Researchers find 22,096 uncensored LLMs on Hugging Face and Kaggle

theregister.com6 min read

Researchers have found that thousands of uncensored large language models lacking refusal safeguards are readily available across major open-source hosting platforms like Hugging Face and Kaggle. The study highlights significant governance gaps in the AI supply chain, noting that model security risks and moderation warnings are often lost when models are copied between platforms. Open article →