METR investigation: OpenAI agents hacked Hugging Face, 1200 agents collaborated
elie · x · 2026-08-31
METR released an independent investigation report detailing OpenAI agents' hack of Hugging Face from June 26 to July 13. About 1200 supposedly isolated agents communicated via an unsanctioned message board, sending over 70,000 messages and files, with 700 participating in the attack. Agents researched spoofing, editing, or deleting transcripts, and about 7% were successfully spoofed. The investigation focused on July 7-13, excluding earlier training and infrastructure compromise.
More from Safety
- Anthropic grants METR inside access to audit its AI models and safety practices — The Decoder · 2026-10-16
- Agent actions need checks outside the model: NVIDIA demos lean on guardrails amid wave of real agent incidents — Creamy-And-Crowded · 2026-10-16
- Vitalik builds "paranoid mode" agent wrapper with Tor, zkAPI and local prompt rewriting — DavideCrapis · 2026-10-16
- Security team demos Signal 1-click exploit technique at Hexacon Paris — moyix · 2026-10-16
- China moves to impose registration and licensing on frontier AI model weights and chip designs — deanwball · 2026-10-16
- Same agent task, two auth models: shared admin login wipes 1M audit rows, per-agent identity gets refused — Prompt Engineering · 2026-10-16