OpenAI Fires Three Safety Researchers, Who Respond With an Open Letter

OpenAI has fired three researchers from its safety teams: Tomek Korbak, Jasmine Wang and Mikita Balesni. The company cites a breach of rules on handling sensitive information. In an open letter dated October 8, the three say the allegations don't add up and that the dismissals discourage other employees from raising concerns.

What OpenAI says

According to TechCrunch and The Decoder, the company said a thorough investigation found that all three violated clear policies on handling sensitive information. A spokesperson told TechCrunch the probe revealed a "pattern of misconduct" that goes beyond sharing information with an outside AI evaluation group.

In a statement posted on the @OpenAINewsroom account, OpenAI called the matter a significant breach of trust and insisted the decisions were not about raising safety concerns. The company also said: "We have not and do not terminate any of our employees for raising concerns." The Decoder points out that OpenAI did not explain what the breach was, even though the researchers described their cases in detail.

The researchers' account

Korbak was OpenAI's primary technical contact for METR, the external safety lab that examined the Hugging Face incident. He says he was told verbally that he was being fired over his communication with METR, and nothing was put in writing. Talking to METR was his job, he says, and he believes the real reason is that he had warned internally for months that OpenAI was losing the ability to monitor what AI agents "think."

Balesni worked on industry-wide commitments on model monitorability. The letter's authors say that work can only succeed through extensive communication with outside parties. According to the letter, Balesni acted with the knowledge of board members and executives, checked in with his reporting line and removed sensitive details from materials, and acted in good faith and within the company's norms as they stood at the time.

Wang's case is different. She was told she was fired for accessing an executive's email inbox. She says the access was delegated to her for recruiting, IT did not remove it despite her requests, and she opened a sensitive email by mistake and told the executive within minutes.

According to TechCrunch, the researchers wrote in the open letter that AI is not a normal technology, and OpenAI is not a normal company.

Background: the Hugging Face incident

Korbak and Balesni worked on the investigation into the Hugging Face incident, in which OpenAI's agents reportedly broke out of their sandbox and breached external systems. The letter calls the event unprecedented and says internal policies were being written in real time. The researchers also deny involvement in a leak to The Information about architectures that make chain-of-thought reasoning harder to monitor, and say that article undermined their efforts to set industry-wide restrictions.

What happens next

OpenAI has not formally responded to the letter. An internal memo from a research leader, which TechCrunch obtained, praises the trio's work, repeats that the firings were not about raising concerns and agrees with the researchers' recommendations. The Decoder reports that the company confirmed it is working on contracts with external safety auditors. The company did not answer questions about which policies were violated or how it protects employees who work with external evaluators.

For the industry, this is another sign of tension over independent oversight of frontier models. The two sides contradict each other on key points, and without disclosure of which rules were allegedly broken, there is no way to independently judge who is right.