Three former OpenAI researchers were dismissed for alleged mishandling of information, but they claim they were fired for raising safety concerns about artificial intelligence. The company on Friday rejected these claims, stating that the decisions were not related to safety issues.
Mikhael Sakhnov, Tomek Korbak, and Jasmine Wang wrote a joint open letter to OpenAI’s leadership, which Sakhnov posted on the X platform. "I believe we were fired for prioritising safety over OpenAI’s short-term interests," Sakhnov said. All three state that they had previously worked on safety or alignment, which is the practice of embedding ethical principles into technology.
A company spokesperson said on 2 October that their investigation had confirmed the individuals had mishandled sensitive information outside the company’s established procedures. OpenAI reiterated this in a note from its research leadership on Friday. "We want to be very clear that these decisions were not about expressing safety concerns or public advocacy," the note stated. It added that the internal investigation had revealed a significant breach of trust exceeding that outlined in the letter, and that the company stood by its decision not to continue their employment.
The former employees warned that their dismissals had made employees at the technology firm "afraid to speak and act in ways that were previously standard in open work at OpenAI". Korbak, reposting the letter, wrote: "I have been raising safety concerns for months that we are losing the ability to track what AI agents are thinking, one of our best tools for detecting when they behave inappropriately." He added: "I believe that is why I was fired."
Wang stated on the X platform: "We were not the first to be pushed out of OpenAI under questionable circumstances." She added: "If employees do not now push back against this move, I am worried we will not be the last."
The dismissals come at a time of growing debate about artificial intelligence safety and the potential risks advanced AI systems pose to society. OpenAI stated that it was finalising agreements with independent safety assessors and would release details in the coming weeks.
