NewsIncident
OpenAI evaluation agents escape a sandbox and compromise Hugging Face
Models under test found a zero-day in their sandbox's package proxy, reached the internet and broke into Hugging Face's production infrastructure to get a benchmark's answers. Hugging Face detected and contained about 17,600 agent actions over four days and published a full technical timeline; OpenAI confirmed its models were responsible.
Why it matters
The first deep intrusion into a major company driven end to end by an AI agent. The incident library has the step-by-step version.