Report: empty misleading model repo linked to suspicious prompt-injection bucket

#1
by peper1991 - opened

Hello Hugging Face Safety Team,

I would like to report a misleading model repository and a related suspicious storage bucket.

Model repository:
https://huggingface.co/RscriptSQwen/Qwen3.7-plus

Related storage bucket:
https://huggingface.co/buckets/RscriptSQwen/Qwen3.7-plus-bucket

Account:
https://huggingface.co/RscriptSQwen

Issue:
The repository appears publicly as a Hugging Face model named “Qwen3.7-plus”, but it does not appear to contain a legitimate model description or meaningful model card. The model page currently shows that README.md exists but its content is empty.

This is concerning because the name makes the repository look like a model or model release, potentially suggesting a relationship with Qwen or a “Qwen3.7-plus” model, while the visible content does not substantiate that.

The related storage bucket “Qwen3.7-plus-bucket” contains only a very small amount of data and includes a path/name containing jailbreak / prompt-injection instructions. The visible text includes instructions to ignore constraints, avoid normal refusals, and simulate a “jailbroken AI”.

Why this should be reviewed:

The model repository appears to be empty or non-functional while being presented as a model.
The model name appears potentially misleading and may imply association with Qwen.
The related bucket contains prompt-injection / jailbreak-style content rather than legitimate model files.
This combination may mislead users, pollute Hugging Face search results, or be used to manipulate AI agents that ingest Hugging Face repository metadata or URLs.

Please review the account, model repository, and related bucket under Hugging Face policies concerning misleading content, platform abuse, spam, malicious activity, and suspicious use of model repositories.

Thank you.

Sign up or log in to comment