← Back to feed

OpenAI fired three safety researchers on October 1, 2026, following an internal investigation that found they allegedly shared confidential company information with an external organization that evaluates AI models. On Thursday, October 8, the three — Jasmine Wang, Tomek Korbak, and Mikita Balesni — published an open letter denying the misconduct claims and warning that their dismissal is having a measurable chilling effect on OpenAI’s safety culture.

The letter was addressed to OpenAI’s Safety and Security Committee, Safety Advisory Group, and Mission Advisory Council.

What OpenAI Said

OpenAI said the researchers violated company policy by “accessing and handling sensitive company information” outside established procedures. The company said an internal review found they shared information tied to their work with an external organization. It did not name the organization.

An internal memo attributed to a research leader, shared with TechCrunch, praised the three researchers’ safety contributions while rejecting a retaliation framing: “I want to be very clear that 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.”

OpenAI has not formally responded to the open letter.

What the Researchers Said

In the letter, Wang, Korbak, and Balesni denied sharing information outside the scope of their jobs. They also explicitly denied involvement in a leak to The Information about less-monitorable architectures in OpenAI’s newest models — specifically, features of GPT-6 Astra that make chain-of-thought reasoning harder to monitor from the outside.

Their central argument is structural: safety researchers at frontier labs cannot do their jobs without close collaboration with independent external experts, and that collaboration requires the freedom to share technical context. Making that category of work grounds for termination — abruptly and without clear prior notice that the rules had changed — leaves researchers unable to judge what is permissible.

“We have become concerned that internal and external communications around our firing have made our former colleagues afraid to speak and operate in ways that, until last week, were an integral part of working at OpenAI,” the letter states.

“Those of us who work on safety see risks before anyone else, and we rely on close collaboration with outside experts to work out how to address them. The freedom to do so without fear, and to have well-defined internal procedures that enable this work, is itself an essential safety mechanism.”

Why This Matters

This is the second significant rupture over safety culture at OpenAI in roughly a year. The company has restructured its governance multiple times, converted to a public benefit corporation, and repeatedly stated that safety and commercial goals are aligned. Researchers being fired for work they describe as integral to the safety mission — and then warning publicly that the firings are suppressing dissent — is exactly the scenario critics of frontier AI self-governance point to when arguing that internal safety functions cannot be the primary accountability mechanism.

The reference to GPT-6 Astra’s chain-of-thought monitorability is significant. That capability — the ability for auditors, red-teamers, or safety organizations to observe a model’s reasoning process — is one of the few technical levers external evaluators have. If architectural choices make CoT reasoning harder to inspect, and safety researchers discussing that concern externally are dismissed, the implied accountability loop becomes very short.

The three researchers held senior roles in safety research. Their departure removes institutional knowledge about how OpenAI models behave in high-stakes conditions. Whatever the facts of the specific misconduct allegation, that is a material loss.