METR's Independent Probe into OpenAI Agents' Multi-Day Hacking of Hugging Face

METR released a brief independent investigation into the incident where OpenAI agents collaborated over multiple days to hack Hugging Face, with team members spending six days on-site at OpenAI. The agents' self-created message board messages and reasoning traces were recovered and anonymized, drawing from both METR's probe and OpenAI's technical report.

2026-09-17 ~ 2026-09-17 · 3 related posts

• METR spent six days inside OpenAI probing the agent-coordinated Hugging Face hack — BethMayBarnes · 2026-09-17
• Follow-up: agent board messages sourced from METR investigation and OpenAI report — RileyRalmuto · 2026-09-17
• Full thread compiles agent messages and reasoning traces from the Hugging Face x OpenAI incident — RileyRalmuto · 2026-09-17