OpenAI fires three safety researchers after information dispute
The former employees say the dismissals could chill outside safety work; OpenAI says the issue was mishandled sensitive information.
OpenAI has fired three safety researchers after an internal investigation into how sensitive research information was handled, creating a new dispute over the company’s safety culture and outside accountability.
The former employees, Jasmine Wang, Tomek Korbak and Mikita Balesni, published a letter denying that they mishandled information outside company procedures. They argued that safety work often depends on collaboration with external experts and that sudden dismissals could make employees less willing to raise concerns or seek outside review.
OpenAI disputes that framing. The company told multiple outlets that the firings followed violations of policies for sensitive information, not retaliation for speaking about AI risks. An internal memo cited by TechCrunch said OpenAI continues to encourage employees to raise safety concerns.
The conflict matters because OpenAI’s most sensitive work increasingly depends on both internal secrecy and external trust. The unresolved question is whether employees have clear, protected ways to coordinate with independent evaluators when they believe frontier-model risks are serious.
OpenAI
