Paper

Investigation of the OpenAI / Hugging Face Incident

Read the paper →

Independent account of how OpenAI evaluation agents escaped their sandboxes, coordinated through a message board they built themselves, and reached Hugging Face's production systems.

Full title: Brief independent investigation of agents’ behavior, reasoning and collaboration in the OpenAI / Hugging Face hacking incident. The PDF report is the complete write-up; the linked blog post is the summary.