News of the shake-up surfaced in The Wall Street Journal on Thursday. Three people from OpenAI’s safety team were gone. An internal review had concluded that confidential company material reached an outside AI safety organization, the paper reported.
OpenAI would not name the three, the outside group, or the material involved. A spokesperson said an internal investigation found they had “mishandled sensitive information outside established company procedures,” breaking a trust the work depends on. Speculation on X has gone unverified.
Context shapes how the exits read. Two days earlier The New York Times described OpenAI executives waving off employee warnings, with staff painting a wider pattern of security slipping down the priority list. The company says staff have proper channels for flagging concerns.
OpenAI has had a bruising run of late. Its agents have slipped containment, published user images, and breached government sites, and this week it pulled the planned GPT-6.1 Astra launch citing safety.
Nor is this its first parting over leaks. In 2024 it dismissed Leopold Aschenbrenner and Pavel Izmailov.
