Three Fired OpenAI Safety Researchers Say They Were Let Go for "Prioritising Safety". OpenAI Says It Was Mishandled Information

The Three-Question Framework

What Happened?

Three former OpenAI safety or alignment researchers, Mikita Balesni, Tomek Korbak and Jasmine Wang, published an open letter saying they were fired for raising safety concerns. OpenAI told the BBC on 2 October that its investigation "confirmed that these individuals mishandled sensitive information outside established company procedures", and repeated that on Friday 9 October in a note from research leaders.

Why It Matters

The dispute turns on one technical concern. Korbak says he had raised for months that "we're losing the ability to monitor what AI agents think", which is the ability to read model reasoning to catch misbehaviour. The case is a test of whether people who work on that kind of oversight can speak up inside a lab. The BBC report does not let us judge who is right.

What Could Happen Next

OpenAI says it is finalising contracts with third-party safety assessors and will announce details in the coming weeks. That announcement is the next checkpoint. The former employees say staff are now "afraid to speak" and want others to take a stand. No legal action is reported in the source.

Executive Key Takeaways

The three researchers are Mikita Balesni, Tomek Korbak and Jasmine Wang.

OpenAI says they mishandled sensitive information outside company procedures and that the firings were not about safety.

Korbak links his dismissal to months of warnings about losing the ability to monitor what AI agents think.

OpenAI is finalising contracts with third-party safety assessors and will give details in coming weeks.

Both accounts are claims. The BBC did not publish the underlying evidence.

On 2 October a spokesperson told the BBC that OpenAI's investigation had confirmed the three had mishandled sensitive information outside established company procedures. Then the former employees published an open letter to OpenAI leadership, and Balesni wrote on X: "I believe we were fired for prioritising safety over the near-term interests of OpenAI as a corporation."

OpenAI replied on Friday 9 October with a note from research leaders: "We want to be very clear that these decisions were not about raising safety concerns or speaking out." It added that its internal investigation found "a significant breach of trust beyond what's outlined in the letter they published" and that it stands by the decision to end their employment.

The letter says the firings left remaining staff "afraid to speak and operate in ways that, until last week, were an integral part of working at OpenAI". Wang wrote that they were "not the first to be pushed out of OpenAI under suspicious circumstances" and warned that unless employees take a stand "we will not be the last".

The technical issue behind Korbak's claim is the monitoring of agent reasoning. Labs and researchers use the visible reasoning of a model as one signal for catching when an agent is doing something it should not. If that signal becomes unreliable, one of the main tools for catching misbehaviour weakens. The BBC quotes Korbak calling it "one of our best tools for catching when they misbehave". OpenAI has not said, in the quoted material, whether it disputes that concern.

OpenAI also said it is finalising contracts with third-party safety assessors and will announce details in the coming weeks. The timing matches a wider public debate on independent audits; the same week Microsoft's president called for outside evaluators to sign off on AI products.

What we do not know: what information was shared, with whom, and why. The BBC article gives no detail on either side's evidence, and neither claim can be checked from public documents. Readers should treat both accounts as statements by interested parties.

The Three-Question Framework

What Happened?

Why It Matters

What Could Happen Next

Executive Key Takeaways

Primary Sources & Citations