OpenAI’s stated reason for the firings

OpenAI said an investigation into the firings found that three safety researchers committed “a significant breach of trust.” In a post on X on Friday, the company said Jasmine Wang, Tomek Korbak and Mikita Balesni had been dismissed for violating “clear policies on handling sensitive information.”

The company presented the decision as a matter of policy and trust, rather than a response to the researchers raising safety concerns. OpenAI said its investigation had concluded that the researchers’ conduct crossed a line in the way sensitive information was handled. It described the relevant rules as clear and characterized their violation as significant.

OpenAI stood by the decision after the investigation, maintaining that the firings were based on the alleged policy breach. Its stated explanation focused on the handling of sensitive information and the company’s assessment that the conduct amounted to a breach of trust.

The company’s response to safety concerns

OpenAI said the researchers’ dismissals were not connected to their speaking out about safety issues. Instead, the company framed the decision as a matter of trust and compliance with its rules for handling sensitive information.

In a post on X, OpenAI said Jasmine Wang, Tomek Korbak and Mikita Balesni had violated “clear policies on handling sensitive information.” The company also said an investigation found that they had committed “a significant breach of trust.”

OpenAI’s response was aimed at separating the firings from the researchers’ safety-related views: it maintained that the decision concerned their conduct and the handling of information, not their comments about safety. The company stood by that explanation as it defended its decision to dismiss the three researchers.

How the decision was announced

OpenAI explained its position in a post on X on Friday, saying the three researchers were dismissed for violating “clear policies on handling sensitive information.” The company described the decision as the result of an investigation that found a “significant breach of trust.”

In the post, OpenAI also stressed that the firings were not about the researchers speaking out about safety. Instead, it framed the matter as a breach of its rules for handling sensitive information. That distinction was central to the company’s public explanation: it said the decision concerned the researchers’ conduct under those policies, rather than their views or comments on safety.

The announcement named Jasmine Wang, Tomek Korbak and Mikita Balesni. By laying out its rationale on X, OpenAI publicly reaffirmed its decision to dismiss them and pushed back against the suggestion that raising safety concerns was the reason. The company said its position followed an investigation and rested on what it characterized as a serious breach of trust.

The researchers affected

The three safety researchers dismissed by OpenAI were Jasmine Wang, Tomek Korbak and Mikita Balesni. The company said they had violated “clear policies on handling sensitive information,” describing the issue as a significant breach of trust.

OpenAI said the researchers had shared internal information with people outside the company without authorization. The information included details about internal deliberations, according to the company. OpenAI also said it had found no evidence that the researchers acted with malicious intent, and described the information involved as limited in scope.

The company’s account was disputed by people familiar with the matter. They told The Verge that the researchers had shared concerns about OpenAI’s safety practices with members of the board, and said the internal investigation focused on their communication with board members. OpenAI denied that the dismissals were connected to raising safety concerns.

Wang, Korbak and Balesni were among OpenAI employees who had raised concerns about the company’s approach to safety. Their dismissals brought renewed attention to the question of how the company handles internal criticism and sensitive information.