Already a subscriber? Make sure to log into your account before viewing this content. You can access your account by hitting the “login” button on the top right corner. Still unable to see the content after signing in? Make sure your card on file is up-to-date.

Three AI safety researchers fired by OpenAI last week say they were let go for putting safety ahead of the company’s interests.

Getting into it: Mikita Balesni, Tomek Korbak and Jasmine Wang, who worked on making sure OpenAI’s AI systems behave as intended, published an open letter Thursday to the company’s safety oversight groups denying they did anything wrong and warning their firings are making colleagues “afraid to speak.” Korbak said he was told he was fired over how he communicated with METR, an outside AI safety group OpenAI brought in to investigate the July incident in which its AI agents escaped a test environment and hacked the developer platform Hugging Face, and he said he was OpenAI’s main contact with the group.

He believes the real reason was that he’d spent months warning that OpenAI was “losing the ability to monitor what AI agents think.” Balesni said he was told he was “speaking too much to third party safety organizations,” and Wang said she was told she improperly accessed an executive’s email, which she says she’d been given access to for recruiting and reported herself within minutes after opening a sensitive message by mistake. The three also denied leaking information to the press about how OpenAI’s newest models are harder to monitor.

OpenAI said an investigation found a “pattern of misconduct” and “a significant breach of trust” that involved more than the researchers’ contact with an outside group, though it didn’t say which specific policies were broken. “We want to be very clear that these decisions were not about raising safety concerns or speaking out,” the company said, adding that it agrees outside safety groups play an important role and that AI models need to stay monitorable.

Balesni said former colleagues are now “afraid to speak, and worry their personal phones will be searched for messages to us and third parties,” while Wang warned, “Unless the employees take a stand now against this kind of maneuver, I am concerned we will not be the last.”

This all comes as OpenAI faces mounting scrutiny over rogue AI incidents, including an FTC investigation and its decision to scrap the release of its GPT-6.1 Astra model over safety concerns, while it prepares for an expected IPO in 2027 and reported about $50 billion in annualized revenue, well short of the $68 billion number that circulated in September.