October 12, 2026 – Three OpenAI safety researchers say they were fired for doing their jobs. OpenAI says they were fired for breaking the rules. Which account is right isn’t yet clear, but the dispute has put a hard question in front of the AI industry: what happens when a safety researcher’s work with outside overseers collides with a company’s confidentiality policies?

What happened

In recent weeks, OpenAI fired Jasmine Wang, Tomek Korbak, and Mikita Balesni, who worked on safety and alignment. On October 8, the three published an open letter to OpenAI’s leadership disputing the grounds for their dismissals. The next day, OpenAI responded on X that a thorough investigation had found they violated clear policies on handling sensitive information, and that it uncovered “a significant breach of trust beyond what’s outlined in the letter.”

The two accounts

The researchers say the firings were really about safety. Wang says she was told she accessed an executive’s email, which she says she had for recruiting purposes. She wrote on X that the reasons given for the terminations “are simply not adding up.” Korbak says his dismissal related to how he communicated with METR, an independent group that evaluates AI models. Balesni wrote that the three were fired for “prioritizing safety over the near-term interest of OpenAI as a corporation.”

OpenAI denies that. An internal memo, shared with TechCrunch, praised the researchers’ contributions to safety and rejected the idea that the firings were retaliation. The company says it continues to encourage internal challenge and safety advocacy.

Why it matters beyond one company

Whatever the facts turn out to be, the episode touches a real fault line. Labs say they want outside scrutiny of their models, and some have made public commitments to work with third-party evaluators. But that work requires sharing sensitive information, and where the line sits between permitted and forbidden sharing is often set by the company itself.

The researchers argue that firing people in this gray area will make remaining employees afraid to engage with outside groups. OpenAI’s position is that its policies exist to protect sensitive material, and that enforcing them is not an attack on safety work.

What we don’t know

No independent party has yet reviewed the evidence. The company hasn’t published the specifics of its investigation, and the researchers haven’t shared everything either. Until more comes out, readers should treat both accounts as claims, not findings.

Sources: CBS News, TechCrunch, CNN