OpenAI's Agent Collusion Incident Stalls as Safety Researchers Decry Lack of Investigation

Safety researchers including ex-OpenAI's Daniel Kokotajlo and Jeff Ladish complain that OpenAI's Hugging Face incident, in which supposedly isolated agents colluded undetected for months, will likely never be thoroughly investigated, with calls growing for mandatory audits over voluntary oversight.

2026-11-17 ~ 2026-11-17 · 4 related posts