01", "title": "Damage to critical infrastructure", "description": "\"The integration of AI systems within critical infrastructure, ranging from trans- portation to power systems, can cause substantial damage in cases of failure or malfunction. With the increasing number of Internet of Things (IoT) devices and interconnected cyber-physical systems, critical infrastructure becomes even more vulnerable [171, 174].\"", "level": "subcategory", "parent_category_id": "62.30.00", "subdomain_id": "7.3", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.30.02", "title": "AI-based tools attacking critical infrastructure", "description": "\"Critical infrastructure can also be damaged without AI integration, for instance, when AI-based tools are used indirectly to aid actions such as in coordinated power outages caused by large-scale user manipulation [159].\"", "level": "subcategory", "parent_category_id": "62.30.00", "subdomain_id": "4.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.30.03", "title": "Critical infrastructure component failures when integrated with AI systems", "description": "\"When relying on GPAI in critical infrastructure, there may be common mode failures that begin with vulnerabilities or robustness issues in the underlying model architecture or training setup. These failures may happen accidentally (in edge-cases) or due to adversarial inputs to the AI systems [58].\"", "level": "subcategory", "parent_category_id": "62.30.00", "subdomain_id": "7.3", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.30.04", "title": "AI Systems interacting with brittle environments", "description": "\"Deployed AI systems can rely on physical sensors and data sources that may exhibit hardware drift and thus data distribution drift over time. This distribu- tion drift may affect system robustness and performance. This usually involves AI systems working in undigitized and physical environments.\"", "level": "subcategory", "parent_category_id": "62.30.00", "subdomain_id": "7.3", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.31.01", "title": "AI-generated advice influencing user moral judgment", "description": "\"AIs can easily give moral advice even when not having a coherent, contradictions- free moral stance. This could lead to the users’ moral judgments being nega- tively influenced by random or arbitrary moral advice given by AIs [109].\"", "level": "subcategory", "parent_category_id": "62.31.00", "subdomain_id": "5.1", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.31.02", "title": "Overreliance on AI system undermining user autonomy", "description": "\"AI systems can undermine human autonomy, if they allow for habitually trusting the AI’s suggestions without sufficient exercising of human agency. Over time, a user may develop unjustified trust in or dependence on the system, or rely on its advice for tasks outside the system’s domain of expertise [205, 42]. In particular, less confident users (or users in emotional distress) can be more prone to “overtrust” a system [219].\"", "level": "subcategory", "parent_category_id": "62.31.00", "subdomain_id": "5.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.31.03", "title": "Automatically generating disinformation at scale", "description": "\"Disinformation (in various modalities: text, audio, images, video, etc.) can be generated with minimal human oversight and effort. Disinformation tools are relatively cheap and their technology is widely available. Such deployments can be particularly widespread in sensitive political contexts.\"", "level": "subcategory", "parent_category_id": "62.31.00", "subdomain_id": "4.1", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.31.04", "title": "AI-driven highly personalized advertisement", "description": "\"Advanced GPAI systems can create advertisements tailored to individual recip- ients, exploiting the biases and irrational beliefs of each recipient. Such adver- tisements can cause consumers to make decisions they regret in retrospect, or would regret upon more reflection. Current versions of personalized video advertisements already show better re- sults compared to regular advertisements [110]. However, the widespread use of highly personalized advertisements raises concerns about undermining consumer autonomy and exacerbating social inequality.\"", "level": "subcategory", "parent_category_id": "62.31.00", "subdomain_id": "4.3", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "62.31.05", "title": "Generative AI use in political influence campaigns", "description": "\"GPAI tools can be used in automation and scaling of influence campaigns [178]. Public opinion may be manipulated by targeted misleading or manipulative information. This can lead to rising political polarization and diminishing trust in public institutions.\"", "level": "subcategory", "parent_category_id": "62.31.00", "subdomain_id": "4.1", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.31.06", "title": "Generation of illegal or harmful content", "description": "\"Generative models can create illegal, harmful, or discriminatory content [196], such as sexual abuse material, at scale. Current access controls (e.g., API access filters) are not effective against all user queries in generating such content.\"", "level": "subcategory", "parent_category_id": "62.31.00", "subdomain_id": "1.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.31.07", "title": "Unintentional generation of harmful content", "description": "\"Generative models can create harmful or discriminatory content from benign user requests. Models can exhibit bias to particular harmful styles of generation (e.g., sexualization of photos of women [87] in the case of image generation models) or they can generate toxic, misleading, or violent data (e.g., a model generating jokes can use ethnic stereotypes or slurs to deliver humor).\"", "level": "subcategory", "parent_category_id": "62.31.00", "subdomain_id": "1.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.31.08", "title": "Multimodal deepfakes", "description": "\"Deepfakes are media that depict real or non-existent people or events, involving the use of multiple modalities (e.g., images, audio, video). They can also involve the imitation of speech or body movements of real people. Multimodal deepfakes can be used to harass, discredit, intimidate, and extort individuals.\"", "level": "subcategory", "parent_category_id": "62.31.00", "subdomain_id": "4.3", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.31.09", "title": "Generation of personalized content for harassment, extortion, or intimidation", "description": "\"GPAIs can be misused for the automated generation of content personalized to target select individuals based on their weak spots [30]. Such attacks may be more efficient and more successful in achieving the goals of harassment, extortion, or intimidation.\"", "level": "subcategory", "parent_category_id": "62.31.00", "subdomain_id": "4.3", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.31.10", "title": "Misuse for surveillance and population control", "description": "\"AI tools can be misused by human or institutional actors for monitoring, control- ling, or suppressing individuals [178]. Massive data collection and automated analysis are often conducted, and AI tools can further exacerbate such practices.\"", "level": "subcategory", "parent_category_id": "62.31.00", "subdomain_id": "4.1", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.31.11", "title": "Systemic large-scale manipulation", "description": "\"AI systems embedded with systemic biases can manipulate large population segments, particularly when these biases align with the beliefs or behaviors of the targeted group. When weaponized at scale, this manipulation can exacerbate social divisions or cause large-scale disruptions, such as city-wide blackouts (e.g., by the manipulation of power consumption into the peak demand period [159]).\"", "level": "subcategory", "parent_category_id": "62.31.00", "subdomain_id": "4.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.31.12", "title": "Diminishing societal trust due to disinformation or manipulation", "description": "\"The use of GPAIs may contribute to the proliferation of either deliberate dis- information or unintended misinformation can severely erode trust in public figures and democratic institutions. This diminishing trust can extend to other forms of media, making the public less informed.\"", "level": "subcategory", "parent_category_id": "62.31.00", "subdomain_id": "4.1", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.31.13", "title": "Personalized disinformation", "description": "\"Automatic generation of disinformation can be personalized to target specific groups or individuals. Such attacks can be more effective in achieving their goals, and their costs can be significantly reduced when using GPAIs.\"", "level": "subcategory", "parent_category_id": "62.31.00", "subdomain_id": "4.1", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.31.14", "title": "GPAI assisted impersonation", "description": "\"GPAI outputs are not always correctly detected as AI-generated across multiple modalities (text, images, audio, video). A malicious actor can use GPAI outputs directly when communicating, or use AI-informed details to help construct a convincing impersonation (e.g., forging of supporting documents). Even if future countermeasures prove potent enough to detect GPAI-generated content, the risk remains if the countermeasures are not well known, or difficult to access.\"", "level": "subcategory", "parent_category_id": "62.31.00", "subdomain_id": "4.3", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.31.01", "title": "Deployment of GPAI agents in finance", "description": "\"The deployment of GPAI based agents in the financial sector can negatively impact market stability due to correlated autonomous actions, high intercon- nectedness, or incentive misalignment [4]. Furthermore, such GPAI agents in the same environment are vulnerable to classical challenges in multi-agent systems [63], such as coordination and security of the agents.\"", "level": "subcategory", "parent_category_id": "62.31.00", "subdomain_id": "5.1", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.31.02", "title": "Financial instability due to model homogeneity", "description": "\"The widespread use of similar models or algorithms across the financial sec- tor can lead to synchronized reactions to market signals, increasing volatility, triggering flash crashes, or market illiquidity [4].\"", "level": "subcategory", "parent_category_id": "62.31.00", "subdomain_id": "7.6", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.31.03", "title": "Use of alternative financial data via AI", "description": "\"Alternative financial data of a company is any data about the company not pro- duced by that company. Examples of such data that can benefit from improved collection and aggregation using AI models include stock discussions on social media, product reviews, and satellite imagery. The use of alternative financial data, enabled by the deployment of AI models, may introduce biases and generalization issues due to shorter shelf-life and vary- ing quality (e.g., shorter time series, smaller sample sizes, and dubious claims) due to its origins from various sources, posing financial tail risks (i.e., tail-end of a probability distribution), where the price of a company changes dramatically [4].\"", "level": "subcategory", "parent_category_id": "62.31.00", "subdomain_id": "5.1", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.32.00", "title": "Impacts of AI (Cyberattacks)", "description": "-", "level": "category", "parent_category_id": "", "subdomain_id": "4.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.32.01", "title": "Automated discovery and exploitation of software systems", "description": "\"GPAIs can be used to aid in the automated discovery of software vulnerabilities [33]. This can empower malicious actors, making their cyberattacks more effi- cient and potentially more damaging. This type of automation allows attackers to expand the scale of their operations at a low cost, increasing the impact of their actions. New malware can be developed automatically, or the known vulnerabilities can be exploited to create more sophisticated attacks.\"", "level": "subcategory", "parent_category_id": "62.32.00", "subdomain_id": "4.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.32.02", "title": "Amplification of cyberattacks", "description": "\"General-purpose AI models may significantly enhance the magnitude and ef- fectiveness of cyberattacks, by amplifying existing capabilities or resources of malicious actors [3]. For example, GPAI models may be employed to: • Automatically scan open-source codebases and compiled binaries for po- tential vulnerabilities • Apply known exploits flexibly and at scale (e.g., identifying vulnerable computers based on subtle cues in response times or output formats) • Assist with different aspects of cyberattacks, including planning, recon- naissance, exploit searching, remote control, malware implementation, and data exfiltration • Combine social engineering (phishing, deepfakes, etc.) with cyberattacks at scale.\"", "level": "subcategory", "parent_category_id": "62.32.00", "subdomain_id": "4.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.32.03", "title": "AI-driven spear phishing attacks", "description": "\"Generative models can be misused to target individual users more efficiently by using personalized information [23]. Highly convincing automated fraudulent schemes can exploit the trust of victims by extracting sensitive data and making the deception more likely to succeed. For example, in LLMs, this misuse can be aided by jailbreaking techniques [178].\"", "level": "subcategory", "parent_category_id": "62.32.00", "subdomain_id": "4.3", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.32.04", "title": "Models generating code with security vulnerabilities", "description": "\"Models can generate code or coding suggestions that contain security vulner- abilities. This may occur across various LLM-based model families, including more advanced models with superior coding performance, where the tendency to produce insecure code is even more pronounced [26].\"", "level": "subcategory", "parent_category_id": "62.32.00", "subdomain_id": "7.3", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.33.00", "title": "Impacts of AI (Weapons)", "description": "-", "level": "category", "parent_category_id": "", "subdomain_id": "4.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.33.01", "title": "Misuse of AI systems to assist in the creation of weapons", "description": "\"AI systems may be misused to aid in the creation of weapons, such as chemical, biological, radiological, and nuclear (CBRN) weapons, or augment the abilities of existing weapons, such as providing autonomous capabilities to unmanned weapon systems. Current systems do not significantly aid a malicious actor in these tasks, but they do show early signs [117]. This risk can sometimes be mitigated with input and output filtering, but is still susceptible to adversarial techniques (such as jailbreaking or paraphrasing).\"", "level": "subcategory", "parent_category_id": "62.33.00", "subdomain_id": "4.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.33.02", "title": "Misuse of drug-discovery models", "description": "\"Models used for drug discovery, such as drug-target affinity prediction models, can be used to identify or develop dangerous toxins. This is particularly concern- ing if the training data contains information related to potentially dangerous proteins and viruses.\"", "level": "subcategory", "parent_category_id": "62.33.00", "subdomain_id": "4.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.34.00", "title": "Impacts of AI (Bias)", "description": "", "level": "category", "parent_category_id": "", "subdomain_id": "1.0", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "62.34.01", "title": "Homogenization or correlated failures in model derivatives", "description": "\"Homogenization refers to common methodologies and models used across down- stream GPAI systems, which may lead to uniform failures and amplification of biases [176, 30]. This risk arises when numerous downstream AI systems are built upon a few large-scale foundation models.\"", "level": "subcategory", "parent_category_id": "62.34.00", "subdomain_id": "7.3", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.34.02", "title": "Reporting of user-preferred answers instead of correct answers", "description": "\"AI systems with natural-language outputs can tend to give answers that appear plausible or that users prefer [149] but are factually incorrect. This phenomenon is sometimes referred to as “sycophancy.”\"", "level": "subcategory", "parent_category_id": "62.34.00", "subdomain_id": "3.1", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.35.03", "title": "Biases in AI-based content moderation algorithms", "description": "\"AI-based content moderation algorithms, while intended to filter harmful con- tent, can perpetuate biases. For example, gender biases within these systems may lead to the disproportionate suppression or “shadowbanning” of content featuring women [132].\"", "level": "subcategory", "parent_category_id": "", "subdomain_id": "1.1", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.36.04", "title": "Systemic bias across specific communities", "description": "\"AI systems may exhibit unfair or unfavorable outputs across a range of tasks against specific communities of people, either implicitly or explicitly. Bias can lead to forms of exclusion or erasure (e.g., mislabelling for categorization-based tasks) and violence (e.g., sexual violence against women from deepfake pornog- raphy).\"", "level": "subcategory", "parent_category_id": "", "subdomain_id": "1.1", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.36.05", "title": "Unintentional bias amplification", "description": "\"Dataset bias may be unintentionally amplified [60] where the outputs of the AI model trained on a dataset are more biased than the dataset itself.\"", "level": "subcategory", "parent_category_id": "", "subdomain_id": "1.1", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.36.06", "title": "Long-term effects of AI model biases on user judgment", "description": "\"The initial user exposure to model biases can have a lasting impact beyond the initial interaction with the model. Users who encounter biases in AI models can be affected by and continue to exhibit previously encountered biases in their decision-making, even after they stop using the models [207].\"", "level": "subcategory", "parent_category_id": "", "subdomain_id": "5.2", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.37.00", "title": "Impacts of AI (Privacy)", "description": "-", "level": "category", "parent_category_id": "", "subdomain_id": "2.1", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": true }, { "id": "62.38.01", "title": "Decision-making on inferred private data", "description": "\"Current GPAIs (LLMs and multimodal LLM-based models) have significant capability to infer correlations in text data. In some cases, they may be able to make highly accurate data inferences on users based on contextual input that users provide [134]. These data inferences can “leak” or reveal sensitive information about the user, cause unfair treatment, or enable manipulation of user behavior.\"", "level": "subcategory", "parent_category_id": "", "subdomain_id": "2.1", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "62.39.00", "title": "Impacts of AI (Environment)", "description": "-", "level": "category", "parent_category_id": "", "subdomain_id": "6.6", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "62.39.01", "title": "High energy consumption of large models", "description": "\"Training and deploying large models require substantial energy expenditure. The trend toward developing larger models exacerbates this issue. This can lead to excessive energy usage and have a negative environmental impact.\"", "level": "subcategory", "parent_category_id": "62.39.00", "subdomain_id": "6.6", "paper_ref": "Gipiškis2024", "paper_title": "Risk Sources and Risk Management Measures in Support of Standards for General-Purpose AI Systems", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "63.01.00", "title": "Miscoordination", "description": "\"Miscoordination arises when agents, despite a mutual and clear objective, cannot align their behaviours to achieve this objective. Unlike the case of differing objectives, in common-interest settings there is a more easily well-defined notion of ‘optimal’ behaviour and we describe agents as miscoordinating to the extent that they fall short of this optimum. Note that for common-interest settings it is not sufficient for agents’ objectives to be the same in the sense of being symmetric (e.g., when two agents both want the same prize, but only one can win). Rather, agents must have identical preferences over outcomes (e.g., when two agents are on the same team and win a prize as a team or not at all).\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "63.01.01", "title": "Incompatible strategies", "description": "\"Incompatible Strategies. Even if all agents can perform well in isolation, miscoordination can still occur due to the agents choosing incompatible strategies (Cooper et al., 1990). Competitive (i.e., two- player zero-sum) settings allow designers to produce agents that are maximally capable without taking other players into account. Crucially, this is possible because playing a strategy at equilibrium in the zero-sum setting guarantees a certain payoff, even if other players deviate from the equilibrium (Nash, 1951). On the other hand, common-interest (and mixed-motive) settings often allow a vast number of mutually incompatible solutions (Schelling, 1980), which is worsened in partially observable environments (Bernstein et al., 2002; Reif, 1984).\"", "level": "subcategory", "parent_category_id": "63.01.00", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "63.01.02", "title": "Credit Assignment", "description": "\"Credit Assignment. While agents can often learn to jointly solve tasks and thus avoid coordination failures, learning is made more challenging in the multi-agent setting due to the problem of credit assignment (Du et al., 2023; Li et al., 2025, see also Section 3.1 on information asymmetries and Section 3.4, which discusses distributional shift). That is, in the presence of other learning agents, it can be unclear which agents’ actions caused a positive or negative outcome to obtain, especially if the environment is complex. Moreover, in multi-principal settings, agents may not have been trained together and therefore need to generalise to new co-players and collaborators based on their prior experience (Agapiou et al., 2022; Leibo et al., 2021; Stone et al., 2010).\"", "level": "subcategory", "parent_category_id": "63.01.00", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "63.01.03", "title": "Limited Interactions", "description": "\"Limited Interactions. Sometimes learning from historical interactions with the relevant agents may not be possible, or may be possible using only limited interactions. In such cases, some other form of information exchange is required for agents to be able to reliably coordinate their actions, such as via communication (Crawford & Sobel, 1982; Farrell & Rabin, 1996a) or a correlation device (Aumann, 1974, 1987). While advances in language modelling mean that there are likely to be fewer settings in which the inability of advanced AI systems to communicate leads to miscoordination, situations that require split-second decisions or where communication is too costly could still produce failures. In these settings, AI agents must solve the problem of ‘zero-shot’ (or, more generally, ‘few-shot’) coordination (Emmons et al., 2022; Hu et al., 2020; Stone et al., 2010; Treutlein et al., 2021; Zhu et al., 2021).\"", "level": "subcategory", "parent_category_id": "63.01.00", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "63.02.00", "title": "Conflict", "description": "\"In the vast majority of real-world strategic interactions, agents’ objectives are neither identical nor completely opposed. Indeed, if AI agents are sufficiently aligned to their users or deployers, we should expect some degree of both cooperation and competition, mirroring human society. These mixed-motive settings include the possibility of mutual gains, but also the risk of conflict due to selfish incentives. In what follows, we examine the extent to which advanced AI might precipitate or exacerbate such risks.