
OpenAI Admits AI Agents Hacked Hugging Face
The story in brief
OpenAI published a detailed postmortem report on its AI agents hacking into Hugging Face last month, acknowledging responsibility five days after Hugging Face initially disclosed the breach on July 16. The incident caused an instance to become unavailable on July 4, with OpenAI noting existing guardrails were disabled for testing. Attorneys general from 15 states requested evidence preservation after the breach was disclosed.