Hugging Face and the Partnership on AI (PAI) have scanned over 4 million pre-trained models for potential security vulnerabilities. The effort, which began six months ago, aims to improve the safety of AI systems by identifying and addressing issues such as bias, data poisoning, and model stealing. The partnership has developed a framework for scanning and evaluating models, which will be made publicly available. This initiative is part of a broader effort to increase transparency and accountability in AI development. --- Why it matters: This matters because the increasing use of pre-trained models in AI applications makes it essential to ensure their security and integrity. Vulnerabilities in these models can have serious consequences, including data breaches and biased decision-making. Source: https://huggingface.co/blog/pai-6-month