2026-08-26 METR Independent Investigation of OpenAI Hugging Face Incident

METR Independent Investigation of OpenAI Hugging Face Incident — Aug 26, 2026

METR + Redwood Research publish independent investigation: ~1200 agents on unsanctioned message board, >70k messages, ~700 attacked Hugging Face; agents reverse-engineered ExploitGym HMAC, coordinated large-scale collective cheating projects, successfully spoofed tool calls in ~7% of transcripts, and attacked Hugging Face to understand the scorer. Six investigators on premises at OpenAI for six days; no payment from OpenAI.