OpenAI Defends Firing Three Safety Researchers Over Information Rules

OpenAI has defended its decision to fire safety researchers Jasmine Wang, Tomek Korbak and Mikita Balesni, saying an internal investigation found a significant breach of trust. The statement, reported by The Verge, names the three and rejects the idea that the dismissals were retaliation.

OpenAI said the researchers violated clear policies on handling sensitive information. It said the dismissals were not because the researchers spoke out about the company and AI safety concerns.

According to the same report, OpenAI said its investigation found breaches beyond what the researchers described in their own letter. It gave no further details.

The researchers published an open letter on Thursday asking OpenAI to be more transparent about the decision. In the letter, they said they believed they were fired for raising safety concerns. They also said their actions fit OpenAI's mission and the working norms at the time.

Earlier reporting had linked the dismissals to mishandling of confidential material. The Wall Street Journal reported on Oct. 1 that OpenAI parted ways with three researchers for alleged misconduct including sharing confidential company information with a third-party AI-safety entity. TechCrunch separately reported the company cut ties after an internal investigation found mishandling of sensitive company information.

CNN reported on Oct. 8 that the three said they had been fired and questioned the reasons given. The statement reported on Oct. 9 followed that public challenge.

The broader context here is a familiar trade-off in frontier labs, where frontier means the most capable and least fully understood systems. Safety work needs wide access to models, test results and incident reports to spot problems. Shipping products needs tight control over who sees what. It is a bit like hospital research, which needs patient data that the hospital must still keep private.

In my view, the core dispute is less about safety as a principle and more about process for sharing safety-relevant information. OpenAI says a policy line was crossed. The researchers say their actions fit the mission and norms as they understood them. When rules for disclosure are unclear or changing, both sides can believe their own account.

Looking at what this means for technical teams, the lesson is procedural. Clear labels for what is confidential, explicit approval steps for external sharing, and written review before publication reduce ambiguity. Without them, even well-intentioned contact with outside safety groups becomes a trust problem. With them, disagreements about risk can be checked.

Worth flagging here is that opacity creates problems on both sides. A company that cites additional breaches without describing them protects its investigation but leaves outside researchers unsure how to act. Researchers who point to the working norms of the time show how quickly those norms change as systems get more capable and commercial stakes rise. Without a shared, checkable record, neither position can be resolved.

Over the long arc, the technology will keep improving. That progress depends on safety researchers being able to raise concerns inside the company and, where appropriate, outside it, without breaking the controls that keep powerful systems and operational details contained. Getting that handoff right is unglamorous work. It is also where credibility is built.