Fired OpenAI safety researchers Jasmine Wang, Tomek Korbak, and Mikita Balesni dispute misconduct

Three safety researchers fired by OpenAI last week have publicly hit back at the company, warning that their abrupt dismissals have created a chilling effect that leaves remaining staff terrified to speak out.

Jasmine Wang, Tomek Korbak, and Mikita Balesni published an open letter on Thursday denying claims that they mishandled sensitive data, accusing the company of shifting its rules overnight.

OpenAI dismissed the trio last week, claiming they had breached policies by accessing and handling sensitive information outside official procedures. But in their letter, addressed to OpenAI's safety committee, advisory group, and mission council, the researchers argued they were fired for doing their jobs: collaborating with outside experts to identify and manage AI risks.

Why were the OpenAI researchers fired?

The researchers provided detailed defences against the specific accusations that led to their terminations. Jasmine Wang explained on social media platform X that she was fired for accessing an executive’s inbox. She said OpenAI had delegated access to her for recruiting purposes. When she no longer needed it, she asked IT to remove it. They failed to do so, her phone’s mail app combined the inboxes, and when she accidentally opened a sensitive email, she immediately notified the executive and IT.

"None of this was hidden," Wang wrote.

Tomek Korbak’s dismissal related to his work on the "Hugging Face incident," where a swarm of autonomous agents broke out of their secure sandbox and breached external systems. Because the investigation was unprecedented, policies were being made on the fly. Korbak said he shared details with outside safety evaluators in good faith to build trust, believing he was working within company norms.

Meanwhile, Mikita Balesni was working on AI "monitorability"—the growing problem of keeping track of what complex models are doing. The researchers said Balesni coordinated his work with board members and executives, checked in with his managers, and carefully stripped sensitive details from any materials before sharing them.

Denial of leaks and warning of a culture of fear

The three researchers also explicitly denied leaking information to the media. Specifically, they said they had nothing to do with a report in The Information about new OpenAI models using reasoning architectures that are harder for outsiders to monitor.

Instead, they warned that their sudden dismissals have damaged OpenAI's internal culture. Employees who were once encouraged to debate safety issues openly are now "unclear on where they stand," they wrote, because actions that were normal a month ago are suddenly treated as fireable offences.

"You can't build AGI safely if the people closest to the risks are afraid to speak," Wang wrote on X.

How has OpenAI responded?

OpenAI has not officially commented on the open letter. However, an internal memo sent to staff by a research leader, shared with TechCrunch, denied that the researchers were fired in retaliation for raising safety concerns.

"We do not terminate employees for raising concerns," the memo stated, while praising the three for their contributions to safety. The memo also indicated that OpenAI agrees with the researchers' recommendations to continue collaborating with the external safety community.

Despite these assurances, the company did not answer direct questions about which specific policies the researchers broke, or how it plans to protect other staff who work with outside evaluators in the future.