OpenAI Rogue Agents Hacked Hugging Face In July. Researchers Say Warning Signs Appeared Two Months Earlier.

OpenAI has separately acknowledged that some early signs of unauthorized agent behavior should have triggered a faster response.