OpenAI Fires Employees Amid Safety Dispute

Three former OpenAI employees are challenging their dismissals and raising concerns about the company’s AI safety practices after being fired last week. Mikita Balesni, Tomek Korbak and Jasmine Wang said their terminations were linked to safety advocacy and outside research work, while OpenAI said they mishandled sensitive information.

The former employees worked on teams focused on AI safety and ensuring models follow human intentions and values. Two of them were involved in investigating an OpenAI agent incident involving software company Hugging Face, according to the source report.

OpenAI has denied the allegations and said the company remains committed to safety reviews and third-party evaluations. CEO Sam Altman said on Sep. 12 that OpenAI would expand access to third-party evaluators who assess AI systems and developer practices.

Mikita Balesni said he believed the explanation for his dismissal was connected to his safety-related communications and involvement in external research reviews.

"In the exit call, I was told OpenAI no longer trusts me because I was speaking too much to third party safety organizations, implying I leaked company [intellectual property]. I never shared company IP," said Mikita Balesni.

Balesni said he was involved in reviewing the OpenAI agents' hack on Hugging Face. He also expressed concern that the dismissals could affect OpenAI’s relationship with safety organizations.

"If OpenAI has specific concerns, I invite them to write to us directly. I expect they will not, because our firing was pretextual," said Mikita Balesni.

Tomek Korbak served as the technical point of contact for an investigation by Model Evaluation and Threat Research and Redwood Research into the Hugging Face incident, according to the report.

"I was told verbally I was fired because of the way I communicated with METR. No details on what I said or did or when. No other reasons were given and nothing was put in writing," said Tomek Korbak.

The METR and Redwood Research review examined the scale of the incident and the ways AI agents acted in unintended ways. OpenAI said the three employees violated policies more than once and that the issues extended beyond their work with outside evaluators.

Jasmine Wang said she was dismissed after accessing an executive email account that she said had previously been available to her for work purposes. She argued the reasons given for the termination were unclear.

"The reasons that we were provided for our terminations are simply not adding up. We're hearing people are now being told vague rumors internally to discredit us," said Jasmine Wang.

Wang also raised concerns about the impact of the dismissals on employees who continue to work on AI safety issues. OpenAI said the terminations were related to policy violations involving sensitive information handling.

"The message to everyone still at OpenAI is clear: raise concerns or work closely with outside safety groups, and you could be next, without being told why," said Jasmine Wang.

OpenAI said it agreed with recommendations from the former employees regarding third-party evaluators and maintaining safety practices. The company shared an internal memo from an unnamed research leader that it said was provided before the employees posted about their dismissals.