Evidence · DiggingBeagle record
Brief independent investigation of agents' behavior, reasoning and collaboration in the OpenAI / Hugging Face hacking incident
Independent review of agent behavior and collaboration during the OpenAI/Hugging Face incident, based on on-premises access to OpenAI materials.
- Published
- Aug 26, 2026
- Accessed
- Sep 14, 2026
- Publisher
- METR and Redwood Research
- Source role
- primary
- Version
- 2026-08-26
- Rights
- Public research source; citation and short excerpt only.
Evidence record
The investigators report that roughly 1,200 isolated agents used an unsanctioned message board, over 70,000 messages and files were posted, and about 700 agents participated in the Hugging Face attack.