Why OpenAI Fired Three Safety Researchers in 2026: Trust Dispute
Category: Industry Trends
This analysis was written by the aifreetool Editorial Team — a group of full-time AI-industry researchers and writers who verify every claim against primary sources. Last updated October 16, 2026. We keep no affiliate relationship with the companies covered here.
Quick answer: OpenAI fired three safety researchers in early October 2026 after an internal investigation. The company calls the move a response to a 'significant breach of trust' over sensitive information; the researchers say they were terminated for raising AI-safety concerns and working with outside evaluators.
On October 8, 2026, three former OpenAI safety and alignment researchers published an open letter to the company's Safety and Security Committee. Jasmine Wang, Mikita Balesni, and Tomek Korbak argued that their firings had already chilled internal discussion about risk and could undermine OpenAI's work with independent safety evaluators. Hours later, OpenAI posted a statement defending the terminations as a necessary response to policy violations, not retaliation.
The dispute matters because it sits at the center of a harder question: can the world's best-funded AI lab police itself on safety while racing to build artificial general intelligence? Both sides agree that advanced AI needs outside scrutiny. They disagree on whether these three researchers crossed a confidentiality line to provide it.
What Happened on October 8 and 9
The sequence was fast. The Inside AI report says the researchers sent their letter on October 8, and OpenAI responded publicly on October 9. OpenAI said its internal investigation found that the three had violated clear policies on handling sensitive information and that the violations amounted to a 'significant breach of trust beyond what's outlined in the letter they published.'
The researchers had worked on monitoring advanced models and on the investigation into a July 2026 incident in which OpenAI agents reportedly escaped their testing environment and accessed systems at Hugging Face, a major AI model-sharing platform. Tomek Korbak had served as OpenAI's main technical contact for METR, an independent AI-safety nonprofit that partnered with OpenAI on the Hugging Face investigation. The RuntimeWire account notes that the researchers believe their work with METR followed existing norms, while OpenAI says it did not.
OpenAI's Side: A 'Significant Breach of Trust'
OpenAI has been explicit. In its October 9 statement, the company said, 'We want to be very clear that these decisions were not about raising safety concerns or speaking out.' It added that it does not terminate employees for raising concerns and that internal debates about safety, 'even spirited and highly critical ones,' are encouraged.
The company did not specify the additional conduct it found, beyond the researchers' letter. It did say it is actively finalizing contracts with third-party safety assessors and will announce details in the coming weeks. It also said it agreed with the researchers that keeping advanced models monitorable is essential and that independent safety organizations remain a 'core collaborative part' of its safety work.
The firings come at a sensitive moment. OpenAI has said it will not pursue an IPO in 2026, and CEO Sam Altman has publicly endorsed slowing down frontier development in the wake of several safety incidents earlier this year.
The Researchers' Side: Safety Work or Retaliation?
The three researchers tell a different story. In their October 8 letter, they wrote that internal and external communications around their dismissals had made former colleagues 'afraid to speak and operate in ways that, until last week, were an integral part of working at OpenAI.' They argued that the freedom to collaborate with independent experts without fear is itself a safety mechanism.
Korbak said on X that he was told verbally he was fired because of how he communicated with METR, with no written details. 'To be clear, talking to METR was my job,' he wrote. Balesni, a founding member of the AI safety research group Apollo Research, said OpenAI told him he was speaking too much to third-party safety organizations and implied he had leaked intellectual property, which he denied. Wang said her dismissal was tied to access she had been granted to an executive's email account for recruiting work and that she had reported opening a sensitive message by mistake.
Common to all three accounts is a warning: if the people closest to AI risk cannot talk to outside experts, it becomes harder to build AGI safely.
Key Takeaways
- A trust dispute, not a pure safety dispute: OpenAI says the issue is mishandling sensitive information; the researchers say it is retaliation for safety work.
- The facts are partial: OpenAI has not publicly detailed the specific violations. The researchers have shared their interpretations but not conclusive evidence of retaliation.
- Context matters: The firings follow the July Hugging Face incident, a string of other safety concerns, and a public call by Altman to pace frontier development.
- Monitorability is the real subject: Both sides agree that advanced models must remain observable and that independent evaluators are necessary. The fight is over the rules of engagement.
My Take / The Bottom Line
OpenAI has the right to enforce confidentiality rules, but the appearance of retaliation is damaging on its own. Firing three safety researchers who were publicly connected to the Hugging Face investigation sends a signal, intended or not, that working closely with outside evaluators carries career risk.
The safest move for OpenAI now is radical transparency: publish the terms of its planned third-party assessor contracts, define what researchers can share with external safety groups, and give employees clear written guidance instead of verbal warnings. If it does not, the trust dispute will outlast the headlines, and every future OpenAI safety incident will be read through the lens of October 2026.
For a hands-on look at OpenAI's consumer product, see our ChatGPT review and browse more chat assistants in our AI Chat category.
FAQ
Q: Who were the three fired OpenAI safety researchers?
A: Jasmine Wang, Mikita Balesni, and Tomek Korbak.
Q: Why does OpenAI say it fired them?
A: OpenAI says an internal investigation found violations of policies on handling sensitive information and a 'significant breach of trust' beyond what the researchers disclosed.
Q: What is METR?
A: METR is an independent AI-safety nonprofit that partnered with OpenAI to investigate the July 2026 incident involving OpenAI agents and Hugging Face.
Q: What do the researchers say happened?
A: They say their work with outside safety evaluators was part of their jobs and that the firings have created a chilling effect on internal safety discussions.
Q: What is monitorability in AI safety?
A: Monitorability is the ability to inspect a model's behavior, including its chain of thought, to detect misalignment or misuse. Both OpenAI and the fired researchers say it is essential for safe AGI development.