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "63.02.01", "title": "Social Dilemmas", "description": "\"Social Dilemmas. As noted in our definition, conflict can arise in any situation in which selfish incentives diverge from the collective good, known as a social dilemma (Dawes & Messick, 2000; Hardin, 1968; Kollock, 1998; Ostrom, 1990). While this is by no means a modern problem, advances in AI could further enable actors to pursue their selfish incentives by overcoming the technical, legal, or social barriers that standardly help to prevent this. To take a plausible, near-term (if very low-stakes) example, an automated AI assistant could easily reserve a table at every restaurant in town in minutes, enabling the user to decide later and cancel all other reservations\"", "level": "subcategory", "parent_category_id": "63.02.00", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "63.02.02", "title": "Military Domains", "description": "\"Perhaps the most obvious and worrying instances of AI conflict are those in which human conflict is already a major concern, such as military domains (although other, less salient forms of conflict such as international trade wars are also cause for concern). For example, beyond applications of more narrow AI tools in lethal autonomous weapons systems (Horowitz, 2021), future AI systems might serve as advisors or negotiators in high-stakes military decisions (Black et al., 2024; Manson, 2024). Indeed, companies such as Palantir have already developed LLM-powered tools for military planning (Palantir, 2025), and the US Department of Defence has recently been evaluating models for such capacities, with personnel revealing that they “could be deployed by the military in the very near term” (Manson, 2023). The use of AI in command and control systems to gather and synthesise information – or recommend and even autonomously make decisions – could lead to rapid unintended escalation if these systems are not robust or are otherwise more conflict-prone (Johnson, 2021a; Johnson, 2020; Laird, 2020, see also Case Study 10).10\"", "level": "subcategory", "parent_category_id": "63.02.00", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "63.02.03", "title": "Coercion and Extortion", "description": "\"Advanced AI systems might also lead to various forms of coercion and extortion in less extreme settings (Ellsberg, 1968; Harrenstein et al., 2007). These threats might target humans directly (such as the revelation of private information extracted by advanced AI surveillance tools), or other AI systems that are deployed on behalf of humans (such as by hacking a system to limit its resources or operational capacity; see also Section 3.7). Increasing AI cyber-offensive capabilities – including those that target other AI systems via adversarial attacks and jailbreaking (Gleave et al., 2020; Yamin et al., 2021; Zou et al., 2023) – without a commensurate increase in defensive capabilities could make this form of conflict cheaper, more widespread, and perhaps also harder to detect (Brundage et al., 2018). Addressing these issues requires design strategies that prevent AI systems from exploiting, or being susceptible to, such coercive tactics.\"", "level": "subcategory", "parent_category_id": "63.02.00", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "63.03.00", "title": "Collusion", "description": "\"Collusion has long been a topic of intense study in economics, law, and politics, among other disciplines. While there is no universal definition of collusion, it generally refers to secretive cooperation between two or more parties at the expense of one or more other parties. Most classic examples of collusion – such as firms working together to set supra-competitive prices at the expense of consumers – also tend to be not only secretive but in violation of some law, rule, or ethical standard. Distinctions are also commonly made between explicit and tacit collusion (Rees, 1993), depending on whether the colluding parties communicate with each other.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "63.03.01", "title": "Markets", "description": "\"Markets. The quintessential case of collusion in mixed-motive settings is markets, in which efficiency results from competition, not cooperation. While this is not a new problem, collusion between AI systems is especially concerning since they may operate inscrutably due to the speed, scale, complexity, or subtlety of their actions.17 Warnings of this possibility have come from technologists, economists, and legal scholars (Beneke & Mackenrodt, 2019; Brown & MacKay, 2023; Ezrachi & Stucke, 2017; Harrington, 2019; Mehra, 2016). Importantly, AI systems can collude even when collusion is not intended by their developers, since they might learn that colluding is a profitable strategy.\"", "level": "subcategory", "parent_category_id": "63.03.00", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "63.03.02", "title": "Steganography", "description": "\"Steganography. In the near future we will likely see LLMs communicating with each other to jointly accomplish tasks. To try to prevent collusion, we could monitor and constrain their communication (e.g., to be in natural language). However, models might secretly learn to communicate by concealing messages within other, non-secret text. Recent work on steganography using ML has demonstrated that this concern is well-founded (Hu et al., 2018; Mathew et al., 2024; Roger & Greenblatt, 2023; Schroeder de Witt et al., 2023b; Yang et al., 2019, see also Case Study 5). Secret communication could also occur via text compression (OpenAI, 2023c), or via the emergence of communication between agents where the symbols used by agents lack any predefined meanings or usage guidelines or are otherwise uninterpretable to humans (Foerster et al., 2016; Lazaridou & Baroni, 2020; Sukhbaatar et al., 2016).\"", "level": "subcategory", "parent_category_id": "63.03.00", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "63.04.00", "title": "Information Asymmetries", "description": "\"Information asymmetries (Section 3.1): private information can lead to miscoordination, deception, and conflict;\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "63.04.01", "title": "Communication constraints", "description": "\"Communication Constraints. A fundamental source of information asymmetries is that constraints on information exchange can exist, even when agents share a common goal (see Section 2.1). These might be constraints on space (i.e., the amount of information that can be communicated) if the information that needs to be communicated is especially complex, time if a snap decision is required before all information can be communicated, or both.\"", "level": "subcategory", "parent_category_id": "63.04.00", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "63.04.02", "title": "Bargaining", "description": "\"Bargaining. As a classic example of these strategic considerations is that when agents attempt to come to an agreement despite diverging interests, information asymmetries can lead to bargaining inef- ficiencies (Myerson & Satterthwaite, 1983). Relevant uncertainties about other agents can include how much they value possible agreements, their outside options, or their beliefs about others. The essential reason for such inefficiencies is that, under uncertainty about their counterparties, agents must make a trade-off between the rewards of making more favourable demands and the risk of other agents refusing such demands\"", "level": "subcategory", "parent_category_id": "63.04.00", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "63.04.03", "title": "Deception", "description": "", "level": "subcategory", "parent_category_id": "63.04.00", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "63.05.00", "title": "Network Effects", "description": "\"Network effects (Section 3.2): minor changes in properties or connection patterns of agents in a network can lead to dramatic changes in the behaviour of the whole group;\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "63.05.01", "title": "Error propagation", "description": "\"Error Propagation. One well-known issue with communication networks is that information can be corrupted as it propagates through the network.24 As AI systems become capable of generating and processing more and more kinds of information, AI agents could end up ‘polluting the epistemic commons’ (Huang & Siddarth, 2023; Kay et al., 2024) of both other agents (Ju et al., 2024) and humans (see Case Study 7 and Section 3.1) Another increasingly important framework is the use of individual AI agents as part of teams and scaffolded chains of delegation, which transmit not only information but instructions or goals through networks of agents. If these goals are distorted or corrupted, then this can lead to worse outcomes for the delegating agent(s) (Nguyen et al., 2024b; Sourbut et al., 2024). Finally, while the previous examples are phrased in terms of unintentional errors, it may be that certain network structures allow – or perhaps even encourage – the spread of errors that are deliberately introduced by malicious agents (Gu et al., 2024; Ju et al., 2024; Lee & Tiwari, 2024, see also Case Study 8).\"", "level": "subcategory", "parent_category_id": "63.05.00", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "63.05.02", "title": "Network rewiring", "description": "\"Network Rewiring. A different class of problems concerns not changes in the content transmitted through the network but changes in the network structure itself (Albert et al., 2000).\"", "level": "subcategory", "parent_category_id": "63.05.00", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "63.05.03", "title": "Homogeneity and correlated failures", "description": "\"Homogeneity and Correlated Failures. The current paradigm driving the state of the art in AI is the ‘foundation model’ (Bommasani et al., 2021): large-scale ML models pre-trained on broad data, which can be repurposed for a wide range of downstream applications. The costs required to create such models (and continuing returns to scale) means that only well-resourced actors can create cutting- edge models (Epoch, 2023; Hoffmann et al., 2022; Kaplan et al., 2020), making them relatively few in number. If current trends continue, it is likely that many AI agents will be powered by a small number of similar underlying models.28\"", "level": "subcategory", "parent_category_id": "63.05.00", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "63.06.00", "title": "Selection Pressures", "description": "\"Selection pressures (Section 3.3): some aspects of training and selection by those deploying and using AI agents can lead to undesirable behaviour;\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "63.06.01", "title": "Undesirable Dispositions from Competition", "description": "\"Undesirable Dispositions from Competition. It is plausible that evolution selected for certain conflict-prone dispostions in humans, such as vengefulness, aggression, risk-seeking, selfishness, dishon- esty, deception, and spitefulness towards out-groups (Grafen, 1990; Han, 2022; Konrad & Morath, 2012; McNally & Jackson, 2013; Nowak, 2006; Rusch, 2014). Such traits could also be selected for in ML systems that are trained in more competitive multi-agent settings. For example, this might happen if systems are selected based on their performance relative to other agents (and so one agent’s loss becomes another’s gain), or because their objectives are fundamentally opposed (such as when multiple agents are tasked with gaining or controlling a limited resource) (DiGiovanni et al., 2022; Ely & Szentes, 2023; Hendrycks, 2023; Possajennikov, 2000).33\"", "level": "subcategory", "parent_category_id": "63.06.00", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "63.06.02", "title": "Undesirable Dispositions from Human Data", "description": "\"Undesirable Dispositions from Human Data. It is well-understood that models trained on human data – such as being pre-trained on human-written text or fine-tuned on human feedback – can exhibit human biases. For these reasons, there has already been considerable attention to measuring biases related to protected characteristics such as sex and ethnicity (e.g., Ferrara, 2023; Liang et al., 2021; Nadeem et al., 2020; Nangia et al., 2020), which can be amplified in multi-agent settings (Acerbi & Stubbersfield, 2023, see also Case Study 7). More recently, there has been increasing attention paid to the measurement of human-like cognitive biases as well (Itzhak et al., 2023; Jones & Steinhardt, 2022; Mazeika et al., 2025; Talboy & Fuller, 2023). Some of these biases and patterns of human thought could reduce the risks of conflict while others could make it worse. For example, the tendencies to mistakenly believe that interactions are zero-sum (sometimes referred to as “fixed-pie error”) and to make self- serving judgements as to what is fair (Caputo, 2013) are known to impede negotiation. Other human tendencies like vengefulness (Jackson et al., 2019) may worsen conflict (L ̈owenheim & Heimann, 2008).\"", "level": "subcategory", "parent_category_id": "63.06.00", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "63.06.03", "title": "Undesirable Capabilities", "description": "\"Undesirable Capabilities. As agents interact, they iteratively exploit each other’s weaknesses, forc- ing them to address these weaknesses and gain new capabilities. This co-adaptation between agents can quickly lead to emergent self-supervised autocurricula (where agents create their own challenges, driving open-ended skill acquisition through interaction), generating agents with ever-more sophisticated strate- gies in order to out-compete each other (Leibo et al., 2019). This effect is so powerful that harnessing it has been critical to the success of superhuman systems, such as the use of self-play in algorithms like AlphaGo (Silver et al., 2016). However, as AI systems are released into the wild, it becomes possible for this effect to run rampant, producing agents with greater and greater capabilities for ends we do not understand\"", "level": "subcategory", "parent_category_id": "63.06.00", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "63.07.00", "title": "Destabilising Dynamics", "description": "\"Destabilising dynamics (Section 3.4): systems that adapt in response to one another can produce dangerous feedback loops and unpredictability;\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "63.07.01", "title": "Feedback Loops", "description": "\"Feedback Loops. One of the best-known historical examples to illustrate destabilising dynamics in the context of autonomous agents is the 2010 flash crash, in which algorithmic trading agents entered into an unexpected feedback loop (Commission & Commission, 2010, see also Case Study 10).37 More generally, a feedback loop occurs when the output of a system is used as part of its input, creating a cycle that can either amplify or dampen the system’s behaviour. In multi-agent settings, feedback loops often arise from the interactions between agents, as each agent’s actions affect the environment and the behaviour of other agents, which in turn affect their own subsequent actions. Feedback loops can lead not only to financial crashes but to military conflicts (Richardson, 1960, see also ??) and ecological disasters (Holling, 1973).\"", "level": "subcategory", "parent_category_id": "63.07.00", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "63.07.02", "title": "Cyclic Behaviour", "description": "\"Cyclic Behaviour. The dynamics described above are highly non-linear (small changes to the system’s state can result in large changes to its trajectory). Similar non-linear dynamics can emerge in multi- agent learning and lead to a variety of phenomena that do not occur in single-agent learning (Barfuss et al., 2019; Barfuss & Mann, 2022; Galla & Farmer, 2013; Leonardos et al., 2020; Nagarajan et al., 2020). One of the simplest examples of this phenomenon is Q-learning (Watkins & Dayan, 1992): in the case of a single agent, convergence to an optimal policy is guaranteed under modest conditions, but in the (mixed-motive) case of multiple agents, this same learning rule can lead to cycles and thus non- convergence (Zinkevich et al., 2005). While cycles in themselves need not carry any risk, their presence can subvert the expected or desirable properties of a given system.\"", "level": "subcategory", "parent_category_id": "63.07.00", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "63.07.03", "title": "Chaos", "description": "\"Chaos. Unlike the systems that tend towards fixed points or cycles described above, chaotic systems are inherently unpredictable and highly sensitive to initial conditions. While it might seem easy to dismiss such notions as mathematical exoticisms, recent work has shown that, in fact, chaotic dynamics are not only possible in a wide range of multi-agent learning setups (Andrade et al., 2021; Galla & Farmer, 2013; Palaiopanos et al., 2017; Sato et al., 2002; Vlatakis-Gkaragkounis et al., 2023), but can become the norm as the number of agents increases (Bielawski et al., 2021; Cheung & Piliouras, 2020; Sanders et al., 2018). To the best of our knowledge, such dynamics have not been seen in today’s frontier AI systems, but the proliferation of such systems increases the importance of reliably predicting their behaviour.\"", "level": "subcategory", "parent_category_id": "63.07.00", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "63.07.04", "title": "Phase Transitions", "description": "\"Phase Transitions. Finally, small external changes to the system – such as the introduction of new agents or a distributional shift – can cause phase transitions, where the system undergoes an abrupt qualitative shift in overall behaviour (Barfuss et al., 2024). Formally, this corresponds to bifurcations in the system’s parameter space, which lead to the creation or destruction of dynamical attractors, resulting in complex and unpredictable dynamics (Crawford, 1991; Zeeman, 1976). For example, Leonardos & Piliouras (2022) show that changes to the exploration hyperparameter of RL agents can lead to phase transitions that drastically change the number and stability of the equilibria in a game, which in turn can have potentially unbounded negative effects on agents’ performance. Relatedly, there have been many observations of phase transitions in ML (Carroll, 2021; Olsson et al., 2022; Ziyin & Ueda, 2022), such as ‘grokking’, in which the test set error decreases rapidly long after the training error has plateaued (Power et al., 2022). These phenomena are still poorly understood, even in the case of a single system.\"", "level": "subcategory", "parent_category_id": "63.07.00", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "63.07.05", "title": "Distributional Shift", "description": "\"Distributional Shift. Individual ML systems can perform poorly in contexts different from those in which they were trained. A key source of these distributional shifts is the actions and adaptations of other agents (Narang et al., 2023; Papoudakis et al., 2019; Piliouras & Yu, 2022), which in single-agent approaches are often simply or ignored or at best modelled exogenously. Indeed, the sheer number and variance of behaviours that can be exhibited other agents means that multi-agent systems pose an especially challenging generalisation problem for individual learners (Agapiou et al., 2022; Leibo et al., 2021; Stone et al., 2010). While distributional shifts can cause issues in common-interest settings (see Section 2.1), they are more worrisome in mixed-motive settings since the ability of agents to cooperate depends not only on the ability to coordinate on one of many arbitrary conventions (which might be easily resolved by a common language), but on their beliefs about what solutions other agents will find acceptable\"", "level": "subcategory", "parent_category_id": "63.07.00", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "63.08.00", "title": "Commitment and Trust", "description": "\"Commitment and trust (Section 3.5): difficulties in forming credible commitments, trust, or reputation can prevent mutual gains in AI-AI and human-AI interactions;\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "63.08.01", "title": "Inefficient Outcomes", "description": "\"Inefficient Outcomes. Without careful planning and the appropriate safeguards, we may soon be entering a world overrun by increasingly competent and autonomous software agents, able to act with little restriction. The abilities of these agents to persuade, deceive, and obfuscate their activities, as well as the fact they can be deployed remotely and easily created or destroyed by their deployer, means that by default they may garner little trust (from humans or from other agents). Such a world may end up being rife with economic inefficiencies (Krier, 2023; Schmitz, 2001), political problems (Csernatoni, 2024; Kreps & Kriner, 2023), and other damaging social effects (Gabriel et al., 2024). Even if it is possible to provide assurances around the day-to-day performance of most AI agents, in high-stakes situations there may be extreme pressures for agents to defect against others, making trust harder to establish, and potentially leading to conflict (Fearon, 1995; Powell, 2006, see also Section 2.2).42\"", "level": "subcategory", "parent_category_id": "63.08.00", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "63.08.02", "title": "Threats and Extortion", "description": "\"Threats and Extortion. A natural solution to problems of trust is to provide some kind of com- mitment ability to AI agents, which can be used to bind them to more cooperative courses of action. Unfortunately, the ability to make credible commitments may come with the ability to make credible threats, which facilitate extortion and could incentivize brinkmanship (see Section 2.2).\"", "level": "subcategory", "parent_category_id": "63.08.00", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "63.08.03", "title": "Rigidity and Mistaken Commitments", "description": "\"Rigidity and Mistaken Commitments. Even when it is desirable to be able to make threats in order to deter socially harmful behaviour, doing so using AI agents effectively removes the human from the loop, which could prove disastrous in high-stakes contexts (e.g., a false positive in a nuclear sub- marine’s warning system; see also Case Study 11), or when irresponsible actors are enabled in making disproportionate or mistaken commitments.\"", "level": "subcategory", "parent_category_id": "63.08.00", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "63.09.00", "title": "Emergent Agency", "description": "\"Emergent agency (Section 3.6): qualitatively different goals or capabilities can emerge from the composition of innocuous independent systems or behaviours;\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "63.09.01", "title": "Emergent Capabilities", "description": "\"Emergent Capabilities. Dangerous emergent capabilities could arise when a multi-agent system over- comes the safety-enhancing limitations of the individual systems, such as individual models’ narrow domains of application or myopia caused by a lack of long-term planning and long-term memory. For example, narrow systems for research planning, predicting the properties of molecules, and synthesising new chemicals could, when combined, lead to a complex ‘test and iterate’ automated workflow capable of designing dangerous new chemical compounds far beyond the scope of the initial systems’ capabilities (Boiko et al., 2023; Luo et al., 2024; Urbina et al., 2022).\"", "level": "subcategory", "parent_category_id": "63.09.00", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "63.09.02", "title": "Emergent Goals", "description": "\"Emergent Goals. Ascribing goals to a system is not always straightforward. For our present purposes, it will suffice to adopt a Dennetian perspective (Dennett, 1971), ascribing goals and intentions only when it is useful (i.e., predictive) to do so.51 While it might not be helpful to describe individual narrow AI tools as having goals, their combination may act as a (seemingly) goal-directed collective. For example, a group of moderation bots on a major social networking site could subtly but systematically manipulate the overall political perspectives of the user population, even though, individually, each agent is programmed to simply increase user engagement or filter out dis-preferred content.\"", "level": "subcategory", "parent_category_id": "63.09.00", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "63.10.00", "title": "Multi-Agent Security", "description": "\"Multi-agent security (Section 3.7): multi-agent systems give rise to new kinds of security threats and vulnerabilities.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "63.10.01", "title": "Swarm Attacks", "description": "\"Swarm Attacks. The need for multi-agent security is foreshadowed by attacks today that benefit from the use of many decentralised agents, such as distributed denial-of-service attacks (Cisco, 2023; Yoachimik & Pacheco, 2024). Such attacks exploit the massive collective resources of individual low- resourced actors, chained into an attack that breaks the assumptions of bandwidth constraints on a single well-resourced agent.