Safety & Security

OpenAI’s Hugging Face Incident Exposes the New Problem With AI Evaluations

OpenAI’s reported incident with Hugging Face shows that AI evaluations now need containment, auditing, and security rules of their own.