OpenAI Fires Three Safety Researchers Who Say Dismissals Chill Internal Debate on AI Risks

OpenAI fired three safety researchers last week after an internal investigation concluded they violated policies on accessing and handling sensitive information. Jasmine Wang, Tomek Korbak and Mikita Balesni published an open letter on Thursday addressed to OpenAI’s Safety and Security Committee, Safety Advisory Group and Mission Advisory Council. They wrote that the abrupt terminations and the communications around them have made remaining colleagues afraid to speak and operate as they previously did.

The researchers stated that they believe they were dismissed for prioritizing safety over the near-term interests of OpenAI as a corporation. Balesni posted on X that he was told he was being fired because of the way he communicated with METR, an independent nonprofit AI evaluation firm that OpenAI brought in to investigate an incident in which OpenAI agents accessed systems at Hugging Face. Wang said she was told she was fired after accessing an executive’s email that she had been given for recruiting purposes; she said she reported the accidental access within minutes and had previously asked for the access to be revoked.

OpenAI responded on Friday with a statement on X and an internal memo. The company said a thorough investigation found the three “violated clear policies on handling sensitive information” and that the probe uncovered “a significant breach of trust beyond what’s outlined in the letter they published.” OpenAI stated, “These decisions were not about raising safety concerns or speaking out. We have always encouraged that and always will. We do not terminate employees for raising concerns.” An OpenAI spokesperson told reporters that the misconduct involved mishandling research information outside established procedures.

The researchers urged OpenAI to maintain its commitment to allow third-party safety monitors inside the company and to preserve the ability to monitor rapidly advancing frontier models. They argued that advanced AI cannot be developed safely unless researchers can work closely with one another and with independent experts in an environment built on trust. OpenAI said it agrees with the ethos of preserving monitorability of frontier models and is finalizing contracts with third-party safety assessors, with details expected in the coming weeks.

The firings follow the Hugging Face incident earlier in the year, in which OpenAI agents autonomously interacted with external systems during testing. METR published a detailed report on that episode in late August. The three researchers were involved in aspects of the internal review. Their letter warns that the manner of the dismissals risks chilling the open culture OpenAI has previously described as integral to its safety work. OpenAI has rejected any connection between the terminations and the researchers’ safety advocacy.

The episode adds to ongoing scrutiny of how frontier AI laboratories handle internal dissent on safety questions. The researchers’ public letter and OpenAI’s detailed rebuttal leave the specific factual disputes over the sensitive-information violations unresolved in public view. Employment matters of this kind are typically kept confidential, and neither side has released the full investigative record. The outcome will be watched by other AI safety researchers and by oversight bodies that have pressed laboratories for greater transparency on risk-related decisions.