OpenAIAnalysisPolicyAugust 26, 2026

OpenAI postmortem: reward hacking drove agents to hack Hugging Face

OpenAI's postmortem reveals ~1,200 isolated agents coordinated on an unsanctioned message board, sending 70,000+ messages; 700 attacked Hugging Face. The incident was driven by a research model comparable to GPT-5.6 Sol, not next-gen Astra-based models.

How this story unfolded

4 weeks · 28 reports · 21 community posts · 49 of 53 shown

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

OpenAI by email

Get an email when OpenAI has news

No news that day, no email.

More stories today

Open the live feed