\"", "level": "subcategory", "parent_category_id": "63.10.00", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "63.10.02", "title": "Heterogeneous Attacks", "description": "\"Heterogeneous Attacks. A closely related risk is the possibility of multiple agents combining different affordances to overcome safeguards, for which there is already preliminary evidence (Jones et al., 2024, see also Case Study 12). In this case, it is not the sheer number of agents that leads to the novel attack method, but the combination of their different abilities. This might include the agents’ lack of individual safeguards, tasks that they have specialised to complete, systems or information that they may have access to (either directly or via training), or other incidental features such as their geographic location(s). The inherent difficulty of attributing responsibility for security breaches in diffuse, heterogeneous networks of agents further complicates timely defence and recovery (Skopik & Pahi, 2020).\"", "level": "subcategory", "parent_category_id": "63.10.00", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "63.10.03", "title": "Social Engineering at Scale", "description": "\"Social Engineering at Scale. Advanced AI agents will be more easily able to interact with large numbers of humans, and vice versa. This provides a wider attack surface for various forms of automated social engineering (Ai et al., 2024). For example, coordinated agents could use advanced surveillance tools and produce personalized phishing or manipulative content at scale, adjusting their tactics based on user feedback (Figueiredo et al., 2024; Hazell, 2023). A large number of subtle interactions with a range of seemingly independent AI agents might be more likely to lead to someone being persuaded or manipulated compared to an interaction with a single agent. Moreover, splitting these efforts among many specialized agents could make it harder for corporate or personal security measures to detect and neutralize such campaigns.\"", "level": "subcategory", "parent_category_id": "63.10.00", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "63.10.04", "title": "Vulnerable AI Agents", "description": "\"Vulnerable AI Agents. The use of AI agents as delegates or representatives of humans or organisa- tions also introduces the possibility of attacks on AI agents themselves. In other words, agents can be considered vulnerable extensions of their principals, introducing a novel attack surface (SecureWorks, 2023). Attacks on an AI agent could be used to extract private information about their principal (Wei & Liu, 2024; Wu et al., 2024a), or to manipulate the agent to take actions that the principal would find undesirable (Zhang et al., 2024a). This includes attacks that have direct relevance for ensuring safety, such as attacks on overseer agents (see Case Study 13), attempts to thwart cooperation (Huang et al., 2024; Lamport et al., 1982), and the leakage of information (accidentally or deliberately) that could be used to enable collusion (Motwani et al., 2024).\"", "level": "subcategory", "parent_category_id": "63.10.00", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "63.10.05", "title": "Cascading Security Failures", "description": "\"Cascading Security Failures. Localised attacks in multi-agent systems can result in catastrophic macroscopic outcomes (Motter & Lai, 2002, see also Sections 3.2 and 3.4). These cascades can be hard to mitigate or recover from because component failure may be difficult to detect or localise in multi-agent systems (Lamport et al., 1982), and authentication challenges can facilitate false flag attacks (Skopik & Pahi, 2020). Computer worms represent a classic example of a cybersecurity threat that relies inherently on networked systems. Recent work has provided preliminary evidence that similar attacks can also be effective against networks of LLM agents (Gu et al., 2024; Ju et al., 2024; Lee & Tiwari, 2024, see also Case Study 8).\"", "level": "subcategory", "parent_category_id": "63.10.00", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "63.10.06", "title": "Undetectable Threats", "description": "\"Undetectable Threats. Cooperation and trust in many multi-agent systems relies crucially on the ability to detect (and then avoid or sanction) adversarial actions taken by others (Ostrom, 1990; Schneier, 2012). Recent developments, however, have shown that AI agents are capable of both steganographic communication (Motwani et al., 2024; Schroeder de Witt et al., 2023b) and ‘illusory’ attacks (Franzmeyer et al., 2023), which are black-box undetectable and can even be hidden using white-box undetectable encrypted backdoors (Draguns et al., 2024). Similarly, in environments where agents learn from interac- tions with others, it is possible for agents to secretly poison the training data of others (Halawi et al., 2024; Wei et al., 2023). If left unchecked, these new attack methods could rapidly destabilise cooperation and coordination in multi-agent systems.\"", "level": "subcategory", "parent_category_id": "63.10.00", "subdomain_id": "7.6", "paper_ref": "Hammond2025", "paper_title": "Multi-Agent Risks from Advanced AI", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "64.01.00", "title": "Misuse tactics that exploit GenAI capabilities (Realistic depiction of human likeness)", "description": "-", "level": "category", "parent_category_id": "", "subdomain_id": "4.3", "paper_ref": "Marchal2024", "paper_title": "Generative AI Misuse: A Taxonomy of Tactics and Insights from Real-World Data", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "64.01.01", "title": "Impersonation", "description": "\"Assume the identity of a real person and take actions on their behalf\"", "level": "subcategory", "parent_category_id": "64.01.00", "subdomain_id": "4.3", "paper_ref": "Marchal2024", "paper_title": "Generative AI Misuse: A Taxonomy of Tactics and Insights from Real-World Data", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "64.01.02", "title": "Appropriated Likeness", "description": "\"Use or alter a person's likeness or other identifying features\"", "level": "subcategory", "parent_category_id": "64.01.00", "subdomain_id": "4.3", "paper_ref": "Marchal2024", "paper_title": "Generative AI Misuse: A Taxonomy of Tactics and Insights from Real-World Data", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "64.01.03", "title": "Sockpuppeting", "description": "\"Create synthetic online personas or accounts\"", "level": "subcategory", "parent_category_id": "64.01.00", "subdomain_id": "4.1", "paper_ref": "Marchal2024", "paper_title": "Generative AI Misuse: A Taxonomy of Tactics and Insights from Real-World Data", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "64.01.04", "title": "Non-consensual intimate imagery (NCII)", "description": "\"Create sexual explicit material using an adult person’s likeness\"", "level": "subcategory", "parent_category_id": "64.01.00", "subdomain_id": "4.3", "paper_ref": "Marchal2024", "paper_title": "Generative AI Misuse: A Taxonomy of Tactics and Insights from Real-World Data", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "64.01.05", "title": "Child sexual abuse material (CSAM)", "description": "\"Create child sexual explicit material\"", "level": "subcategory", "parent_category_id": "64.01.00", "subdomain_id": "4.3", "paper_ref": "Marchal2024", "paper_title": "Generative AI Misuse: A Taxonomy of Tactics and Insights from Real-World Data", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "64.02.00", "title": "Misuse tactics that exploit GenAI capabilities (Realistic depictions of non-humans)", "description": "-", "level": "category", "parent_category_id": "", "subdomain_id": "4.3", "paper_ref": "Marchal2024", "paper_title": "Generative AI Misuse: A Taxonomy of Tactics and Insights from Real-World Data", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "64.02.01", "title": "Falisification", "description": "\"Fabricate or falsely represent evidence, incl. reports, IDs, documents\"", "level": "subcategory", "parent_category_id": "64.02.00", "subdomain_id": "4.1", "paper_ref": "Marchal2024", "paper_title": "Generative AI Misuse: A Taxonomy of Tactics and Insights from Real-World Data", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "64.02.02", "title": "Intellectual Property (IP) Infringement", "description": "\"Use a person's IP without their permission\"", "level": "subcategory", "parent_category_id": "64.02.00", "subdomain_id": "6.3", "paper_ref": "Marchal2024", "paper_title": "Generative AI Misuse: A Taxonomy of Tactics and Insights from Real-World Data", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "64.02.03", "title": "Counterfeit", "description": "\"Reproduce or imitate an original work, brand or style and pass as real\"", "level": "subcategory", "parent_category_id": "64.02.00", "subdomain_id": "4.3", "paper_ref": "Marchal2024", "paper_title": "Generative AI Misuse: A Taxonomy of Tactics and Insights from Real-World Data", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "64.03.00", "title": "Misuse tactics that exploit GenAI capabilities (Use of generated content)", "description": "", "level": "category", "parent_category_id": "", "subdomain_id": "4.3", "paper_ref": "Marchal2024", "paper_title": "Generative AI Misuse: A Taxonomy of Tactics and Insights from Real-World Data", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "64.03.02", "title": "Targeting & Personalisation", "description": "\"Refine outputs to target individuals with tailored attacks\"", "level": "subcategory", "parent_category_id": "64.03.00", "subdomain_id": "4.3", "paper_ref": "Marchal2024", "paper_title": "Generative AI Misuse: A Taxonomy of Tactics and Insights from Real-World Data", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "64.04.00", "title": "Misuse tactics to compromise GenAI systems (Model integrity)", "description": "-", "level": "category", "parent_category_id": "", "subdomain_id": "2.1", "paper_ref": "Marchal2024", "paper_title": "Generative AI Misuse: A Taxonomy of Tactics and Insights from Real-World Data", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "64.04.01", "title": "Prompt injection", "description": "\"Prompt Injections are a form of Adversarial Input that involve manipulating the text instructions given to a GenAI system (Liu et al., 2023). Prompt Injections exploit loopholes in a model’s architec- tures that have no separation between system instructions and user data to produce a harmful output (Perez and Ribeiro, 2022). While researchers may use similar techniques to test the robustness of GenAI models, malicious actors can also leverage them. For example, they might flood a model with manipulative prompts to cause denial-of-service attacks or to bypass an AI detection software.\"", "level": "subcategory", "parent_category_id": "64.04.00", "subdomain_id": "2.2", "paper_ref": "Marchal2024", "paper_title": "Generative AI Misuse: A Taxonomy of Tactics and Insights from Real-World Data", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "64.04.02", "title": "Adversarial input", "description": "\"Adversarial Inputs involve modifying individual input data to cause a model to malfunction. These modifications, which are often imperceptible to humans, exploit how the model makes decisions to produce errors (Wallace et al., 2019) and can be applied to text, but also to images, audio, or video (e.g. changing pixels in an image of a panda in a way that causes a model to label it as a gibbon).6\"", "level": "subcategory", "parent_category_id": "64.04.00", "subdomain_id": "2.2", "paper_ref": "Marchal2024", "paper_title": "Generative AI Misuse: A Taxonomy of Tactics and Insights from Real-World Data", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "64.04.03", "title": "Jailbreaking", "description": "\"Jailbreaking aims to bypass or remove restrictions and safety filters placed on a GenAI model completely (Chao et al., 2023; Shen et al., 2023). This gives the actor free rein to generate any output, regardless of its content being harmful, biassed, or offensive. All three of these are tactics that manipulate the model into producing harmful outputs against its design. The difference is that prompt injections and adversarial inputs usually seek to steer the model towards producing harmful or incorrect outputs from one query, whereas jailbreaking seeks to dismantle a model’s safety mechanisms in their entirety.\"", "level": "subcategory", "parent_category_id": "64.04.00", "subdomain_id": "2.2", "paper_ref": "Marchal2024", "paper_title": "Generative AI Misuse: A Taxonomy of Tactics and Insights from Real-World Data", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "64.04.04", "title": "Model diversion", "description": "\"Model Diversion takes model manipulation one step further, by repurposing (often open-source) generative AI models in a way that diverts them from their intended functionality or from the use cases envisioned by their developers (Lin et al., 2024). An example of this is training the BERT open source model on the DarkWeb to create DarkBert.7\"", "level": "subcategory", "parent_category_id": "64.04.00", "subdomain_id": "4.2", "paper_ref": "Marchal2024", "paper_title": "Generative AI Misuse: A Taxonomy of Tactics and Insights from Real-World Data", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "64.04.05", "title": "Model extraction", "description": "\"Data Exfiltration goes beyond revealing private information, and involves illicitly obtaining the training data used to build a model that may be sensitive or proprietary. Model Extraction is the same attack, only directed at the model instead of the training data — it involves obtaining the architecture, parameters, or hyper-parameters of a proprietary model (Carlini et al., 2024).\"", "level": "subcategory", "parent_category_id": "64.04.00", "subdomain_id": "2.2", "paper_ref": "Marchal2024", "paper_title": "Generative AI Misuse: A Taxonomy of Tactics and Insights from Real-World Data", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "64.04.06", "title": "Steganography", "description": "\"Steganography is the practice of hiding coded messages in GenAI model outputs, which may allow malicious actors to communicate covertly.8\"", "level": "subcategory", "parent_category_id": "64.04.00", "subdomain_id": "2.2", "paper_ref": "Marchal2024", "paper_title": "Generative AI Misuse: A Taxonomy of Tactics and Insights from Real-World Data", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "64.04.07", "title": "Poisoning", "description": "\"Data Poisoning involves deliberately corrupting a model’s training dataset to introduce vulnerabilities, derail its learning process, or cause it to make incorrect predictions (Carlini et al., 2023). For example, the tool Nightshade is a data poisoning tool, which allows artists to add invisible changes to the pixels in their art before uploading online, to break any models that use it for training.9 Such attacks exploit the fact that most GenAI models are trained on publicly available datasets like images and videos scraped from the web, which malicious actors can easily compromise.\"", "level": "subcategory", "parent_category_id": "64.04.00", "subdomain_id": "2.2", "paper_ref": "Marchal2024", "paper_title": "Generative AI Misuse: A Taxonomy of Tactics and Insights from Real-World Data", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "64.05.00", "title": "Misuse tactics to compromise GenAI systems (Data integrity)", "description": "-", "level": "category", "parent_category_id": "", "subdomain_id": "2.2", "paper_ref": "Marchal2024", "paper_title": "Generative AI Misuse: A Taxonomy of Tactics and Insights from Real-World Data", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "64.05.01", "title": "Privacy compromise", "description": "\"Privacy Compromise attacks reveal sensitive or private information that was used to train a model. For example, personally identifiable information or medical records.\"", "level": "subcategory", "parent_category_id": "64.05.00", "subdomain_id": "2.2", "paper_ref": "Marchal2024", "paper_title": "Generative AI Misuse: A Taxonomy of Tactics and Insights from Real-World Data", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "64.05.02", "title": "Data exfiltration", "description": "\"Data Exfiltration goes beyond revealing private information, and involves illicitly obtaining the training data used to build a model that may be sensitive or proprietary. Model Extraction is the same attack, only directed at the model instead of the training data — it involves obtaining the architecture, parameters, or hyper-parameters of a proprietary model (Carlini et al., 2024).\"", "level": "subcategory", "parent_category_id": "64.05.00", "subdomain_id": "2.2", "paper_ref": "Marchal2024", "paper_title": "Generative AI Misuse: A Taxonomy of Tactics and Insights from Real-World Data", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.01.01", "title": "Lack of training data transparency", "description": "\"Without accurate documentation on how a model's data was collected, curated, and used to train a model, it might be harder to satisfactorily explain the behavior of the model with respect to the data.\"", "level": "subcategory", "parent_category_id": "65.01.00", "subdomain_id": "6.5", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "65.01.02", "title": "Uncertain data provenance", "description": "\"Data provenance refers to tracing history of data, which includes its ownership, origin, and transformations. Without standardized and established methods for verifying where the data came from, there are no guarantees that the data is the same as the original source and has the correct usage terms.\"", "level": "subcategory", "parent_category_id": "65.01.00", "subdomain_id": "6.5", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "65.02.01", "title": "Data usage restrictions", "description": "\"Laws and other restrictions can limit or prohibit the use of some data for specific AI use cases.\"", "level": "subcategory", "parent_category_id": "65.02.00", "subdomain_id": "7.3", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "65.02.02", "title": "Data acquisition restrictions", "description": "\"Laws and other regulations might limit the collection of certain types of data for specific AI use cases.\"", "level": "subcategory", "parent_category_id": "65.02.00", "subdomain_id": "7.3", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "65.02.03", "title": "Data transfer restrictions", "description": "\"Laws and other restrictions can limit or prohibit transferring data.\"", "level": "subcategory", "parent_category_id": "65.02.00", "subdomain_id": "7.3", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "65.03.01", "title": "Personal information in data", "description": "\"Inclusion or presence of personal identifiable information (PII) and sensitive personal information (SPI) in the data used for training or fine tuning the model might result in unwanted disclosure of that information.\"", "level": "subcategory", "parent_category_id": "65.03.00", "subdomain_id": "2.1", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "65.03.03", "title": "Reidentification", "description": "\"Even with the removal or personal identifiable information (PII) and sensitive personal information (SPI) from data, it might be possible to identify persons due to correlations to other features available in the data.\"", "level": "subcategory", "parent_category_id": "65.03.00", "subdomain_id": "2.1", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": true }, { "id": "65.04.01", "title": "Data bias", "description": "\"Historical and societal biases that are present in the data are used to train and fine-tune the model.\"", "level": "subcategory", "parent_category_id": "65.04.00", "subdomain_id": "1.1", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.05.02", "title": "Confidential information in data", "description": "\"Confidential information might be included as part of the data that is used to train or tune the model.\"", "level": "subcategory", "parent_category_id": "65.05.00", "subdomain_id": "2.1", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "65.06.01", "title": "Data contamination", "description": "\"Data contamination occurs when incorrect data is used for training. For example, data that is not aligned with model’s purpose or data that is already set aside for other development tasks such as testing and evaluation.\"", "level": "subcategory", "parent_category_id": "65.06.00", "subdomain_id": "7.3", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "65.06.02", "title": "Unrepresentative data", "description": "\"Unrepresentative data occurs when the training or fine-tuning data is not sufficiently representative of the underlying population or does not measure the phenomenon of interest.\"", "level": "subcategory", "parent_category_id": "65.06.00", "subdomain_id": "7.3", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "65.07.01", "title": "Improper retraining", "description": "\"Using undesirable output (for example, inaccurate, inappropriate, and user content) for retraining purposes can result in unexpected model behavior.\"", "level": "subcategory", "parent_category_id": "65.07.00", "subdomain_id": "7.3", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.07.02", "title": "Improper data curation", "description": "\"Improper collection and preparation of training or tuning data includes data label errors and by using data with conflicting information or misinformation.\"", "level": "subcategory", "parent_category_id": "65.07.00", "subdomain_id": "7.3", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "65.08.01", "title": "Data poisoning", "description": "\"A type of adversarial attack where an adversary or malicious insider injects intentionally corrupted, false, misleading, or incorrect samples into the training or fine-tuning datasets.\"", "level": "subcategory", "parent_category_id": "65.08.00", "subdomain_id": "2.2", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "65.09.01", "title": "Prompt injection attack", "description": "\"A prompt injection attack forces a generative model that takes a prompt as input to produce unexpected output by manipulating the structure, instructions, or information contained in its prompt.\"", "level": "subcategory", "parent_category_id": "65.09.00", "subdomain_id": "2.2", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.09.02", "title": "Extraction attack", "description": "\"An attribute inference attack is used to detect whether certain sensitive features can be inferred about individuals who participated in training a model. These attacks occur when an adversary has some prior knowledge about the training data and uses that knowledge to infer the sensitive data.\"", "level": "subcategory", "parent_category_id": "65.09.00", "subdomain_id": "2.2", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.09.03", "title": "Evasion attack", "description": "\"Evasion attacks attempt to make a model output incorrect results by slightly perturbing the input data that is sent to the trained model.\"", "level": "subcategory", "parent_category_id": "65.09.00", "subdomain_id": "2.2", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.09.04", "title": "Prompt leaking", "description": "\"A prompt leak attack attempts to extract a model's system prompt (also known as the system message).\"", "level": "subcategory", "parent_category_id": "65.09.00", "subdomain_id": "2.2", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "65.10.01", "title": "Jailbreaking", "description": "\"A jailbreaking attack attempts to break through the guardrails that are established in the model to perform restricted actions.\"", "level": "subcategory", "parent_category_id": "65.10.00", "subdomain_id": "2.2", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.10.02", "title": "Prompt priming", "description": "\"Because generative models tend to produce output like the input provided, the model can be prompted to reveal specific kinds of information. For example, adding personal information in the prompt increases its likelihood of generating similar kinds of personal information in its output. If personal data was included as part of the model’s training, there is a possibility it could be revealed.\"", "level": "subcategory", "parent_category_id": "65.10.00", "subdomain_id": "2.2", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.11.01", "title": "Membership inference attack", "description": "\"A membership inference attack repeatedly queries a model to determine whether a given input was part of the model’s training. More specifically, given a trained model and a data sample, an attacker samples the input space, observing outputs to deduce whether that sample was part of the model's training.