Three OpenAI researchers who spent their careers warning about AI risks say the company fired them for doing exactly that.
Quick Take
- OpenAI fired researchers Jasmine Wang, Tomek Korbak, and Mikita Balesni, all tied to its safety and alignment work.
- The three wrote an open letter to OpenAI’s board claiming the firings created a “chilling effect” on safety speech.
- OpenAI says an internal investigation found the trio broke rules on handling sensitive information, calling it a “significant breach of trust.”
- The company insists it “has not and does not terminate any of our employees for raising concerns.”
A Public Fight Over Why Three Safety Experts Lost Their Jobs
OpenAI confirmed it parted ways with Jasmine Wang, Tomek Korbak, and Mikita Balesni after what it called a thorough investigation. All three worked on the company’s safety and alignment teams, the group tasked with catching dangerous behavior in AI models before it reaches the public. Their firing landed differently than a routine staff change because of who they were and what they had been saying before they were shown the door.
Within days, the researchers published a letter addressed to OpenAI’s board and its safety committees. They wrote that internal and external messaging around their firing had left former colleagues “afraid to speak and operate” the way they used to inside the company. That is a serious charge. It suggests a culture where raising red flags about powerful AI systems now comes with a career risk attached.
OpenAI’s Direct Rebuttal to the Retaliation Claim
OpenAI did not stay quiet. The company posted a response stating plainly, “We want to be very clear that these decisions were not about raising safety concerns or speaking out”. It added that its probe “uncovered a significant breach of trust beyond what’s outlined in the letter they published,” pointing to alleged mishandling of sensitive company information shared with an outside AI safety organization.
That is a specific, on-the-record denial, not a vague dismissal. OpenAI says the researchers “violated clear policies on handling sensitive information,” language repeated across multiple independent outlets covering the story. The company also noted that an internal memo from a research leader praised the three for their safety contributions while still insisting the terminations were unrelated to that work, an odd but telling detail that cuts against a simple villain narrative.
Why Both Stories Can Be Partly True
Here is the uncomfortable reality for anyone wanting a clean verdict. OpenAI says the researchers’ safety concerns and their alleged policy violation are not mutually exclusive. The company can genuinely value safety debate, as it claims happens “every day” internally, while still firing people over a separate confidentiality breach. Nothing in the public record proves OpenAI is lying about the investigation, and nothing proves the researchers invented a retaliation claim either.
What is missing is the paper trail. No termination letter, HR file, or board minutes have been made public to settle which story holds up. The researchers’ letter rests on their own account of a chilling effect. OpenAI’s defense rests on its own internal probe. Until someone releases documents, depositions, or a timeline showing which came first, the public is left weighing two competing, sourced claims rather than a proven fact pattern.
This Is Not OpenAI’s First Rodeo With This Exact Complaint
Skeptics have reason to raise an eyebrow given history. Back in 2024, OpenAI fired researcher Leopold Aschenbrenner, who later said he was let go after sharing a safety document with outside researchers, something OpenAI treated as a leak at the time. That earlier episode followed the same script: a safety-minded employee pushed out, framed by the company as a confidentiality matter. A pattern like that does not prove retaliation this time, but it does make blanket denials harder for the public to simply accept at face value.
For an industry racing to build systems smarter than anything humans have made before, this dispute matters beyond office politics. If genuine whistleblowers get silenced under the cover of confidentiality rules, the public loses its best early-warning system for AI gone wrong. If instead real security breaches get waved off as noble dissent, companies lose the ability to protect legitimate secrets. Both outcomes carry real cost, and right now, nobody outside OpenAI’s walls can say for certain which one happened here.
OpenAI fires three safety researchers, dispute goes public
OpenAI just confirmed one of the messiest internal disputes in its recent history: three researchers working on AI safety and model monitorability were dismissed, and the two sides are now telling very different stories.… pic.twitter.com/pJG4e3pfjT
— Eugenio Fierro (@EugenioFierro3) October 9, 2026
Congress and AI watchdogs have pushed for stronger whistleblower protections at frontier AI labs for exactly this reason, arguing that self-policing companies cannot be the only judge of their own safety culture. This case, with its dueling on-record statements and zero released documents, is a live test of whether that push has teeth or whether it remains just another talking point in Washington.
Sources:
insiderpaper.com, cnbc.com, wsj.com, newsweek.com, gizmodo.com, aiweekly.co
