OpenAI Fires Three Safety Researchers Over Signal Chats, Suspecting Leak Of Sensitive Information To Outside Expert
OpenAI fired three safety researchers over suspected leaks of model designs to an outside expert via Signal. They deny wrongdoing. The dismissals follow the arrest of a Saudi OpenAI engineer for allegedly bugging an executive meeting room.
OpenAI Fires Three Safety Researchers After Probe Into Signal Chats With Outside AI Expert | X
OpenAI has fired three safety researchers after an internal investigation into suspected leaks of sensitive information. The company's security team found that some of them had been messaging an outside AI-safety researcher on Signal, reportedly using disappearing messages. The three dispute the company's account. This comes just weeks after a Saudi national was arrested for allegedly bugging OpenAI's sensitive meeting room discussions.
Phones, then chats
The company suspected the researchers, Jasmine Wang, Tomek Korbak and Mikita Balesni, of passing details of OpenAI's internal model designs to outside researcher Ryan Greenblatt, including through disappearing Signal chats. According to the WSJ, OpenAI's security team called the three into meetings in September and asked them to unlock their personal phones. The company then asked to see their Signal chats.
Fired on October 1
The three were dismissed on October 1 for what OpenAI described as a significant breach of trust. The company has framed the matter as one of mishandling sensitive information.
What triggered the probe
The investigation reportedly began after a September article about Astra, a new OpenAI model said to use a design that improves performance while making its reasoning steps less visible to humans. OpenAI is said to have treated the report as a possible leak of internal design details and set out to trace its source.
OpenAI researchers push back
The three deny breaking any rules. They say they were trying to encourage more open collaboration across the AI field. Balesni has said he was dismissed for "speaking too much to third-party safety organisations." The WSJ describes the trio as having campaigned internally to keep the workings of AI systems visible and open to outside scrutiny.
A familiar flashpoint
The episode revives a recurring tension at OpenAI between internal safety advocacy and tight control over technical information. In April 2024, the company fired researcher Leopold Aschenbrenner over an alleged leak, which he disputed, saying he had only shared a benign brainstorming document with three external researchers.
RECENT STORIES
-
-
-
-
-