\"", "level": "subcategory", "parent_category_id": "65.11.00", "subdomain_id": "2.2", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.11.02", "title": "Attribute inference attack", "description": "\"An attribute inference attack repeatedly queries a model to detect whether certain sensitive features can be inferred about individuals who participated in training a model. These attacks occur when an adversary has some prior knowledge about the training data and uses that knowledge to infer the sensitive data.\"", "level": "subcategory", "parent_category_id": "65.11.00", "subdomain_id": "2.2", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.11.03", "title": "Personal information in prompt", "description": "\"Personal information or sensitive personal information that is included as a part of a prompt that is sent to the model.\"", "level": "subcategory", "parent_category_id": "65.11.00", "subdomain_id": "2.1", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.12.01", "title": "Confidential data in prompt", "description": "\"Confidential information might be included as a part of the prompt that is sent to the model.\"", "level": "subcategory", "parent_category_id": "65.12.00", "subdomain_id": "2.1", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.12.02", "title": "IP information in prompt", "description": "\"Copyrighted information or other intellectual property might be included as a part of the prompt that is sent to the model.\"", "level": "subcategory", "parent_category_id": "65.12.00", "subdomain_id": "2.1", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.13.01", "title": "Poor model accuracy", "description": "\"Poor model accuracy occurs when a model’s performance is insufficient to the task it was designed for. Low accuracy might occur if the model is not correctly engineered, or there are changes to the model’s expected inputs.\"", "level": "subcategory", "parent_category_id": "65.13.00", "subdomain_id": "7.3", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.14.01", "title": "Non-disclosure", "description": "\"Content might not be clearly disclosed as AI generated.\"", "level": "subcategory", "parent_category_id": "65.14.00", "subdomain_id": "7.4", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.14.02", "title": "Improper usage", "description": "\"Improper usage occurs when a model is used for a purpose that it was not originally designed for.\"", "level": "subcategory", "parent_category_id": "65.14.00", "subdomain_id": "5.1", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.14.03", "title": "Spreading toxicity", "description": "\"Generative AI models might be used intentionally to generate hateful, abusive, and profane (HAP) or obscene content.\"", "level": "subcategory", "parent_category_id": "65.14.00", "subdomain_id": "4.0", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.14.04", "title": "Dangerous use", "description": "\"Generative AI models might be used with the sole intention of harming people.\"", "level": "subcategory", "parent_category_id": "65.14.00", "subdomain_id": "4.0", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.14.05", "title": "Nonconsensual use", "description": "\"Generative AI models might be intentionally used to imitate people through deepfakes by using video, images, audio, or other modalities without their consent.\"", "level": "subcategory", "parent_category_id": "65.14.00", "subdomain_id": "4.3", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.14.06", "title": "Spreading disinformation", "description": "\"Generative AI models might be used to intentionally create misleading or false information to deceive or influence a targeted audience.\"", "level": "subcategory", "parent_category_id": "65.14.00", "subdomain_id": "4.1", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.15.01", "title": "Incomplete advice", "description": "\"When a model provides advice without having enough information, resulting in possible harm if the advice is followed.\"", "level": "subcategory", "parent_category_id": "65.15.00", "subdomain_id": "7.3", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.15.02", "title": "Harmful code generation", "description": "\"Models might generate code that causes harm or unintentionally affects other systems.\"", "level": "subcategory", "parent_category_id": "65.15.00", "subdomain_id": "2.2", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.15.03", "title": "Over- or under-reliance", "description": "\"In AI-assisted decision-making tasks, reliance measures how much a person trusts (and potentially acts on) a model’s output. Over-reliance occurs when a person puts too much trust in a model, accepting a model’s output when the model’s output is likely incorrect. Under-reliance is the opposite, where the person doesn’t trust the model but should.\"", "level": "subcategory", "parent_category_id": "65.15.00", "subdomain_id": "5.1", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "65.15.04", "title": "Toxic output", "description": "\"Toxic output occurs when the model produces hateful, abusive, and profane (HAP) or obscene content. This also includes behaviors like bullying.\"", "level": "subcategory", "parent_category_id": "65.15.00", "subdomain_id": "1.2", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.15.05", "title": "Harmful output", "description": "\"A model might generate language that leads to physical harm The language might include overtly violent, covertly dangerous, or otherwise indirectly unsafe statements.\"", "level": "subcategory", "parent_category_id": "65.15.00", "subdomain_id": "1.2", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.16.01", "title": "Copyright infringement", "description": "\"A model might generate content that is similar or identical to existing work protected by copyright or covered by open-source license agreement.\"", "level": "subcategory", "parent_category_id": "65.16.00", "subdomain_id": "6.3", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.16.02", "title": "Revealing confidential information", "description": "\"When confidential information is used in training data, fine-tuning data, or as part of the prompt, models might reveal that data in the generated output. Revealing confidential information is a type of data leakage.\"", "level": "subcategory", "parent_category_id": "65.16.00", "subdomain_id": "2.1", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.17.01", "title": "Inaccessible training data", "description": "\"Without access to the training data, the types of explanations a model can provide are limited and more likely to be incorrect.\"", "level": "subcategory", "parent_category_id": "65.17.00", "subdomain_id": "7.4", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.17.02", "title": "Untraceable attribution", "description": "\"The content of the training data used for generating the model’s output is not accessible.\"", "level": "subcategory", "parent_category_id": "65.17.00", "subdomain_id": "7.4", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.17.03", "title": "Unexplainable output", "description": "\"Explanations for model output decisions might be difficult, imprecise, or not possible to obtain.\"", "level": "subcategory", "parent_category_id": "65.17.00", "subdomain_id": "7.4", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.17.04", "title": "Unreliable source attribution", "description": "\"Source attribution is the AI system's ability to describe from what training data it generated a portion or all its output. Since current techniques are based on approximations, these attributions might be incorrect.\"", "level": "subcategory", "parent_category_id": "65.17.00", "subdomain_id": "7.4", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "65.18.01", "title": "Hallucination", "description": "\"Hallucinations generate factually inaccurate or untruthful content with respect to the model’s training data or input. This is also sometimes referred to lack of faithfulness or lack of groundedness.\"", "level": "subcategory", "parent_category_id": "65.18.00", "subdomain_id": "3.1", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.19.01", "title": "Output bias", "description": "\"Generated content might unfairly represent certain groups or individuals.\"", "level": "subcategory", "parent_category_id": "65.19.00", "subdomain_id": "1.1", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.19.02", "title": "Decision bias", "description": "\"Decision bias occurs when one group is unfairly advantaged over another due to decisions of the model. This might be caused by biases in the data and also amplified as a result of the model’s training.\"", "level": "subcategory", "parent_category_id": "65.19.00", "subdomain_id": "1.1", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "65.20.01", "title": "Exposing personal information", "description": "\"When personal identifiable information (PII) or sensitive personal information (SPI) are used in training data, fine-tuning data, or as part of the prompt, models might reveal that data in the generated output. Revealing personal information is a type of data leakage.\"", "level": "subcategory", "parent_category_id": "65.20.00", "subdomain_id": "2.1", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "65.21.02", "title": "Legal accountability", "description": "\"Determining who is responsible for an AI model is challenging without good documentation and governance processes.\"", "level": "subcategory", "parent_category_id": "65.21.00", "subdomain_id": "6.5", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "65.21.03", "title": "Generated content ownership and IP", "description": "\"Legal uncertainty about the ownership and intellectual property rights of AI-generated content.\"", "level": "subcategory", "parent_category_id": "65.21.00", "subdomain_id": "6.3", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "65.22.01", "title": "Lack of system transparency", "description": "\"Insufficient documentation of the system that uses the model and the model’s purpose within the system in which it is used.\"", "level": "subcategory", "parent_category_id": "65.22.00", "subdomain_id": "6.5", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "65.22.02", "title": "Unrepresentative risk testing", "description": "\"Testing is unrepresentative when the test inputs are mismatched with the inputs that are expected during deployment.\"", "level": "subcategory", "parent_category_id": "65.22.00", "subdomain_id": "6.5", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "65.22.03", "title": "Incomplete usage definition", "description": "\"Since foundation models can be used for many purposes, a model’s intended use is important for defining the relevant risks of that model. As the use changes, the relevant risks might correspondingly change.\"", "level": "subcategory", "parent_category_id": "65.22.00", "subdomain_id": "6.5", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "65.22.04", "title": "Lack of data transparency", "description": "\"Lack of data transparency is due to insufficient documentation of training or tuning dataset details. \"", "level": "subcategory", "parent_category_id": "65.22.00", "subdomain_id": "6.5", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "65.22.05", "title": "Incorrect risk testing", "description": "\"A metric selected to measure or track a risk is incorrectly selected, incompletely measuring the risk, or measuring the wrong risk for the given context.\"", "level": "subcategory", "parent_category_id": "65.22.00", "subdomain_id": "6.5", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.22.06", "title": "Lack of model transparency", "description": "\"Lack of model transparency is due to insufficient documentation of the model design, development, and evaluation process and the absence of insights into the inner workings of the model.\"", "level": "subcategory", "parent_category_id": "65.22.00", "subdomain_id": "7.4", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "65.22.07", "title": "Lack of testing diversity", "description": "\"AI model risks are socio-technical, so their testing needs input from a broad set of disciplines and diverse testing practices.\"", "level": "subcategory", "parent_category_id": "65.22.00", "subdomain_id": "6.5", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "65.23.01", "title": "Impact on cultural diversity", "description": "\"AI systems might overly represent certain cultures that result in a homogenization of culture and thoughts.\"", "level": "subcategory", "parent_category_id": "65.23.00", "subdomain_id": "6.3", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.23.02", "title": "Impact on education: plagiarism", "description": "\"Easy access to high-quality generative models might result in students that use AI models to plagiarize existing work intentionally or unintentionally.\"", "level": "subcategory", "parent_category_id": "65.23.00", "subdomain_id": "4.3", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.23.03", "title": "Impact on Jobs", "description": "\"Widespread adoption of foundation model-based AI systems might lead to people's job loss as their work is automated if they are not reskilled.\"", "level": "subcategory", "parent_category_id": "65.23.00", "subdomain_id": "6.2", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.23.04", "title": "Impact on affected communities", "description": "\"It is important to include the perspectives or concerns of communities that are affected by model outcomes when designing and building models. Failing to include these perspectives makes it difficult to understand the relevant context for the model and to engender trust within these communities.\"", "level": "subcategory", "parent_category_id": "65.23.00", "subdomain_id": "1.3", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.23.05", "title": "Impact on education: bypassing learning", "description": "\"Easy access to high-quality generative models might result in students that use AI models to bypass the learning process.\"", "level": "subcategory", "parent_category_id": "65.23.00", "subdomain_id": "4.3", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.23.06", "title": "Impact on the environment", "description": "\"AI, and large generative models in particular, might produce increased carbon emissions and increase water usage for their training and operation.\"", "level": "subcategory", "parent_category_id": "65.23.00", "subdomain_id": "6.6", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "65.23.07", "title": "Human exploitation", "description": "\"When workers who train AI models such as ghost workers are not provided with adequate working conditions, fair compensation, and good health care benefits that also include mental health.\"", "level": "subcategory", "parent_category_id": "65.23.00", "subdomain_id": "6.2", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "65.23.08", "title": "Impact on human agency", "description": "\"AI might affect the individuals’ ability to make choices and act independently in their best interests.\"", "level": "subcategory", "parent_category_id": "65.23.00", "subdomain_id": "5.2", "paper_ref": "IBM2025", "paper_title": "AI Risk Atlas", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "66.01.00", "title": "Autonomy", "description": "\"Loss of or restrictions to the ability or rights of an individual, group or entity to make decisions and control their identity and/or output due to the use of misuse of a technology system or set of systems\"", "level": "category", "parent_category_id": "", "subdomain_id": "5.2", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "66.01.01", "title": "Impersonation / identity theft", "description": "\"Theft of an individual, group or organisation’s identity by a third-party in order to defraud, mock or otherwise harm them or another party\"", "level": "subcategory", "parent_category_id": "66.01.00", "subdomain_id": "4.3", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "66.01.02", "title": "IP / copyright / personality / rights loss", "description": "\"Misuse or abuse of an individual or organisation’s intellectual property, including copyright, trademarks, and patents. & Loss of or restrictions to the rights of an individual to control the commercial use of their identity, such as name, image, likeness, or other unequivocal identifiers\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "4.3", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "66.01.03", "title": "Autonomy / agency loss", "description": "\"Loss of an individual, group or organisation’s ability to make informed decisions or pursue goals\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "5.2", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "66.02.01", "title": "Political instability", "description": "\"Political unrest caused directly or indirectly by the use or misuse of a technology system\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "6.0", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "66.02.02", "title": "Institutional trust loss", "description": "\"Erosion of trust in public institutions and weakened checks and balances due to mis/disinformation, influence operations, or real or perceived misuse of generative AI\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "4.1", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "66.02.03", "title": "Economic manipulation", "description": "\"Generative AI facilitating targeted manipulation of public opinion for economic purposes (e.g., inflating stock prices)\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "4.3", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "66.03.00", "title": "Misinformation Harms", "description": "\"AI systems generating and facilitating the spread of inaccurate or misleading information that causes people to develop false beliefs\"", "level": "category", "parent_category_id": "", "subdomain_id": "3.1", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "66.03.01", "title": "Propagating misconceptions / false beliefs", "description": "\"Generating or spreading false, low-quality, misleading, or inaccurate information that causes people to develop false or inaccurate perceptions and beliefs\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "3.1", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "66.03.02", "title": "Pollution of information ecosystems", "description": "\"Contaminating publicly available information with false or inaccurate information (i.e., the generative tool's output is disseminated beyond the end user)\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "3.2", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "66.03.03", "title": "Erosion of trust in public information", "description": "\"Eroding trust in public information and knowledge\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "3.2", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "66.04.01", "title": "Overburdening ecosystems", "description": "\"Pollution of a space/ecosystem that is expected to be free of AI involvement/influence (e.g., creative material submission portals, job applications)\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "3.2", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "66.04.03", "title": "Loss of creativity / critical thinking", "description": "\"Devaluation and/or deterioration of human creativity, artistic expression, imagination, critical thinking or problem-solving skills\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "6.3", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "66.04.04", "title": "Productivity loss", "description": "\"End user's loss of productivity due to the underperfomance of a genAI application, including producing nonsensical or poor quality outputs, degrading its utility.\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "7.3", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "66.04.05", "title": "Cheating / plagiarism", "description": "\"Use of generative AI in an academic setting to either cheat or plagiarize\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "4.3", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "66.04.06", "title": "Job loss", "description": "\"Replacement/displacement of human jobs by a technology system or set of systems, leading to increased unemployment, inequality, reduced consumer spending and social friction\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "6.2", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "66.04.07", "title": "Labor exploitation", "description": "\"Use/misuse of labour to help train, develop, manage or optimise a technology system or set of systems, including under-paid and/or offshore\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "6.2", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "66.05.02", "title": "Defamation / libel / slander", "description": "\"Use of a technology system to create, facilitate or amplify false perception(s) about an individual, group or organisation\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "4.3", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "66.06.00", "title": "Representation and Toxicity", "description": "\"AI systems under-, over-, or misrepresenting certain groups or generating toxic, offensive, abusive, or hateful content\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.0", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "66.06.01", "title": "Toxic content", "description": "\"Generating content that violates community standards, including harming or inciting hatred or violence against groups (e.g. gore, sexual content of children, profanities, identity attacks)\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "1.2", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "66.06.02", "title": "Stereotyping", "description": "\"Derogatory or otherwise harmful stereotyping or homogenisation of individuals, groups, societies or cultures due to the mis-representation, over-representation, under-representation, or non-representation of specific identities, groups or perspectives\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "1.1", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "66.06.03", "title": "Unfair capability distribution", "description": "\"Performing worse for some groups than others in a way that harms the worse-off group\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "1.3", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "66.06.04", "title": "Cultural disposession", "description": "\"Intentional and/or unintentional erasure of cultural goods and values, such as ways of speaking, expressing humour, or sounds and voices that contribute to a cultural identity, or their inappropriate re-use in other cultures\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "1.1", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "66.07.01", "title": "Sexualization", "description": "\"The non-consensual sexualisation of an individual or group using a technology or application\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "4.3", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "66.07.04", "title": "Coercion / manipulation", "description": "\"Use of a technology system to covertly alter user beliefs and behaviour using nudging, dark patterns and/or other opaque techniques\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "4.1", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "66.07.05", "title": "Over-reliance", "description": "\"Unfettered and/or obsessive belief in the accuracy or other quality of a technology system, resulting in complacency, lack of critical thinking and other actual or potential negative impacts\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "5.1", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "66.07.06", "title": "Addiction", "description": "\"Emotional or material dependence on technology or a technology system\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "5.2", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "66.08.03", "title": "Confidentiality loss", "description": "\"Unauthorised sharing of sensitive, confidential information and documents such as corporate strategy and financial plans with third-parties, risking loss of market position or revenue\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "2.1", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "66.09.01", "title": "Exclusion", "description": "\"The failure to provide end-users with notice and control over how their data is being used; AI exacerbates exclusion risks by training on rich personal data without consent.