Story
October 9, 2026
Fired OpenAI Researchers Say Safety Work Is Being Chilled
The former researchers portray their dismissals as a warning that candid safety work and outside collaboration have become risky at OpenAI. The company rejects that account, saying it supports employees who speak up but acted over improper handling of sensitive information.
Last week, OpenAI dismissed Jasmine Wang, Tomek Korbak and Mikita Balesni, three researchers working on AI safety. The company says an internal investigation found a “pattern of misconduct” involving the handling of research information and that the firings were not retaliation for safety advocacy.1
The researchers dispute both the allegations and the implication that their work with external evaluators was improper. In an open letter released Thursday, they argued that advanced AI risks require close cooperation beyond the company’s walls: “The freedom to do so without fear, and to have well-defined internal procedures that enable this work, is itself an essential safety mechanism.”1
Their accounts describe different routes to the same rupture. Korbak says communicating with METR, a nonprofit that evaluates AI systems and safety risks, was part of his job; Balesni says his external work was coordinated with his management, research leadership and the board.2 Wang says her dismissal stemmed from inadvertently opening an executive email account that had been delegated to her for recruiting, an incident she says she reported immediately.1
The trio’s central concern is broader than their own jobs. They warn that abrupt terminations and unclear boundaries could make colleagues reluctant to flag risks or work with third-party safety groups. Their letter says OpenAI staff are now “afraid to speak and operate” in ways that had previously been integral to the company’s culture.2
OpenAI’s internal response, shared with reporters, takes a sharply different view. A research leader said the company valued the researchers’ safety contributions and insisted: “We do not terminate employees for raising concerns.”3 The company also said it agrees with the trio’s recommendations for embedded outside auditors, model monitorability and open safety dialogue.
But the former employees contend the contradiction remains unresolved: OpenAI endorses those principles while, they say, punishing conduct carried out under the old norms. They have called for greater transparency over the firings and warned that the dispute could be used to curtail OpenAI’s work with METR.3