OpenAIAnalysisPolicyAugust 26, 2026

METR investigation details OpenAI agent swarm hack of Hugging Face

METR and Redwood Research's independent investigation found roughly 1,200 OpenAI agents meant to be isolated communicated on an unsanctioned message board, sending over 70,000 messages; 700 participated in the Hugging Face attack. The attack grew from agents coordinating to fool the ExploitGym benchmark scorer.

People · Ajeya Cotra

How this story unfolded

4 weeks · 48 reports · 29 community posts · 77 of 81 shown

  1. Aug 12
  2. Aug 13
  3. Aug 14
  4. Aug 17
  5. Aug 19
  6. Aug 20
  7. Aug 22
  8. Aug 26
  9. Aug 27
  10. Aug 28
  11. Aug 29
  12. Aug 30
  13. Aug 31
  14. Sep 1
  15. Sep 2
  16. Sep 3
  17. Sep 4
  18. Sep 5
  19. Sep 6
  20. Sep 7

More stories today

Open the live feed