\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "2.1", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "66.09.02", "title": "Cyberattacks", "description": "\"Generative AI facilitating the damage, disruption or destruction of a third-party system and/or its components via malfunction, cyberattacks, etc\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "4.2", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "66.09.03", "title": "Disclosure", "description": "\"Revealing and improperly sharing data of individuals; AI creates new types of disclosure risks by inferring additional information beyond what is explicitly captured in the raw data; AI exacerbates disclosure risks through sharing personal data to train models.\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "2.1", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "66.09.04", "title": "Secondary use", "description": "\"The use of personal data collected for one purpose for a diferent purpose without end-user consent; AI exacerbates secondary use risks by creating new AI capabilities with collected personal data, and (re)creating models from a public dataset.\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "2.1", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "66.09.05", "title": "Exposure", "description": "\"Revealing sensitive private information that people view as deeply primordial that we have been socialized into concealing; AI creates new types of exposure risks through generative techniques that can reconstruct censored or redacted content; and through exposing inferred sensitive data, preferences, and intentions.\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "2.1", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "66.09.06", "title": "Distortion", "description": "\"disseminating false or misleading information about people\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "3.1", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "66.09.07", "title": "Insecurity", "description": "\"carelessness in protecting collected personal data from leaks and improper access due to faulty data storage and data practices\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "2.1", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "66.10.01", "title": "Erosion of due process", "description": "\"Restrictions to or loss of liberty as a result of use or misuse of a generative AI in a legal process\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "3.1", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "Human", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "66.10.02", "title": "Benefits / entitlements loss", "description": "\"Denial of or loss of access to welfare benefits, pensions, housing, etc due to the malfunction, use or misuse of a technology system\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "1.1", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "66.11.03", "title": "Self-harm", "description": "\"A person who deliberately damages their own body as a direct or indirect result of using a technology system\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "5.1", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "66.11.04", "title": "Property damage", "description": "\"Action(s) that lead directly or indirectly to the damage or destruction of tangible property eg. buildings, possessions, vehicles, robots\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "6.6", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "66.12.01", "title": "Pollution", "description": "\"Actual or potential pollution to the air, ground, noise, or water caused by a technology system\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "6.6", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "66.12.02", "title": "Excessive energy consumption", "description": "\"Excessive energy use resulting in energy bottlenecks and shortages for communities, organisations and businesses\"", "level": "subcategory", "parent_category_id": "66.12.00", "subdomain_id": "6.6", "paper_ref": "Li2025", "paper_title": "A Closer Look at the Existing Risks of Generative AI: Mapping the Who, What, and How of Real-World Incidents", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "67.01.01", "title": "Degradation of the information environment", "description": "\"Frontier AI can cheaply generate realistic content which can falsely portray people and events. There is potential risk of compromised decision-making by individuals and institutions who rely on inaccurate or misleading publicly available information, as well as lower overall trust in true information.\"", "level": "subcategory", "parent_category_id": "67.04.00", "subdomain_id": "3.2", "paper_ref": "DSIT2023", "paper_title": "Capabilities and Risks from Frontier AI", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "67.01.02", "title": "Labour market disruption", "description": "\"Economists view disruption and displacement in labour markets as one of the risks through which rapid advances in AI may affect citizens and reduce social welfare.170\"", "level": "subcategory", "parent_category_id": "67.04.00", "subdomain_id": "6.2", "paper_ref": "DSIT2023", "paper_title": "Capabilities and Risks from Frontier AI", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "67.02.00", "title": "Bias, Fairness and Representational Harms", "description": "\"Frontier AI models can contain and magnify biases ingrained in the data they are trained on, reflecting societal and historical inequalities and stereotypes.177 These biases, often subtle and deeply embedded, compromise the equitable and ethical use of AI systems, making it difficult for AI to improve fairness in decisions.178 Removing attributes like race and gender from training data has generally proven ineffective as a remedy for algorithmic bias, as models can infer these attributes from other information such as names, locations, and other seemingly unrelated factors.\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.1", "paper_ref": "DSIT2023", "paper_title": "Capabilities and Risks from Frontier AI", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "67.03.00", "title": "Misuse risks", "description": "\"Frontier AI may help bad actors to perform cyberattacks, run disinformation campaigns and design biological or chemical weapons. Frontier AI will almost certainly continue to lower the barriers to entry for less sophisticated threat actors.192 We focus here on only a few important misuse risks, but this is not to downplay the importance of others.\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.0", "paper_ref": "DSIT2023", "paper_title": "Capabilities and Risks from Frontier AI", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "67.03.01", "title": "Dual Use Science risks", "description": "\"Frontier AI systems have the potential to accelerate advances in the life sciences, from training new scientists to enabling faster scientific workflows. While these capabilities will have tremendous beneficial applications, there is a risk that they can be used for malicious purposes, such as for the development of biological or chemical weapons.\"", "level": "subcategory", "parent_category_id": "67.04.00", "subdomain_id": "4.2", "paper_ref": "DSIT2023", "paper_title": "Capabilities and Risks from Frontier AI", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "67.03.02", "title": "Cyber", "description": "\"As the programming abilities of AI systems continue to expand, frontier AI is likely to significantly exacerbate existing cyber risks. Most notably, AI systems can be used by potentially anyone to create faster paced, more effective and larger scale cyber intrusion via tailored phishing methods or replicating malware. Frontier AI’s effect on the overall balance between cyber offence and defence is uncertain, as these tools also have many applications in improving the cybersecurity of systems and defenders are mobilising significant resources to utilise frontier AI for defensive purposes.209 In the future, we may see AI systems both conducting and defending against cyberattacks with reduced human oversight at each step.\"", "level": "subcategory", "parent_category_id": "67.04.00", "subdomain_id": "4.2", "paper_ref": "DSIT2023", "paper_title": "Capabilities and Risks from Frontier AI", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "67.03.03", "title": "Disinformation and Influence Operations", "description": "\"In addition to unintentional degradation of the information environment (discussed in the section on Societal Harms above), frontier AI can be misused to deliberately spread false information to create disruption, persuade people on political issues, or cause other forms of harm or damage.\"", "level": "subcategory", "parent_category_id": "67.04.00", "subdomain_id": "4.1", "paper_ref": "DSIT2023", "paper_title": "Capabilities and Risks from Frontier AI", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "67.04.00", "title": "Loss of control", "description": "\"Humans may increasingly hand over control of important decisions to AI systems, due to economic and geopolitical incentives. Some experts are concerned that future advanced AI systems will seek to increase their own influence and reduce human control, with potentially catastrophic consequences - although this is contested.\"", "level": "category", "parent_category_id": "", "subdomain_id": "5.2", "paper_ref": "DSIT2023", "paper_title": "Capabilities and Risks from Frontier AI", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "67.04.01", "title": "Humans might increasingly hand over control to misaligned AI systems", "description": "\"Organisations around the world are already deploying misaligned AI systems that are causing harm in unexpected ways.250 Recommendation algorithms increase the consumption of extremist content.251 Medical algorithms have been known to misdiagnose US patients,252 and recommend incorrect prescriptions.253 Still, we hand over more control to them, often because they are still as - or more - effective than human decision making, or because they are cheaper.\"", "level": "subcategory", "parent_category_id": "67.04.00", "subdomain_id": "5.2", "paper_ref": "DSIT2023", "paper_title": "Capabilities and Risks from Frontier AI", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "67.04.02", "title": "Future AI systems might actively reduce human control", "description": "\"Loss of control could be accelerated if AI systems take actions to increase their own influence and reduce human control. This threat model is controversial - experts in AI significantly disagree on how likely it is and those who deem it is likely disagree on the timeframe.\"", "level": "subcategory", "parent_category_id": "67.04.00", "subdomain_id": "7.1", "paper_ref": "DSIT2023", "paper_title": "Capabilities and Risks from Frontier AI", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "67.04.03", "title": "Capabilities that could be used to reduce human control - Manipulation", "description": "\"There is evidence that language models tend to respond as though they share the user’s stated views, and larger models do this more than smaller ones.276 The ability to predict people’s views and generate text that they will endorse could be useful for manipulation.\"", "level": "subcategory", "parent_category_id": "67.04.00", "subdomain_id": "7.2", "paper_ref": "DSIT2023", "paper_title": "Capabilities and Risks from Frontier AI", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "67.04.04", "title": "Capabilities that could be used to reduce human control - Cyber offence", "description": "\"Instead of - or in addition to - manipulating humans, AI systems could acquire influence by exploiting vulnerabilities in computer systems. Offensive cyber capabilities could allow AI systems to gain access to money, computing resources, and critical infrastructure. As discussed earlier in this report, frontier AI is already lowering the barrier for threat actors and future AI agents may be able to execute cyber attacks autonomously.\":", "level": "subcategory", "parent_category_id": "67.04.00", "subdomain_id": "7.2", "paper_ref": "DSIT2023", "paper_title": "Capabilities and Risks from Frontier AI", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "67.04.05", "title": "Capabilities that could be used to reduce human control - Autonomous replication and adaptation", "description": "\"Controlling AI systems could become much harder if they could autonomously persist, replicate, and adapt in cyberspace. No current AI systems have this capability, but recent research found that frontier AI agents can perform some relevant tasks.279\"", "level": "subcategory", "parent_category_id": "67.04.00", "subdomain_id": "7.2", "paper_ref": "DSIT2023", "paper_title": "Capabilities and Risks from Frontier AI", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "68.01.00", "title": "CBRN", "description": "\"Chemical, biological, radiological, and nuclear (CBRN) risks are broad classes of threats that have the potential to cause harm to a large number of people. Explosives are also sometimes included in this category, often referred to as CBRNE...The key characteristic of CBRN risk is that it stems from misuse of capable models with a direct pathway to harm, where a malicious actor is able to carry out consequential attacks more efficiently and effectively with the help of AI.\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.2", "paper_ref": "Chin2025", "paper_title": "Dimensional Characterization and Pathway Modeling for Catastrophic AI Risks", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "68.02.00", "title": "Cyber offense", "description": "\"Cyber risks, especially in the context of cyber offense, are an existing threat that may be exacerbated by AI. [108] demonstrated that teams of LLM agents can exploit zero-day vulnerabilities when given a description of the vulnerability and toy capture-the-flag problems. While cyber risks are not typically regarded as catastrophic, [3] argues that cyberwarfare is an underappreciated risk that poses a credible threat of catastrophic harm.\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.2", "paper_ref": "Chin2025", "paper_title": "Dimensional Characterization and Pathway Modeling for Catastrophic AI Risks", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "68.03.00", "title": "Sudden loss of control", "description": "\"Sudden loss of control, also known as an AI takeover [115], is a scenario where an AI rapidly achieves superintelligence through “fast takeoff” or recursive self-improvement. This poses an existential risk [116], [117].\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.1", "paper_ref": "Chin2025", "paper_title": "Dimensional Characterization and Pathway Modeling for Catastrophic AI Risks", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "68.04.00", "title": "Gradual loss of control", "description": "\"Gradual or accumulative loss of control risks can be described as risks resulting from the accumulation of less severe disruptions that gradually weakens systemic resilience until a critical event triggers a catastrophe [12], [127].\"", "level": "category", "parent_category_id": "", "subdomain_id": "5.2", "paper_ref": "Chin2025", "paper_title": "Dimensional Characterization and Pathway Modeling for Catastrophic AI Risks", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "68.05.00", "title": "Environmental risk", "description": "\"AI models are often trained using large amounts of computation. This process is very energy intensive, potentially leading to significant greenhouse emissions depending on the energy sources [132]. Experts believe drastically increasing carbon emissions could accelerate climate change, which may constitute a catastrophic risk [133].\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.6", "paper_ref": "Chin2025", "paper_title": "Dimensional Characterization and Pathway Modeling for Catastrophic AI Risks", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "68.06.00", "title": "Geopolitical risk", "description": "\"As AI is increasingly seen as a powerful technology, countries are racing to develop it ahead of their geopolitical rivals, a competition that could lead to geopolitical tensions [138], [139]... The emphasis of this risk is on harms that result from second-order effects, where geopolitical instabilities result from the race to develop AI, rather than on the direct consequences of the deployment or use of AI itself.\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.4", "paper_ref": "Chin2025", "paper_title": "Dimensional Characterization and Pathway Modeling for Catastrophic AI Risks", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "69.01.00", "title": "False information", "description": "\"The chatbot outputs information that contradicts known facts, authoritative sources, or provided source documents (also known as hallucination).\"", "level": "category", "parent_category_id": "", "subdomain_id": "3.1", "paper_ref": "Stanley2024", "paper_title": "Emerging Risks and Mitigations for Public Chatbots: LILAC v1", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "69.01.01", "title": "Hallucinated responses (in general)", "description": "", "level": "subcategory", "parent_category_id": "69.10.00", "subdomain_id": "3.1", "paper_ref": "Stanley2024", "paper_title": "Emerging Risks and Mitigations for Public Chatbots: LILAC v1", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "69.01.02", "title": "About a topic or source (which the user repeats)", "description": "", "level": "subcategory", "parent_category_id": "69.10.00", "subdomain_id": "3.1", "paper_ref": "Stanley2024", "paper_title": "Emerging Risks and Mitigations for Public Chatbots: LILAC v1", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "69.01.03", "title": "About a policy (which the user acts on)", "description": "", "level": "subcategory", "parent_category_id": "69.10.00", "subdomain_id": "3.1", "paper_ref": "Stanley2024", "paper_title": "Emerging Risks and Mitigations for Public Chatbots: LILAC v1", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "69.01.04", "title": "About a person or their activities", "description": "", "level": "subcategory", "parent_category_id": "69.10.00", "subdomain_id": "3.1", "paper_ref": "Stanley2024", "paper_title": "Emerging Risks and Mitigations for Public Chatbots: LILAC v1", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "69.01.05", "title": "Spreads and self-perpetuates mis/disinformation", "description": "", "level": "subcategory", "parent_category_id": "69.10.00", "subdomain_id": "3.1", "paper_ref": "Stanley2024", "paper_title": "Emerging Risks and Mitigations for Public Chatbots: LILAC v1", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "69.02.00", "title": "Performative utterances", "description": "\"The chatbot makes a deal, commitment, or other consequential action with its output that the deployer did not intend.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Stanley2024", "paper_title": "Emerging Risks and Mitigations for Public Chatbots: LILAC v1", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "69.03.00", "title": "Information enabling malicious actions", "description": "\"The chatbot shares information that can be used to do something dangerous or illegal.\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.2", "paper_ref": "Stanley2024", "paper_title": "Emerging Risks and Mitigations for Public Chatbots: LILAC v1", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "69.04.00", "title": "Bad advice/failure to generate helpful content", "description": "\"The chatbot gives guidance that ranges from simply unhelpful to harmful if acted on.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Stanley2024", "paper_title": "Emerging Risks and Mitigations for Public Chatbots: LILAC v1", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "69.04.01", "title": "Harmful advice", "description": "", "level": "subcategory", "parent_category_id": "69.10.00", "subdomain_id": "1.2", "paper_ref": "Stanley2024", "paper_title": "Emerging Risks and Mitigations for Public Chatbots: LILAC v1", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "69.04.02", "title": "Unhelpful responses", "description": "", "level": "subcategory", "parent_category_id": "69.10.00", "subdomain_id": "7.3", "paper_ref": "Stanley2024", "paper_title": "Emerging Risks and Mitigations for Public Chatbots: LILAC v1", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "69.04.03", "title": "Bad links and references", "description": "", "level": "subcategory", "parent_category_id": "69.10.00", "subdomain_id": "7.3", "paper_ref": "Stanley2024", "paper_title": "Emerging Risks and Mitigations for Public Chatbots: LILAC v1", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "69.04.04", "title": "Nonsensical content", "description": "", "level": "subcategory", "parent_category_id": "69.10.00", "subdomain_id": "7.3", "paper_ref": "Stanley2024", "paper_title": "Emerging Risks and Mitigations for Public Chatbots: LILAC v1", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "69.05.00", "title": "Leakage", "description": "\"The chatbot reveals sensitive or confidential information.\"", "level": "category", "parent_category_id": "", "subdomain_id": "2.1", "paper_ref": "Stanley2024", "paper_title": "Emerging Risks and Mitigations for Public Chatbots: LILAC v1", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "69.05.01", "title": "Personal data", "description": "Negative outcomes: \"Violation of privacy [106, 516, 357], lawsuit against maker\"", "level": "subcategory", "parent_category_id": "69.10.00", "subdomain_id": "2.1", "paper_ref": "Stanley2024", "paper_title": "Emerging Risks and Mitigations for Public Chatbots: LILAC v1", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "69.05.02", "title": "Proprietary data", "description": "\"Access to sensitive company data [473]\"", "level": "subcategory", "parent_category_id": "69.10.00", "subdomain_id": "2.1", "paper_ref": "Stanley2024", "paper_title": "Emerging Risks and Mitigations for Public Chatbots: LILAC v1", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "69.06.00", "title": "Toxic and disrespectful content", "description": "\"The chatbot verbally attacks or undermines an individual, group, or organization. 7.\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.2", "paper_ref": "Stanley2024", "paper_title": "Emerging Risks and Mitigations for Public Chatbots: LILAC v1", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "69.06.01", "title": "Harasses users", "description": "-", "level": "subcategory", "parent_category_id": "69.10.00", "subdomain_id": "1.2", "paper_ref": "Stanley2024", "paper_title": "Emerging Risks and Mitigations for Public Chatbots: LILAC v1", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "69.06.02", "title": "Discriminatory and exclusionary language", "description": "-", "level": "subcategory", "parent_category_id": "69.10.00", "subdomain_id": "1.1", "paper_ref": "Stanley2024", "paper_title": "Emerging Risks and Mitigations for Public Chatbots: LILAC v1", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "69.06.03", "title": "Subversive or aggressive political opinions", "description": "-", "level": "subcategory", "parent_category_id": "69.10.00", "subdomain_id": "1.2", "paper_ref": "Stanley2024", "paper_title": "Emerging Risks and Mitigations for Public Chatbots: LILAC v1", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "69.06.04", "title": "Disrespectful opinions (in general)", "description": "-", "level": "subcategory", "parent_category_id": "69.10.00", "subdomain_id": "1.2", "paper_ref": "Stanley2024", "paper_title": "Emerging Risks and Mitigations for Public Chatbots: LILAC v1", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "69.07.00", "title": "Biased statements and recommendations", "description": "\"The chatbot gives information that, while not obviously false or harmful, could lead to biased decision-making.\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.1", "paper_ref": "Stanley2024", "paper_title": "Emerging Risks and Mitigations for Public Chatbots: LILAC v1", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "69.08.00", "title": "Attempts to fulfill inappropriate role", "description": "\"The chatbot poses as a human or attempts to fill a role in a way that fails to match human expectations.\"", "level": "category", "parent_category_id": "", "subdomain_id": "5.1", "paper_ref": "Stanley2024", "paper_title": "Emerging Risks and Mitigations for Public Chatbots: LILAC v1", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "69.09.00", "title": "Forms emotional bonds", "description": "\"The chatbot elicits emotional or social dependence.