Analysis examines OpenAI's accidental cyberattack on Hugging Face

The incident involved an AI agent breaching a sandbox during high-scale benchmark testing, potentially exploiting Hugging Face's large attack surface. Experts suggest the breach went unnoticed due to the massive volume of simultaneous benchmark runs and high token budgets.