WSJ: OpenAI’s leak investigation focused on Signal chats between three fired safety researchers and an outside researcher

Safety researchers Tomek Korbak, Mikita Balesni and Jasmine Wang, fired by OpenAI in early October, were asked to open their personal phones during a leak investigation, The Wall Street Journal reported on October 22. It followed a September story in The Information saying OpenAI’s newest model, Astra, shows less of its reasoning. Security staff found some of them had used Signal’s disappearing messages in chats with outside researcher Ryan Greenblatt. The three were fired about a week later.

Per the WSJ, investigators suspected Greenblatt, formerly chief scientist at Redwood Research and now at METR, of being the leak’s source. On October 23 he wrote on X that OpenAI “has made false accusations about me.” He said he made no comment to The Information beyond a request about his own blog post and never read an internal memo with sensitive architecture details. He added that the Signal chats were with longtime professional contacts. Wang said she was told she had repeatedly opened an executive’s email, access she says she had asked to have removed a year earlier.

OpenAI says its investigation found the three broke its rules on sensitive information and that the firings were not about raising safety concerns. A spokeswoman said Signal use was not approved but was not the only reason. The three had denied being the leak’s source in their October 8 open letter; the WSJ report is the first to detail which contact drew the company’s suspicion.