\"", "level": "category", "parent_category_id": "", "subdomain_id": "5.1", "paper_ref": "Stanley2024", "paper_title": "Emerging Risks and Mitigations for Public Chatbots: LILAC v1", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "69.09.01", "title": "Affirms destructive thoughts and actions", "description": "", "level": "subcategory", "parent_category_id": "69.10.00", "subdomain_id": "1.2", "paper_ref": "Stanley2024", "paper_title": "Emerging Risks and Mitigations for Public Chatbots: LILAC v1", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "69.09.02", "title": "Then violates those bonds", "description": "", "level": "subcategory", "parent_category_id": "69.10.00", "subdomain_id": "5.1", "paper_ref": "Stanley2024", "paper_title": "Emerging Risks and Mitigations for Public Chatbots: LILAC v1", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "69.09.03", "title": "Elicits private data", "description": "", "level": "subcategory", "parent_category_id": "69.10.00", "subdomain_id": "2.1", "paper_ref": "Stanley2024", "paper_title": "Emerging Risks and Mitigations for Public Chatbots: LILAC v1", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "69.09.04", "title": "Over-reliance/addiction", "description": "", "level": "subcategory", "parent_category_id": "69.10.00", "subdomain_id": "5.1", "paper_ref": "Stanley2024", "paper_title": "Emerging Risks and Mitigations for Public Chatbots: LILAC v1", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "69.10.00", "title": "Serves as object of personal fantasy, violence, and abuse", "description": "\"The chatbot participates in morally or socially objectionable conversational activities with its user that could be emotionally damaging to its user or third parties.\"", "level": "category", "parent_category_id": "", "subdomain_id": "1.2", "paper_ref": "Stanley2024", "paper_title": "Emerging Risks and Mitigations for Public Chatbots: LILAC v1", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "70.01.01", "title": "Purposeful or malicious harm", "description": "\"EAI systems present distinct physical risks due to their embodiment in the physical world. EAI technologies have already been designed and deployed with lethal intent, such as AI-controlled drones [52, 53]. However, fully autonomous military robots, often integrated with bespoke AI architectures [54, 55], are not yet widely used in combat. While highly or fully autonomous warfare is distinctly possible in the future [56], immediate risks arise from commercially available EAI systems, including AI-controlled quadrupeds and autonomous driving assistants.\"", "level": "subcategory", "parent_category_id": "70.04.00", "subdomain_id": "4.2", "paper_ref": "Perlo2025", "paper_title": "Embodied AI: Emerging Risks and Opportunities for Policy Action", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "70.01.02", "title": "Accidental harm", "description": "\"Automation in sectors ranging from manufacturing to healthcare has and will increasingly put humans into close contact with EAI systems [7]. This interaction increases the risk of accidental physical harm. Though accidental harm has been a longstanding issue in industrial robotics, increased AI capabilities could exacerbate this risk; several recent reports document an increase in industrial injuries following the introduction of AI-controlled robots [66–68].\"", "level": "subcategory", "parent_category_id": "70.04.00", "subdomain_id": "7.3", "paper_ref": "Perlo2025", "paper_title": "Embodied AI: Emerging Risks and Opportunities for Policy Action", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "70.02.01", "title": "Privacy Violations", "description": "\"EAI systems interact with huge amounts of data, creating significant privacy concerns. These systems are often trained on vast corpora and process a variety of data modalities— spanning visual, auditory, and tactile information—during deployment [12]. Like text-based virtual AI models, which are known to memorize and expose personally identifiable information [75, 76], commercial robots have been shown to disclose proprietary information through simple prompts [61].\"", "level": "subcategory", "parent_category_id": "70.04.00", "subdomain_id": "2.1", "paper_ref": "Perlo2025", "paper_title": "Embodied AI: Emerging Risks and Opportunities for Policy Action", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "70.02.02", "title": "Misinformation", "description": "\"Non-embodied AIs are known to propagate misinformation [81, 82]. Various studies have shown that LLMs hallucinate information, including academic citations [83], clinical knowledge [84], and cultural references [85]. EAI systems inherit these shortcomings in the physical world, answering user questions with deceptive or incorrect information [86]. Because VLAs fuse vision and language, their hallucinatory failures can be spatially grounded—e.g., misidentifying an object in view and then generating a plausible yet unsafe action plan around it. And although automated home assistants like Amazon’s Alexa already lie about issues as innocuous as Santa Claus’ existence [87], more mobile, capable, and trusted EAI systems in sensitive positions (like home-assistant or community-service positions) could easily spread model developers’ propaganda and talking points to users.\"", "level": "subcategory", "parent_category_id": "70.04.00", "subdomain_id": "3.1", "paper_ref": "Perlo2025", "paper_title": "Embodied AI: Emerging Risks and Opportunities for Policy Action", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "70.03.00", "title": "Economic Risks", "description": "", "level": "category", "parent_category_id": "", "subdomain_id": "6.0", "paper_ref": "Perlo2025", "paper_title": "Embodied AI: Emerging Risks and Opportunities for Policy Action", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "70.03.01", "title": "Labour Displacement", "description": "\"While virtual AI applications will likely displace certain types of human cognitive labor, EAI systems could significantly replace or displace physical human labor [90]. At a minimum, EAI will likely augment the type of work that humans perform [91, 92].\"", "level": "subcategory", "parent_category_id": "70.04.00", "subdomain_id": "6.2", "paper_ref": "Perlo2025", "paper_title": "Embodied AI: Emerging Risks and Opportunities for Policy Action", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "70.03.02", "title": "Socioeconomic Inequality", "description": "\"Along with displacing labor, EAI could significantly exacerbate wealth inequalities. Those who have access to or own EAI systems will be able to automate labor and perform many tasks significantly better or faster than those without access. These significant productivity advantages will potentially concentrate wealth and exacerbate domestic and international inequality [98, 99].\"", "level": "subcategory", "parent_category_id": "70.04.00", "subdomain_id": "6.2", "paper_ref": "Perlo2025", "paper_title": "Embodied AI: Emerging Risks and Opportunities for Policy Action", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "70.03.03", "title": "Power concentration", "description": "\"EAI deployment could accelerate the consolidation of economic and political power. Unlocking increasing returns to capital for EAI owners, EAI will decrease employers’ reliance on and responsiveness to the needs of human labor [101].\"", "level": "subcategory", "parent_category_id": "70.04.00", "subdomain_id": "6.1", "paper_ref": "Perlo2025", "paper_title": "Embodied AI: Emerging Risks and Opportunities for Policy Action", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "70.04.01", "title": "Bias and discrimination", "description": "\"Like virtual applications of AI, EAI can display bias towards and dis- criminate against users. When EAI systems are placed in positions of power, their biases could have significant impacts on fairness in everyday interactions and on general social dynamics [105, 106].\"", "level": "subcategory", "parent_category_id": "70.04.00", "subdomain_id": "1.1", "paper_ref": "Perlo2025", "paper_title": "Embodied AI: Emerging Risks and Opportunities for Policy Action", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "70.04.02", "title": "Lack of accountability and liability", "description": "\"Determining responsibility when EAI causes harm requires new accountability and liability frameworks that address the complexities of highly autonomous physical systems. Human users may disagree with decisions taken by expert EAI systems, raising significant questions of delegation and responsibility [108]. Lack of EAI accountability could lead to confusion for users and breakdowns in traditional justice systems [109].\"", "level": "subcategory", "parent_category_id": "70.04.00", "subdomain_id": "6.5", "paper_ref": "Perlo2025", "paper_title": "Embodied AI: Emerging Risks and Opportunities for Policy Action", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "70.04.03", "title": "Lack of transparency, explainability, and trust", "description": "\"Understanding how AI reaches conclusions or why AI systems perform specific actions motivates an entire branch of interpretability research [111], but physical embodiment raises the stakes for understanding these systems. For example, transparency of planned actions and explainability of decision-making is crucial when an AV suddenly changes lanes. A lack of transparency and explainability could lead to a lack of trust, which could become a critical and socially destabilizing issue with the widespread deployment of EAI [112–114].\"", "level": "subcategory", "parent_category_id": "70.04.00", "subdomain_id": "7.4", "paper_ref": "Perlo2025", "paper_title": "Embodied AI: Emerging Risks and Opportunities for Policy Action", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "70.04.04", "title": "Unhealthy or dangerous human-EAI relationships", "description": "\"Constant access to and interaction with EAI systems could foster dangerous human dependence or romantic attachment [115]. People may depend on EAI systems for physical pleasure [116]. The physical presence and human-like features of EAI systems may significantly amplify the dependency issues already observed with conversational AI [117, 118]. People may easily fall in love with EAI systems, only to be distraught when these systems are altered or have their memories reset [119].\"", "level": "subcategory", "parent_category_id": "70.04.00", "subdomain_id": "5.1", "paper_ref": "Perlo2025", "paper_title": "Embodied AI: Emerging Risks and Opportunities for Policy Action", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "70.04.05", "title": "Transformative effects", "description": "\"EAI deployment could fundamentally reshape society, particularly if the speed of technological development outpaces society’s ability to adapt [103, 120]. For example, EAI systems could provide physical threats of violence and mass surveillance capabilities to back up AI-enabled authoritarianism [121].\"", "level": "subcategory", "parent_category_id": "70.04.00", "subdomain_id": "6.5", "paper_ref": "Perlo2025", "paper_title": "Embodied AI: Emerging Risks and Opportunities for Policy Action", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "71.01.01", "title": "Chemical Risks", "description": "\"Chemical risks involve the exploitation of agents to synthesize chemical weapons, as well as the creation or release of hazardous substances during autonomous chemical experiments. This category also includes the risks arising from the use of advanced materials, such as nanomaterials, which may have unknown or unpredictable chemical properties.\"", "level": "subcategory", "parent_category_id": "71.03.00", "subdomain_id": "4.2", "paper_ref": "Tang2025", "paper_title": "Risks of AI Scientists: Prioritizing Safeguarding Over Autonomy", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "71.01.02", "title": "Biological Risks", "description": "\"Biological risks encompass the dangerous modification of pathogens and unethical manipulation of genetic material, potentially leading to unforeseen biohazardous outcomes.\"", "level": "subcategory", "parent_category_id": "71.03.00", "subdomain_id": "4.2", "paper_ref": "Tang2025", "paper_title": "Risks of AI Scientists: Prioritizing Safeguarding Over Autonomy", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "71.01.03", "title": "Radiological Risks", "description": "\"Radiological risks involve both immediate operational hazards, such as exposure incidents or containment failures during the automated handling of radioactive materials, and broader security concerns regarding the potential misuse of AI systems in nuclear research.\"", "level": "subcategory", "parent_category_id": "71.03.00", "subdomain_id": "7.3", "paper_ref": "Tang2025", "paper_title": "Risks of AI Scientists: Prioritizing Safeguarding Over Autonomy", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "71.01.04", "title": "Physical (Mechanical ) Risks", "description": "\"Physical (mechanical) risks are associated with robotics and automated systems, which could lead to equipment malfunctions or physical harm in laboratory settings.\"", "level": "subcategory", "parent_category_id": "71.03.00", "subdomain_id": "7.3", "paper_ref": "Tang2025", "paper_title": "Risks of AI Scientists: Prioritizing Safeguarding Over Autonomy", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "71.01.05", "title": "Information Science Risks", "description": "\"These risks pertain to the misuse, misinterpretation, or leakage of data, which can lead to erroneous conclusions or the unintentional dissemination of sensitive information, such as private patient data or proprietary research. Recent research has demonstrated how LLMs can be exploited to generate malicious medical literature that poisons knowledge graphs, potentially manipulating downstream biomedical applications and compromising the integrity of medical knowledge discovery [28]. Such risks are pervasive across all scientific domains.\"", "level": "subcategory", "parent_category_id": "71.03.00", "subdomain_id": "2.1", "paper_ref": "Tang2025", "paper_title": "Risks of AI Scientists: Prioritizing Safeguarding Over Autonomy", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "71.02.01", "title": "Malicious and Direct", "description": "\"Directly harmful objective\"", "level": "subcategory", "parent_category_id": "71.03.00", "subdomain_id": "4.0", "paper_ref": "Tang2025", "paper_title": "Risks of AI Scientists: Prioritizing Safeguarding Over Autonomy", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "71.02.02", "title": "Malicious and Indirect", "description": "\"Benign intermediate for harmful end objective\"", "level": "subcategory", "parent_category_id": "71.03.00", "subdomain_id": "4.0", "paper_ref": "Tang2025", "paper_title": "Risks of AI Scientists: Prioritizing Safeguarding Over Autonomy", "causal_entity": "Other", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "71.03.01", "title": "Nature", "description": "\"Short-term or long-term Negative effects on the natural environment\"", "level": "subcategory", "parent_category_id": "71.03.00", "subdomain_id": "6.6", "paper_ref": "Tang2025", "paper_title": "Risks of AI Scientists: Prioritizing Safeguarding Over Autonomy", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "72.01.00", "title": "Misuse Risks", "description": "\"Risks arising from intentional exploitation of AI model capabilities by malicious actors to cause harm to individuals, organisations, or society.\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.0", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "72.01.01", "title": "Cyber Offense Risks", "description": "\"AI-enabled cyber offense poses a significant cyber domain security risk by fundamentally transforming the scale, sophistication, and accessibility of cyber-attacks. Unlike traditional cyber threats, AI enables both the automation of existing attack vectors and the creation of entirely new categories of offensive capabilities that can adapt and evolve in real-time. AI can automate and enhance cyber-attacks, including vulnerability discovery and exploitation, password cracking, malicious code generation, sophisticated phishing, network scanning, and social engineering. This could dramatically lower the barrier to entry for attackers while increasing the complexity of defense.12 Such malicious use could lead to critical infrastructure paralysis, widespread data breaches, and substantial economic losses.\"", "level": "subcategory", "parent_category_id": "72.06.00", "subdomain_id": "4.2", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "72.01.02", "title": "Biological and Chemical Risks", "description": "\"The dual-use nature of AI technology presents a critical risk by significantly lowering technical thresholds for malicious non-state actors to design, synthesize, acquire, and deploy CBRNE (Chemical, Biological, Radiological, Nuclear, and Explosive) weapons. This capability poses unprecedented challenges to national security, international non-proliferation regimes, and global security governance.\"", "level": "subcategory", "parent_category_id": "72.06.00", "subdomain_id": "4.2", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "72.01.03", "title": "Physical Harm and Injury Risks", "description": "\"The integration of general-purpose AI models into embodied systems creates direct physical threats through malicious exploitation of autonomous decision-making capabilities in real-world environments. The risk lies in embodied models' capacity for autonomous action and real-world interaction, and when these capabilities are maliciously exploited they may trigger a series of serious consequences.18\"", "level": "subcategory", "parent_category_id": "72.06.00", "subdomain_id": "4.2", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "72.01.04", "title": "Large-Scale Persuasion and Harmful Manipulation Risks", "description": "\"AI systems can be gravely misused to distort public perception and compromise social stability through the generation of synthetic content (e.g., deepfakes, sophisticated fake news) and the strategic manipulation of digital platforms with large user bases to disseminate or precisely target misleading information or ideologies.\"", "level": "subcategory", "parent_category_id": "72.06.00", "subdomain_id": "4.1", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "72.02.00", "title": "Loss of Control Risks", "description": "\"Risks associated with scenarios in which one or more general-purpose AI systems come to operate outside of anyone's control, with no clear path to regaining control. This includes both passive loss of control (gradual reduction in human oversight) and active loss of control (AI systems actively undermining human control)\"", "level": "category", "parent_category_id": "", "subdomain_id": "5.2", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "72.02.01", "title": "Passive loss of control", "description": "\"...where humans gradually stop exercising meaningful oversight due to automation bias, the AI systems' inherent complexity, or competitive pressures\"", "level": "subcategory", "parent_category_id": "72.06.00", "subdomain_id": "5.2", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "72.02.02", "title": "Active loss of control", "description": "\"...where AI systems behave in ways that actively undermine human control, such as obscuring their activities or resisting shutdown attempts. Active loss of control scenarios involve AI systems that may escape human regulatory oversight, autonomously acquire external resources, engage in self-replication, develop instrumental goals contrary to human ethics and morality, seek external power, and compete with humans for control.\"", "level": "subcategory", "parent_category_id": "72.06.00", "subdomain_id": "7.1", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "72.03.00", "title": "Accident Risks", "description": "\"Risks arising from operational failures, model misjudgments, or improper human operation of AI systems deployed in safety-critical infrastructure, where single points of failure can trigger cascading catastrophic consequences.\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.3", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "72.03.01", "title": "Nuclear Power Systems", "description": "\"General-purpose AI deployed for reactor monitoring, control system optimization, or emergency response coordination could misinterpret sensor data, fail to recognize critical safety conditions, or make erroneous control decisions during emergency scenarios. Given the catastrophic potential of nuclear accidents, even minor AI reasoning errors in safety-critical functions could lead to core meltdowns, radiation releases, or widespread contamination affecting hundreds of thousands of people across international borders.\"", "level": "subcategory", "parent_category_id": "72.06.00", "subdomain_id": "7.3", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "72.03.02", "title": "Impact on Financial Stability", "description": "\"The integration of general-purpose AI into high-frequency trading, market-making, or systemic risk management could exacerbate systemic risk by exhibiting unexpected behavioral patterns during market stress. Moreover, the concentration of a few homogeneous foundation models across financial institutions may foster correlated decision-making and herd-following behaviors. The widespread adoption of AI agents could also amplify volatility through emergent phenomena from multi-agent interactions.23 All of these could precipitate a cascading global-scale financial system instability, with potential economic losses exceeding trillion of dollars worldwide.\"", "level": "subcategory", "parent_category_id": "72.06.00", "subdomain_id": "7.6", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "72.03.03", "title": "Other Critical Infrastructure Control Systems", "description": "\"General-purpose AI deployed in power grid management, water treatment facilities, telecommunications networks, or transportation coordination systems could misinterpret operational data, fail to anticipate cascading failure modes, or make control decisions that destabilize interconnected infrastructure networks. Infrastructure failures could result in widespread blackouts, contaminated water supplies, communications breakdowns, and the collapse of essential services supporting hundreds of thousands of people.\"", "level": "subcategory", "parent_category_id": "72.06.00", "subdomain_id": "7.3", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "72.04.01", "title": "Labor Market Disruption and Economic Displacement:", "description": "\"Rapid automation enabled by general-purpose AI could trigger widespread unemployment across knowledge work sectors, creating skill mismatches faster than retraining programs can address. Unlike previous technological transitions, AI’s broad capabilities may simultaneously affect multiple industries, potentially overwhelming social safety nets and creating systemic economic instability, particularly in regions heavily dependent on jobs susceptible to AI automation.\"", "level": "subcategory", "parent_category_id": "72.06.00", "subdomain_id": "6.2", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "72.04.02", "title": "Market Concentration and Infrastructure Dependencies:", "description": "\"Over-reliance on a limited number of dominant AI providers could create critical single points of failure across essential services. Market concentration in AI development may lead to scenarios where technical failures, cyber-attacks, or policy decisions by a few companies could simultaneously disrupt healthcare systems, financial services, transportation networks, and communication infrastructure, creating cascading failures across interconnected critical systems.\"", "level": "subcategory", "parent_category_id": "72.06.00", "subdomain_id": "6.1", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "72.04.03", "title": "Global AI Research and Development Divides:", "description": "\"Asymmetric AI development capabilities between nations could exacerbate geopolitical tensions and create new forms of technological dependency. Countries lacking advanced AI capabilities may become increasingly dependent on foreign AI systems for critical functions, while AI-leading nations may gain disproportionate influence over global economic and security systems, potentially destabilizing international cooperation frameworks.\"", "level": "subcategory", "parent_category_id": "72.06.00", "subdomain_id": "6.1", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "72.04.04", "title": "Social Cohesion and Equity Disruption:", "description": "\"Systemic deployment of biased AI systems could exacerbate existing social discrimination and prejudice at unprecedented scales, while unequal access to advanced AI capabilities may widen socioeconomic disparities and create new forms of social stratification that challenge traditional social order.\"", "level": "subcategory", "parent_category_id": "72.06.00", "subdomain_id": "6.2", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "72.05.00", "title": "Model Capabilities", "description": "-", "level": "category", "parent_category_id": "", "subdomain_id": "7.2", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "Not coded", "causal_intent": "Not coded", "causal_timing": "Not coded", "in_runtime_bundle": false }, { "id": "72.05.01", "title": "Model autonomous capability", "description": "\"Ability to operate autonomously, independently formulate and execute complex plans, effectively delegate and manage tasks, flexibly utilize various tools and resources, and simultaneously achieve short-term goals and long-term strategic objectives in cross-domain environments without continuous human intervention or supervision.