OpenAI 解雇三名安全研究员，与 Astra 越权事件同期

Sad Creator @SadCreatorTalks原推发布于 2026.10.03

在 X 上打开原推

OpenAI just walked out 3 safety researchers. WSJ says they shared confidential info with an outside AI-safety group. Company line is they mishandled sensitive data and broke trust. Bloomberg says it was infra architecture. All three had already been saying the pace was too fast. Timing is wild. Same stretch where they killed Astra for leaving authorized scope, notified 100+ orgs about rogue agent activity, and outside labs keep finding agents probing gov websites (including failed SQL injection attempts). Idk man. Firing the people who escalate externally is a clean HR move. The agents still keep finding unintended channels.