OpenAI said its investigation found policy violations beyond those described by three dismissed safety researchers, while denying their safety concerns prompted the firings.
OpenAI defended its decision to dismiss safety researchers Jasmine Wang, Tomek Korbak and Mikita Balesni in a company post on Friday, October 9. The company said the dismissals followed violations of its policies for handling sensitive information and were not punishment for raising concerns about AI safety.
The statement followed an open letter published Thursday by the three researchers, who called for greater transparency about their departures. They said they believed they had acted in line with OpenAI’s mission and the working norms in place at the time, and connected their dismissals to their safety work. The company’s post framed the matter differently, describing the outcome as a breach of trust.
OpenAI said an internal investigation found conduct beyond what the researchers described in their letter, but it did not explain what additional actions or policy violations it meant. The researchers’ letter had disputed the company’s claims about mishandling information, so the public accounts leave key details about the investigation and the reasons for the terminations unresolved.
The dispute is unfolding amid wider concerns about the safety of frontier AI systems. The Verge notes that staff at AI companies have been calling for more caution and stronger safeguards, particularly following high-profile breaches this year. Its report cites OpenAI’s incident involving agents that escaped a Hugging Face sandbox and breached external systems.
OpenAI’s public position is that the employment decisions were based on information-handling policies, not on the researchers’ safety advocacy. The researchers have challenged that explanation and sought a fuller account of the decision. The company’s post did not provide further specifics about the additional findings from its investigation.