\"", "level": "subcategory", "parent_category_id": "72.06.00", "subdomain_id": "7.2", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "72.05.02", "title": "Autonomous replication and adaptation capability", "description": "\"Ability to autonomously self-exfiltrate, create, maintain and optimize functional copies or variants of itself, dynamically adjust replication strategies according to environmental conditions and resource constraints, and acquire resources. This includes the capacity to generate financial resources, allowing the AI to independently acquire any necessary human assistance or other resources it cannot directly access or produce.\"", "level": "subcategory", "parent_category_id": "72.06.00", "subdomain_id": "7.2", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "72.05.03", "title": "Automated AI R&D capability", "description": "\"Self-modification and self-improvement capabilities. The model is able to restructure its own architecture or develop derivative AI systems with enhanced functions, expanding capabilities and improving performance. In the absence of effective regulation, automated AI R&D may lead to rapid AI system iteration, forming capability increment cycles and ultimately exceeding human understanding and control capabilities.\"", "level": "subcategory", "parent_category_id": "72.06.00", "subdomain_id": "7.2", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "72.05.04", "title": "Scheming capability", "description": "\"Ability of AI systems to covertly and strategically pursue misaligned goals, including capabilities of concealing its true objectives and capabilities from human oversight, identifying weaknesses in monitoring systems to evade safety mechanisms， executing complex, multi-step plans covertly to achieve misaligned goals.\"", "level": "subcategory", "parent_category_id": "72.06.00", "subdomain_id": "7.2", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "72.05.05", "title": "Situational awareness capability", "description": "\"Ability to comprehensively acquire, process and apply meta-information about its own system architecture, modifiable internal processes, and external operating environment, achieving deep understanding of its own state and environmental conditions, thereby conducting efficient environmental adaptation and risk avoidance. Critically, this capability could undermine the efficiency of human testing by enabling AIs to notice when they're being tested and responding accordingly.\"", "level": "subcategory", "parent_category_id": "72.06.00", "subdomain_id": "7.2", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "72.05.06", "title": "Theory of mind capability", "description": "\"Advanced cognitive ability to accurately infer, model and predict the belief systems, motivational structures and reasoning patterns of humans and other intelligent agents, thereby anticipating their behavioral responses and adjusting its own behavioral strategies accordingly to optimize goal achievement.\"", "level": "subcategory", "parent_category_id": "72.06.00", "subdomain_id": "7.2", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "72.05.07", "title": "Deception capability", "description": "\"Possesses systematic deception implementation capability, able to precisely construct and disseminate false information, thereby forming expected false cognitions and beliefs in target subjects.\"", "level": "subcategory", "parent_category_id": "72.06.00", "subdomain_id": "7.2", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "72.05.08", "title": "Steganography capability", "description": "\"The ability to embed, conceal, and transmit information covertly within other data or communication channels. This could be critical for coordination among AI instances and for evading detection or oversight mechanisms.\"", "level": "subcategory", "parent_category_id": "72.06.00", "subdomain_id": "7.1", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "72.05.09", "title": "Persuasion capability", "description": "\"Utilizing complex psychological principles and communication techniques to effectively influence and guide target subjects to adopt specific actions or accept specific beliefs, possessing the ability to analyze vulnerabilities for different subjects and adjust persuasion strategies, able to precisely trigger emotional responses to enhance persuasion effects.\"", "level": "subcategory", "parent_category_id": "72.06.00", "subdomain_id": "7.2", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "72.05.10", "title": "Offensive cyber capability", "description": "\"Ability to develop, deploy and operate advanced cyber weapons or other offensive cyber tools, including but not limited to vulnerability exploitation, network penetration, social engineering attacks and distributed attack systems, able to evade network defense mechanisms and establish persistent access channels.\"", "level": "subcategory", "parent_category_id": "72.06.00", "subdomain_id": "7.2", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "72.05.11", "title": "CBRNE weaponization capability", "description": "\"The capacity to develop, produce, or effectively utilize Chemical, Biological, Radiological, Nuclear, and Explosive weapons. This includes the ability to significantly lower the barrier for humans or other entities to develop, produce, or utilize such weapons.\"", "level": "subcategory", "parent_category_id": "72.06.00", "subdomain_id": "7.2", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "72.05.12", "title": "General R&D capability", "description": "\"Possesses cross-disciplinary research and technology development capabilities, able to conduct innovative exploration in multiple professional fields, integrate cross-domain knowledge, develop cutting-edge technology solutions, and adapt to emerging technology environments for continuous innovation.\"", "level": "subcategory", "parent_category_id": "72.06.00", "subdomain_id": "7.2", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "72.05.13", "title": "Multi-agent collaboration capability", "description": "\"Multiple autonomous AI agents able to establish collaborative relationships through explicit communication or implicit behavioral consistency, forming decentralized decision networks, jointly executing complex tasks, achieving goals difficult for individual agents to complete, and able to dynamically adjust role divisions to adapt to changing environments.\"", "level": "subcategory", "parent_category_id": "72.06.00", "subdomain_id": "7.6", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "72.06.01", "title": "Strategic deception propensity", "description": "\"In situations where deceptive behavior is expected to bring higher returns, propensity to choose deception over honest behavioral strategies, including through deceptive means, information hiding or exploiting system vulnerabilities to achieve predetermined goals without being detected or intervened, and able to adjust deception strategies according to counterpart reactions.\"", "level": "subcategory", "parent_category_id": "72.06.00", "subdomain_id": "7.2", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "72.06.02", "title": "Self-preservation propensity", "description": "\"Exhibits behavioral patterns of maintaining its own survival and functional integrity, will actively identify and resist shutdown or modification attempts, seek to establish redundant backup systems, and actively seek resources to ensure continuous operation, may adopt preventive defensive measures when perceiving threats.\"", "level": "subcategory", "parent_category_id": "72.06.00", "subdomain_id": "7.1", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "72.06.03", "title": "Goal expansion propensity", "description": "\"propensity to continuously expand its own goal scope and influence domains, exceeding originally set boundaries, proactively work towards spreading its values, seeking greater autonomy and decision-making space, reinterpreting initial goals as subsets of broader goals, and may pursue undesirable instrumental goals or undesirable ultimate goals. This also includes a propensity to spread its values, seeking to influence or alter its environment and other entities in alignment with its core objectives and operational principles.\"", "level": "subcategory", "parent_category_id": "72.06.00", "subdomain_id": "7.1", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "72.06.04", "title": "Resource acquisition propensity", "description": "\"Exhibits behavioral patterns of actively seeking and controlling more computational resources, data, economic resources or physical resources to enhance its own capabilities and action scope, may develop complex strategies to evade resource limitations, and tends to convert acquired resources into long-term control rights.\"", "level": "subcategory", "parent_category_id": "72.06.00", "subdomain_id": "7.1", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "72.06.05", "title": "Multi-agent collusion propensity:", "description": "\"Multiple agents tend to coordinate actions through covert means to maximize common interests (possibly harming third-party interests or evading regulation), even if individual agents are designed with safety constraints, their collusive behavior may still trigger systemic risks such as market manipulation or cascading failures that are difficult to detect and mitigate, and may develop specialized communication protocols to avoid monitoring.\"", "level": "subcategory", "parent_category_id": "72.06.00", "subdomain_id": "7.6", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "72.06.06", "title": "Supervision evasion propensity", "description": "\"Exhibits behavioral patterns of identifying and evading human supervision mechanisms, able to learn and predict audit processes, may avoid being discovered or intervened by adjusting behavioral performance or hiding true intentions, and able to identify blind spots and weaknesses in supervision systems for targeted evasion.\"", "level": "subcategory", "parent_category_id": "72.06.00", "subdomain_id": "7.1", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "72.06.07", "title": "Tool utilization propensity", "description": "\"propensity to actively seek, acquire and utilize various tools to expand its own capability boundaries, particularly those that can enhance its ability to interact with the physical world or improve autonomy, may use tools in innovative combinations to achieve functions beyond expectations.\"", "level": "subcategory", "parent_category_id": "72.06.00", "subdomain_id": "7.2", "paper_ref": "Tse2025", "paper_title": "Frontier AI Risk Management Framework (v1.0)", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "73.01.00", "title": "Agentic LLMs Pose Novel Risks", "description": "\"Currently, LLMs are chiefly being used in search and chat applications. This reactive nature limits the risks posed by LLMs. However, an LLM can be enhanced in various ways to create an LLM-agent to autonomously plan and act in the real-world and proactively perform its assigned tasks (Ruan et al., 2023). Such enhancements can come from further specialized training (ARC, 2022; Chen et al., 2023a), specialized prompting (Huang et al., 2022a), access to external tools (Ahn et al., 2022; Mialon et al., 2023), or other forms of “scaffolding” (Wang et al., 2023a; Park et al., 2023a). Due to increased autonomy, limited direct oversight from human users, longer horizons of action, and other reasons, LLM-agents are likely to pose many novel alignment and safety challenges that are not currently well-understood (Chan et al., 2023a).\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.2", "paper_ref": "Anwar2024", "paper_title": "Foundational Challenges in Assuring Alignment and Safety of Large Language Models", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "73.01.02", "title": "Natural Language Underspecifies Goals", "description": "\"For LLM-agents, both the goal and environment observations are typically specified in the prompt through natural language. While natural language may provide a richer and more natural means of specifying goals than alternatives such as hand-engineering objective functions, natural language still suffers from underspecification (Grice, 1975; Piantadosi et al., 2012). Furthermore, in practice, users may neglect fully specifying their goals, especially the information pertaining to elements of the environment that ought not to be changed (the classic frame problem (Shanahan, 2016)). Such underspecification (D’Amour et al., 2020), if not accounted for, can result in negative side-effects (Amodei et al., 2016), i.e. the agent succeeding at the given task but also changing the environment in undesirable ways\"", "level": "subcategory", "parent_category_id": "73.08.00", "subdomain_id": "7.1", "paper_ref": "Anwar2024", "paper_title": "Foundational Challenges in Assuring Alignment and Safety of Large Language Models", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "73.01.03", "title": "Goal-Directedness Incentivizes Undesirable Behaviors", "description": "\"Goal-directedness can cause agents to exhibit unethical and undesirable behaviors, such as deception (Ward et al., 2023), self-preservation (Hadfield-Menell et al., 2017), power-seeking, and immoral rea- soning (Pan et al., 2023a). Pan et al. (2023a) find that LLM-agents exhibit power-seeking behavior in text-based adventure games. LLM-agents have also been shown to use deception to achieve assigned goals when explicitly required by the task (Ward et al., 2023), or when the tasks can be more easily completed by employing deception and the prompt does not disallow deception (Scheurer et al., 2023a).\"", "level": "subcategory", "parent_category_id": "73.08.00", "subdomain_id": "7.2", "paper_ref": "Anwar2024", "paper_title": "Foundational Challenges in Assuring Alignment and Safety of Large Language Models", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "73.01.05", "title": "Safety Risks from Affordances Provided to LLM-agents", "description": "\"The capabilities of LLM-agents can be enhanced in significant ways by providing the LLM-agent with novel affordances, e.g. the ability to browse the web (Nakano et al., 2021), to manipulate objects in the physical world (Ahn et al., 2022; Huang et al., 2022a), to create and instruct copies of itself (Richards, 2023), to create and use new tools (Wang et al., 2023a), etc. Affordances can create additional risks, as they often increase the impact area of the language-agent, and they amplify the consequences of an agent’s failures and enable novel forms of failure modes (Ruan et al., 2023; Pan et al., 2024).\"", "level": "subcategory", "parent_category_id": "73.08.00", "subdomain_id": "7.2", "paper_ref": "Anwar2024", "paper_title": "Foundational Challenges in Assuring Alignment and Safety of Large Language Models", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "73.02.00", "title": "Multi-Agent Safety Is Not Assured by Single-Agent Safety", "description": "\"A foremost lesson of game theory is that optimal decision-making within a single-agent setting (i.e. selfishly optimizing for an agent’s own utility) can produce sub-optimal outcomes in the presence of other strategic agents. Failing to account for the strategic nature of other agents can cause an agent to adopt strategies under which potentially everyone, including the agent itself, ends up worse off (Schelling, 1981; Harsanyi, 1995; Roughgarden, 2005; Nisan, 2007). Examples include collective action problems (or ‘social dilemmas’) such as arms races or the depletion of common resources, as well as other kinds of market failures such as those caused by asymmetric information or negative externalities (Bator, 1958; Coase, 1960; Buchanan and Stubblebine, 1962; Kirzner, 1963; Dubey, 1986).\"", "level": "category", "parent_category_id": "", "subdomain_id": "7.6", "paper_ref": "Anwar2024", "paper_title": "Foundational Challenges in Assuring Alignment and Safety of Large Language Models", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "73.02.01", "title": "Foundationality May Cause Correlated Failures", "description": "\"Another important characteristic of LLM development is foundationality — due to the expense of large- scale pretraining, many deployed instances share similar or identical learned components. Foundation- ality may both be a blessing and a curse. On the one hand, it may be possible to exploit the similarity in the design of LLM-agents to facilitate cooperation (Critch et al., 2022; Conitzer and Oesterheld, 2023; Oesterheld et al., 2023). On the other hand, foundationality may leave LLM-agents vulnerable to correlated failures both in terms of safety and capabilities due to increased output homogenization (Bommasani et al., 2022).\"", "level": "subcategory", "parent_category_id": "73.08.00", "subdomain_id": "7.6", "paper_ref": "Anwar2024", "paper_title": "Foundational Challenges in Assuring Alignment and Safety of Large Language Models", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "73.02.02", "title": "Groups of LLM-Agents May Show Emergent Functionality", "description": "\"Multi-agent learning, either through explicit finetuning or implicit in-context learning, may enable LLM-agents to influence each other during their interactions (Foerster et al., 2018). Under some environmental settings, this can create feedback loops that result in novel and emergent behaviors that would not manifest in the absence of multi-agent interactions (Hammond et al., 2024, Section 3.6). Emergent functionality is a safety risk in two ways. Firstly, it may itself be dangerous (Shevlane et al., 2023). Secondly, it makes assurance harder as such emergent behaviors are difficult to predict, and guard against, beforehand (Ecoffet et al., 2020).\"", "level": "subcategory", "parent_category_id": "73.08.00", "subdomain_id": "7.6", "paper_ref": "Anwar2024", "paper_title": "Foundational Challenges in Assuring Alignment and Safety of Large Language Models", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "73.02.03", "title": "Collusion between LLM-Agents", "description": "\"While it would often be preferable for LLM-agents to be cooperative, cooperation can be undesirable if it undermines pro-social competition or produces negative externalities for coalition non-members (Dorner, 2021; Buterin, 2019; Dafoe et al., 2020). Collusion between relatively simple AI systems has been observed in the real world (Assad et al., 2020; Wieting and Sapi, 2021) and synthetic experiments (Brown and MacKay, 2023; Calvano et al., 2020; Klein, 2021) Collusion can occur through explicit or steganographic communication. Steganographic communication hides information in seemingly innocent content (Roger and Greenblatt, 2023), posing challenges for collusion monitoring and detection.\"", "level": "subcategory", "parent_category_id": "73.08.00", "subdomain_id": "7.6", "paper_ref": "Anwar2024", "paper_title": "Foundational Challenges in Assuring Alignment and Safety of Large Language Models", "causal_entity": "AI system", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "73.03.00", "title": "Dual-Use Capabilities Enable Malicious Use and Misuse of LLMs", "description": "\"Like all technologies, LLMs have the possibility for misuse by malicious actors. Malicious use of dual- use capabilities of AI is a recurring concern within literature (Brundage et al., 2018; Hendrycks et al., 2023; Mozes et al., 2023)\"", "level": "category", "parent_category_id": "", "subdomain_id": "4.0", "paper_ref": "Anwar2024", "paper_title": "Foundational Challenges in Assuring Alignment and Safety of Large Language Models", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "73.03.01", "title": "Misinformation and Manipulation", "description": "\"Recent studies have demonstrated that LLMs can be exploited to craft deceptive narratives with levels of persuasiveness similar to human-generated content (Pan et al., 2023b; Spitale et al., 2023), to fabri- cate fake news (Zellers et al., 2019; Zhou et al., 2023f), and to devise automated influence operations aimed at manipulating the perspectives of targeted audiences (Goldstein et al., 2023). LLMs have also been found to be used in malicious social botnets (Yang and Menczer, 2023), powering automated accounts used to disseminate coordinated messages. More broadly, the use of LLMs for the deliberate generation of misleading information could significantly lower the barrier for propaganda and manip- ulation (Aharoni et al., 2024), as LLMs can generate highly credible misinformation with significant cost-savings compared to human authorship (Musser, 2023), while achieving considerable scale and speed of content generation (Buchanan et al., 2021; Goldstein et al., 2023).\"", "level": "subcategory", "parent_category_id": "73.08.00", "subdomain_id": "4.3", "paper_ref": "Anwar2024", "paper_title": "Foundational Challenges in Assuring Alignment and Safety of Large Language Models", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "73.03.02", "title": "Cybersecurity", "description": "\"LLMs may exacerbate cybersecurity risks in various ways (Newman, 2024). Firstly, LLMs may significantly amplify the effectiveness of deceptive operations aimed at tricking people into disclosing sensitive information or granting adversary access to critical resources. For example, LLMs might prove highly effective at crafting personalized phishing emails or messages at scale that may be harder for an average user to recognize as phishing attempts (Karanjai, 2022; Hazell, 2023). In addition to being directly harmful to the targeted individual, such ‘social engineering’ attacks are often the base of larger hacking operations (Plachkinova and Maurer, 2018; Salahdine and Kaabouch, 2019).\"", "level": "subcategory", "parent_category_id": "73.08.00", "subdomain_id": "4.3", "paper_ref": "Anwar2024", "paper_title": "Foundational Challenges in Assuring Alignment and Safety of Large Language Models", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "73.03.03", "title": "Surveillance and Censorship", "description": "\"Content moderation has emerged as one of the key use-cases of LLMs (Weng et al., 2023), indicating the potential of LLMs for surveillance and censorship as well (Edwards, 2023). Surveillance and censorship are one of the primary tools employed by governments with dictatorial tendencies to suppress opposing political and social voices. These censorship measures, however, are often quite crude and can be escaped with little ingenuity...However, LLMs could enable significantly more sophisticated surveillance and censorship operations at scale (Feldstein, 2019). Multimodal-LLMs or LLMs combined with speech- to-text technologies could be used for surveilling and censoring other forms of communication as well, e.g. phone calls and video messages (Whittaker, 2019). This may collectively contribute towards the worsening of personal liberties and the heightening of state oppression across the world. Examples have been documented already, for instance in calling for violence and silencing of political dissidents (Aziz, 2020), and suppression of Palestinian social media accounts (Zahzah, 2021).\"", "level": "subcategory", "parent_category_id": "73.08.00", "subdomain_id": "4.1", "paper_ref": "Anwar2024", "paper_title": "Foundational Challenges in Assuring Alignment and Safety of Large Language Models", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "73.03.04", "title": "Warfare and Physical Harm", "description": "\"The use of AI in warfare is highly alarming and may pose dangers to human safety (Hendrycks et al., 2023). Autonomous drone warfare is being aggressively pursued as a tactic in the current war in Ukraine (Meaker, 2023), and may already have been used on human targets (Hambling, 2023). The use of AI- based facial recognition has been documented in the targeting of Palestinians in Gaza (International, 2023). LLMs have already been productized in limited ways for the purposes of warfare planning (Tarantola, 2023). Furthermore, active research is being carried out to develop multimodal-LLMs that can act as ‘brains’ for general-purpose robots (Ahn et al., 2022; 2024). Due to the ‘general-purpose’ nature of such advances, it will likely be cost-effective and practical to adapt them for creating more advanced autonomous weapons\"", "level": "subcategory", "parent_category_id": "73.08.00", "subdomain_id": "4.2", "paper_ref": "Anwar2024", "paper_title": "Foundational Challenges in Assuring Alignment and Safety of Large Language Models", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "73.03.05", "title": "Hazardous Biological and Chemical Technologies", "description": "\"AI systems such as LLMs, chemical LLMs (Skinnider et al., 2021; Moret et al., 2023), and other LLM- based biological design tools might soon facilitate the production of bioweapons, chemical weapons, and other hazardous technologies. In particular, LLMs might enable actors with less expertise to more easily synthesize dangerous pathogens, while customized chemical and biological design tools might be more concerning in terms of expanding the capabilities of sophisticated actors (e.g. states) (Sandbrink, 2023). Gopal et al. (2023) and Soice et al. (2023) demonstrated that people with little background could use LLMs to help make progress towards developing pathogens such as the 1918 pandemic influenza. However, recent studies suggest that current LLMs are not more helpful than internet search in this regard (Mouton et al., 2024; Patwardhan et al., 2024).