OpenAI Checked Fired Researchers' Phones for Signal Chats

OpenAI's security team asked three AI safety researchers to open their personal phones during a leak investigation, and the three were fired about a week later, according to a Wall Street Journal report that one of them confirmed. "I was told that there was a breach of trust because I interacted with a third party through unsanctioned channels," said Tomek Korbak, one of the three.

What the investigation found

Investigators named AI safety researcher Ryan Greenblatt as the suspected leaker. He denies it. The inquiry followed a 1 September story in The Information, which said Astra, OpenAI's latest model, used a new technique that made it more powerful but showed less of its thinking. Chief scientist Jakub Pachocki called the reporting "confused".

On the phones, security officers saw that some researchers had turned on disappearing messages in Signal chats with Greenblatt. The company then called Korbak, Mikita Balesni and Jasmine Wang into meetings at 2pm on the same day and fired them. Balesni and Korbak say leaders cited the disappearing messages as a reason. All three say they never shared confidential information outside the company. This article has not independently verified their account.

Competing accounts

OpenAI says an internal investigation found the three broke its policies on sensitive information, and that the firings were "not for raising safety concerns or speaking out". A spokeswoman said Signal use was not approved but was not the only reason. Korbak calls disappearing messages a "basic security best practice" in Silicon Valley.

Wang says she was told she was fired for repeatedly opening an executive's email. She says she had asked OpenAI to remove that access nearly a year earlier, opened one sensitive email by mistake and reported it at once. OpenAI said it had no problem with the access itself.

Greenblatt, who now works at METR, wrote on X that OpenAI "has made false accusations" about him. He says he did not help leak architecture details and saw only a version of an internal memo approved for sharing with him.

Why it matters

The three had spent the summer pushing to keep AI reasoning, the "chain-of-thought", visible, and in an October letter urged OpenAI's board to work with outside safety groups. "Talking to auditors was too much safety, disappearing messages were too much security," Korbak wrote on X. The dispute now tests how AI labs balance secrecy with outside oversight.