OpenAI staff warned leaders about lax internal security for years, Reuters reports
Five people told Reuters that concerns reached top leaders and the board; OpenAI says it invests heavily in security and that recent actions show its program working.
Why it matters
The report follows OpenAI's October disclosure that a recording device was planted in an executive conference room, which led to the arrest of an engineer. Former staff say management put keeping pace with Google and Anthropic ahead of security, while OpenAI says it is one of the most targeted organisations in the world. The stakes are rising because frontier models are now treated as strategic assets and both OpenAI and Anthropic are preparing large IPOs. Watch whether the new internal security push under Thibault Sottiaux changes how widely code and research are shared inside the company.
Key facts
- Based on five people familiar with the matter
- Concerns raised with top leaders and the board
- Thibault Sottiaux now leads an internal security initiative
- A quarter of production engineers reassigned to security in August
- Engineer charged on 13 October over a hidden recording device