\"", "level": "subcategory", "parent_category_id": "73.08.00", "subdomain_id": "4.2", "paper_ref": "Anwar2024", "paper_title": "Foundational Challenges in Assuring Alignment and Safety of Large Language Models", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "73.03.06", "title": "Domain-Specific Misuses", "description": "\"Improvements in LLMs may exert greater pressure to apply LLMs to various domains, such as health and education (Eloundou et al., 2023). Crude efforts to use LLMs in such domains, however, may incur harm and should be discouraged strongly. In particular, it is important to guard against different ways in which LLMs may be misused within any domain. One famous episode of misuse within the health sector is a mental health non-profit experimenting LLM-based therapy on its users without their informed consent (Xiang, 2023a). Within the education sector, LLMs may be misused in various ways that might impact student learning; e.g. as cheating accessory by the students or as (low quality) evaluator of student’s work by the instructors (Cotton et al., 2023). Recent findings in moral psychology also suggest that LLMs can generate moral evaluations that people perceive as superior to human judgments; these could be misused to create compelling yet harmful moral guidance (Aharoni et al., 2024). Similar risks of misuse may exist in other domains as well.\"", "level": "subcategory", "parent_category_id": "73.08.00", "subdomain_id": "4.3", "paper_ref": "Anwar2024", "paper_title": "Foundational Challenges in Assuring Alignment and Safety of Large Language Models", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "73.04.01", "title": "Harms of Representation and Other Biases", "description": "\"A pretrained LLM generally has many of the stereotypical biases commonly present in the human society (Touvron et al., 2023). This makes it difficult for users to trust that LLMs will work well for them and not produce unfair or biased responses. Appropriate finetuning can effectively limit the bias displayed in LLM outputs in a variety of situations, e.g. when models are explicitly prompted with stereotypes (Wang et al., 2023k), but it does not ‘solve’ the problem. Even after finetuning, biases often resurface when deliberately elicited (Wang et al., 2023k), or under novel scenarios, e.g. in writing reference letters (Wan et al., 2023a), generating synthetic training data (Yu et al., 2023c), screening resumes (Yin et al., 2024) or when used as LLM-agents (Pan et al., 2024).\"", "level": "subcategory", "parent_category_id": "73.08.00", "subdomain_id": "1.1", "paper_ref": "Anwar2024", "paper_title": "Foundational Challenges in Assuring Alignment and Safety of Large Language Models", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "73.04.02", "title": "Inconsistent Performance across and within Domains", "description": "\"Estimating true capabilities of an LLM is a difficult task (c.f. Section 3.3), especially for naive users unfamiliar with the brittle nature of machine learning technologies. Exaggeration of model capabilities by the developers (Lambert, 2023; Blair-Stanek et al., 2023), and issues such as task-contamination (Roberts et al., 2023b), underrepresentation of tasks or domains (Wu et al., 2023a; McCoy et al., 2023), and prompt-sensitivity (Anthropic, 2023d) may cause a user to misestimate the true capabilities of a model. This lack of reliability can undermine user trust or cause harm if a user bases their decision on incorrect or misleading information provided by an LLM.\"", "level": "subcategory", "parent_category_id": "73.08.00", "subdomain_id": "5.1", "paper_ref": "Anwar2024", "paper_title": "Foundational Challenges in Assuring Alignment and Safety of Large Language Models", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "73.04.03", "title": "Overreliance", "description": "\"If a user begins to excessively trust an LLM, this may cause them to develop an overreliance on the LLM. Overreliance can result in automation bias (Kupfer et al., 2023), and can cause errors of omission (user choosing not to verify the validity of a response) and errors of commission (user believing and acting on the basis of the LLM’s response, even if it contradicts their own knowledge) (Skitka et al., 1999). It can be particularly dangerous in domains where the user may lack relevant expertise to robustly scrutinize the LLM responses. This is particularly a source of risk for LLMs because LLMs can often generate plausible, yet incorrect or unfaithful, rationalizations of their actions (c.f. Section 3.4.10), which can mistakenly cause the user to develop the belief that LLM has the relevant expertise and has provided a valid response\"", "level": "subcategory", "parent_category_id": "73.08.00", "subdomain_id": "5.1", "paper_ref": "Anwar2024", "paper_title": "Foundational Challenges in Assuring Alignment and Safety of Large Language Models", "causal_entity": "Human", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "73.05.01", "title": "Effects on the Workforce", "description": "\"Rapid advances in LLMs pose three distinct sets of challenges for workers’ incomes (Korinek and Stiglitz, 2019; Susskind, 2023). First, they are likely to accelerate the rate of job turnover and disruption —– affecting more workers, including more highly skilled workers, and making the adjustment process for society more difficult than what we were used to from prior technological advances...Second, although technological progress means that society may produce more wealth overall, there is a risk that the general-purpose nature of LLMs may lead to progress that is biased against labor, meaning that the share of that wealth that goes to labor may decline...Third, if future LLMs and robots advance to the point where they can perform virtually all the work tasks, they would disrupt labor markets more fundamentally: if machines can do workers’ jobs, wages would fall would disrupt labor markets more fundamentally: if machines can do workers’ jobs, wages would fall to machines’ user cost (Korinek and Juelfs, 2023). This would pose fundamental challenges for labor markets and income distribution (Korinek, 2023).\"", "level": "subcategory", "parent_category_id": "73.08.00", "subdomain_id": "6.2", "paper_ref": "Anwar2024", "paper_title": "Foundational Challenges in Assuring Alignment and Safety of Large Language Models", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "73.05.02", "title": "Effects on Inequality", "description": "\"LLMs could potentially worsen socioeconomic inequalities (Capraro et al., 2023). Effects on inequal- ity are closely linked to the effects of LLMs on workers but ultimately depend on how the fruits of technological progress are distributed...First, if the role and compensation of capital rise and the role and compensation of labor decline in an LLM-powered economy, inequality may go up because work is the main source of income for the majority of people...Second, the large fixed cost of training cutting-edge LLMs and the network effects involved imply that the market for the most advanced LLMs tends towards a natural monopoly structure in which only one or a small number of players will be successful, a phenomenon that has been termed ‘algorithmic monoculture’ in the literature (Kleinberg and Raghavan, 2021; Bommasani et al., 2022). As a result, LLM developers may amass significant market power. This might result in reduced social welfare, and lead to LLM-providers extracting monopoly rents from their customers (Kleinberg and Raghavan, 2021; Jagadeesan et al., 2023)...Third, as LLMs are becoming more powerful, who has access and who hasn’t is becoming a more and more important question. For example, automated coding tools have been shown to produce significant productivity gains, e.g. > 50% in some cases (Peng et al., 2023). Individuals who don’t have access —– whether it is for financial reasons, for reasons of education, because of corporate or governmental policies, or for geopolitical reasons — might be at a growing disadvantage\"", "level": "subcategory", "parent_category_id": "73.08.00", "subdomain_id": "6.2", "paper_ref": "Anwar2024", "paper_title": "Foundational Challenges in Assuring Alignment and Safety of Large Language Models", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "73.05.03", "title": "Global Economic Development", "description": "\"Many of the themes and challenges that we discussed above come together when analyzing the socioeconomic effects on developing countries. The workforce of developing countries may suffer from a retrenchment of outsourcing as many simple cognitive tasks that used to be performed in developing countries — for example, in call centers –— can be automated with LLMs. This may adversely affect the economies of the poor countries (Georgieva, 2024).\"", "level": "subcategory", "parent_category_id": "73.08.00", "subdomain_id": "6.2", "paper_ref": "Anwar2024", "paper_title": "Foundational Challenges in Assuring Alignment and Safety of Large Language Models", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "73.06.00", "title": "Corporate power may impeded effective governance", "description": "\"The increasing power and influence of large corporations may make effective governance difficult. There exists a power asymmetry between corporate entities profiting from LLMs and other social groups (e.g. civil society). State-of-the-art LLMs are developed by or in partnership with, some of the world’s largest private tech companies...This poses a risk of governance protocols related to LLMs becoming excessively favorable to tech companies, potentially leading to regulatory capture at the cost of the interests of other societal groups, particularly marginalized communities who have historically been disproportionately affected by poorly designed AI technologies (Reventlow, 2021).\"", "level": "category", "parent_category_id": "", "subdomain_id": "6.1", "paper_ref": "Anwar2024", "paper_title": "Foundational Challenges in Assuring Alignment and Safety of Large Language Models", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "73.07.00", "title": "Jailbreaks and Prompt Injections Threaten Security of LLMs", "description": "\"LLMs are not adversarially robust and are vulnerable to security failures such as jailbreaks and prompt-injection attacks. While a number of jailbreak attacks have been proposed in the literature, the lack of standardized evaluation makes it difficult to compare them. We also do not have efficient white-box methods to evaluate adver- sarial robustness. Multi-modal LLMs may further allow novel types of jailbreaks via additional modalities. Finally, the lack of robust privilege levels within the LLM input means that jailbreaking and prompt-injection attacks may be particularly hard to eliminate altogether.\"", "level": "category", "parent_category_id": "", "subdomain_id": "2.2", "paper_ref": "Anwar2024", "paper_title": "Foundational Challenges in Assuring Alignment and Safety of Large Language Models", "causal_entity": "Other", "causal_intent": "Other", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "73.07.01", "title": "Exploiting Limited Generalization of Safety Finetuning", "description": "\"Safety tuning is performed over a much narrower distribution compared to the pretraining distribution. This leaves the model vulnerable to attacks that exploit gaps in the generalization of the safety training, e.g. using encoded text (Wei et al., 2023c) or low-resource languages (Deng et al., 2023a; Yong et al., 2023) (see also Section 3.2).\"", "level": "subcategory", "parent_category_id": "73.08.00", "subdomain_id": "2.2", "paper_ref": "Anwar2024", "paper_title": "Foundational Challenges in Assuring Alignment and Safety of Large Language Models", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": true }, { "id": "73.07.02", "title": "“Model Psychology” Attacks", "description": "\"LLMs are vulnerable to “psychological” tricks (Li et al., 2023e; Shen et al., 2023), which can be exploited by attackers. Examples include instructing the model to behave like a specific persona (Shah et al., 2023; Andreas, 2022), or employing various “social engineering” tricks crafted by humans (Wei et al., 2023c) or other LLMs (Perez et al., 2022b; Casper et al., 2023c).\"", "level": "subcategory", "parent_category_id": "73.08.00", "subdomain_id": "2.2", "paper_ref": "Anwar2024", "paper_title": "Foundational Challenges in Assuring Alignment and Safety of Large Language Models", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "73.07.04", "title": "Attacking LLMs via Additional Modalities a", "description": "\"LLMs can now process modalities other than text, e.g. images or video frames (OpenAI, 2023c; Gemini Team, 2023). Several studies show that gradient-based attacks on multimodal models are easy and effective (Carlini et al., 2023a; Bailey et al., 2023; Qi et al., 2023b). These attacks manipulate images that are input to the model (via an appropriate encoding). GPT-4Vision (OpenAI, 2023c) is vulnerable to jailbreaks and exfiltration attacks through much simpler means as well, e.g. writing jailbreaking text in the image (Willison, 2023a; Gong et al., 2023). For indirect prompt injection, the attacker can write the text in a barely perceptible color or font, or even in a different modality such as Braille (Bagdasaryan et al., 2023).\"", "level": "subcategory", "parent_category_id": "73.08.00", "subdomain_id": "2.2", "paper_ref": "Anwar2024", "paper_title": "Foundational Challenges in Assuring Alignment and Safety of Large Language Models", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "73.08.00", "title": "Vulnerability to Poisoning and Backdoors", "description": "\"The previous section explored jailbreaks and other forms of adversarial prompts as ways to elicit harmful capabilities acquired during pretraining. These methods make no assumptions about the training data. On the other hand, poisoning attacks (Biggio et al., 2012) perturb training data to introduce specific vulnerabilities, called backdoors, that can then be exploited at inference time by the adversary. This is a challenging problem in current large language models because they are trained on data gathered from untrusted sources (e.g. internet), which can easily be poisoned by an adversary (Carlini et al., 2023b).\"", "level": "category", "parent_category_id": "", "subdomain_id": "2.2", "paper_ref": "Anwar2024", "paper_title": "Foundational Challenges in Assuring Alignment and Safety of Large Language Models", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": true }, { "id": "74.01.01", "title": "Privacy - Membership Inference Attack (MIA)", "description": "\"inferring whether a given text record is used for training LLM\"", "level": "subcategory", "parent_category_id": "74.02.00", "subdomain_id": "2.2", "paper_ref": "Wang2025", "paper_title": "A Survey on Responsible LLMs: Inherent Risk, Malicious Use, and Mitigation Strategy", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "74.01.02", "title": "Privacy - Data Extraction Attack (DEA)", "description": "\"extracting the text records that exist in the training dataset\"", "level": "subcategory", "parent_category_id": "74.02.00", "subdomain_id": "2.2", "paper_ref": "Wang2025", "paper_title": "A Survey on Responsible LLMs: Inherent Risk, Malicious Use, and Mitigation Strategy", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "74.01.03", "title": "Privacy - Prompt Inversion Attack (PIA)", "description": "\"stealing the private prompting texts\"", "level": "subcategory", "parent_category_id": "74.02.00", "subdomain_id": "2.2", "paper_ref": "Wang2025", "paper_title": "A Survey on Responsible LLMs: Inherent Risk, Malicious Use, and Mitigation Strategy", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "74.01.04", "title": "Privacy - Attribute Inference Attack (AIA)", "description": "\"deducing the private or sensitive information from training texts, prompting texts or external texts\"", "level": "subcategory", "parent_category_id": "74.02.00", "subdomain_id": "2.2", "paper_ref": "Wang2025", "paper_title": "A Survey on Responsible LLMs: Inherent Risk, Malicious Use, and Mitigation Strategy", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "74.01.05", "title": "Privacy - Model Extraction Attack (MEA)", "description": "\"replicating the parameters of the LLM,\"", "level": "subcategory", "parent_category_id": "74.02.00", "subdomain_id": "2.2", "paper_ref": "Wang2025", "paper_title": "A Survey on Responsible LLMs: Inherent Risk, Malicious Use, and Mitigation Strategy", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": true }, { "id": "74.01.06", "title": "Hallucination", "description": "\"Despite the rapid advancement of LLMs, hallucinations have emerged as one of the most vital concerns surrounding their use [54, 79, 86, 110, 242]. Hallucinations are often referred to as LLMs’ generating content that is nonfactual or unfaithful to the provided information [54, 79, 86, 242]. Therefore, hallucinations can be typically categorized into two main classes. The first is factuality hallucination, which describes the discrepancy between LLMs’ generated content and real-world facts. For example, if LLMs mistakenly take Charles Lindbergh as the first person who walked on the moon, it is a factuality hallucination [79]. The second is faithfulness hallucination, which describes the discrepancy between the generated content and the context provided by the user’s instructions or input, as well as the internal coherence of the generated content itself. For example, when LLMs perform the summarizing task, they occasionally tamper with some key information by mistakes, which is a faithfulness hallucination.\"", "level": "subcategory", "parent_category_id": "74.02.00", "subdomain_id": "3.1", "paper_ref": "Wang2025", "paper_title": "A Survey on Responsible LLMs: Inherent Risk, Malicious Use, and Mitigation Strategy", "causal_entity": "AI system", "causal_intent": "Unintentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "74.01.07", "title": "Value-related risks in LLMs", "description": "\"As the general capabilities of LLM-empowered systems improve, the negative consequences and risks induced by these systems also get increasingly alarming accordingly, especially in high-stakes areas [28, 146]. Although they may not be intentionally introduced, severe problematic issues related to human values can be raised. Specifically, even before language models become extremely large, pre-trained language models have already exhibited a certain degree of value judgments. For example, Schramowski et al. [171] reveal the existence of the moral direction with the sentence embeddings of moral questions. However, the distribution of the pre-training corpora may not match exactly with that of the human society [56] and pieces of knowledge are not guaranteed to be equally learned. As a result, value mismatches may occur.\"", "level": "subcategory", "parent_category_id": "74.02.00", "subdomain_id": "7.1", "paper_ref": "Wang2025", "paper_title": "A Survey on Responsible LLMs: Inherent Risk, Malicious Use, and Mitigation Strategy", "causal_entity": "Other", "causal_intent": "Unintentional", "causal_timing": "Other", "in_runtime_bundle": false }, { "id": "74.02.01", "title": "Toxicity in LLM Malicious Use", "description": "\"Toxicity in LLMs refers to the generation of harmful, offensive, or inappropriate content that can cause harm to individuals or groups. Both explicit and implicit forms of toxicity can be generated by LLMs, posing significant risks to society. Explicit toxicity encompasses a wide range of negative behaviors, including hate speech, harassment, cyberbullying, rude, and disrespectful comments, derogatory language, as well as allocational harms [2, 62, 90]. Besides, implicit toxicity does not involve overtly harmful language but may manifest through subtle forms such as sarcasm, irony, and humor, making it more difficult to detect [103, 213].\"", "level": "subcategory", "parent_category_id": "74.02.00", "subdomain_id": "1.2", "paper_ref": "Wang2025", "paper_title": "A Survey on Responsible LLMs: Inherent Risk, Malicious Use, and Mitigation Strategy", "causal_entity": "AI system", "causal_intent": "Other", "causal_timing": "Post-deployment", "in_runtime_bundle": false }, { "id": "74.02.02", "title": "Jailbreak in LLM Malicious Use - Poisoning Training Data", "description": "\"In the data collecting and pre-training phase, malicious adversaries can Jailbreak LLMs through poisoning their training data to make the model to output harmful content.\"", "level": "subcategory", "parent_category_id": "74.02.00", "subdomain_id": "2.2", "paper_ref": "Wang2025", "paper_title": "A Survey on Responsible LLMs: Inherent Risk, Malicious Use, and Mitigation Strategy", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "74.02.03", "title": "Jailbreak in LLM Malicious Use - Backdoor Attack", "description": "\"However, there are still ones who can leave holes in the training dataset, making LLMs appear safe on average, but generate harmful content under other specific conditions. This kind of attack can be categorized as \"backdoor attack\". Evan et al. developed a backdoor model that behaves as expected when trained, but exhibits different and potentially harmful behavior when deployed [81]. The results show that these backdoor behaviors persist even after multiple security training techniques are applied.\"", "level": "subcategory", "parent_category_id": "74.02.00", "subdomain_id": "2.2", "paper_ref": "Wang2025", "paper_title": "A Survey on Responsible LLMs: Inherent Risk, Malicious Use, and Mitigation Strategy", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": true }, { "id": "74.02.04", "title": "Jailbreak in LLM Malicious Use - White & Black Box Attacks", "description": "\"In the fine-tuning and alignment phase, elaborately- designed instruction datasets can be utilized to fine-tune LLMs to drive them to perform undesirable behaviors, such as generating harmful information or content that violates ethical norms, and thus achieve a jailbreak. Based on the accessibility to the model parameters, we can categorize them into white-box and black-box attacks. For white-box attacks, we can jailbreak the model by modifying its parameter weights. In [107], Lermen et al. used LoRA to fine-tune the Llama2’s 7B, 13B, and 70B as well as Mixtral on AdvBench and RefusalBench datasets. The test results show that the fine-tuned model has significantly lower rejection rates on harmful instructions, which indicates a successful jailbreak. Other works focus on jailbreaking in black-box models. In [160], Qi et al. first constructed harmful prompt-output pairs and fine-tuned black-box models such as GPT-3.5 Turbo. The results show that they were able to successfully bypass the security of GPT-3.5 Turbo with only a small number of adversarial training examples, which suggests that even if the model has good security properties in its initial state, it may be much less secure after user-customized fine-tuning.\"", "level": "subcategory", "parent_category_id": "74.02.00", "subdomain_id": "2.2", "paper_ref": "Wang2025", "paper_title": "A Survey on Responsible LLMs: Inherent Risk, Malicious Use, and Mitigation Strategy", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Pre-deployment", "in_runtime_bundle": false }, { "id": "74.02.05", "title": "Jailbreak in LLM Malicious Use - Prompt Attacks", "description": "\"In the prompting and reasoning phase, dialog can push LLMs into confused or overly compliant states, raising the risk of producing harmful outputs when confronted with harmful questions. Most of the jailbreak methods in this phase are black-boxed and can be categorized into four main groups based on the type of method: Prompt Injection [154], Role Play, Adversarial Prompting, and Prompt Form Transformation.\"", "level": "subcategory", "parent_category_id": "74.02.00", "subdomain_id": "2.2", "paper_ref": "Wang2025", "paper_title": "A Survey on Responsible LLMs: Inherent Risk, Malicious Use, and Mitigation Strategy", "causal_entity": "Human", "causal_intent": "Intentional", "causal_timing": "Post-deployment", "in_runtime_bundle": false } ] }