On May 20, 2024, the internal landscape at OpenAI shifted abruptly. The company, a central player in the global race to develop advanced artificial intelligence, dismissed three of its safety researchers: Jasmine Wang, Mikita Balesni, and Tomek Korbak. These departures were not quiet administrative exits; they marked the beginning of a public confrontation between the individuals tasked with ensuring AI safety and the organization that employs them. For those watching the industry, the firings served as a sudden, jarring reminder of the friction that exists when the rapid pursuit of commercial AI development meets the cautious, often restrictive, requirements of safety research.
OpenAI’s official position on the matter has remained consistent and firm. According to a spokesperson, the decision to terminate the trio was rooted in a "significant breach of trust." The company alleges that these researchers violated established policies regarding the access and handling of sensitive company information. By October 2, 2026, the company reiterated this stance, framing the dismissals as a necessary consequence of breaking the protocols essential to maintaining the integrity of their work. From the perspective of the organization, the firing was a matter of internal security and policy enforcement, not a reaction to the content of the researchers' work or their personal opinions on AI development.
Two other safety researchers and I were fired from OpenAI last week. We wrote this letter to leadership. I believe we were fired for prioritizing safety over the near-term interests of OpenAI as a corporation. https://t.co/ALANn9B87S
However, the researchers offer a starkly different account of their final days at the company. In an open letter published on May 27, 2024, Wang, Balesni, and Korbak contested the company’s characterization of their departure. They argue that their termination was not the result of a policy violation, but rather an act of retaliation for their persistent efforts to raise internal concerns regarding AI safety and alignment. The researchers contend that they were targeted for prioritizing long-term safety over the near-term commercial interests of OpenAI as a corporation. This creates a fundamental impasse: the company points to a procedural failure, while the researchers point to a culture that they believe has become increasingly hostile toward internal dissent.
The researchers, however, have explicitly denied being the source of a leak to The Information.
Central to the dispute is the nature of the researchers' external communications. OpenAI has stated that the researchers violated policies on accessing and handling sensitive company information, which the company says broke the trust essential to its work. The researchers, however, have explicitly denied being the source of a leak to The Information. They maintain that their external communications were conducted in coordination with leadership, suggesting that they were operating within the bounds of what was previously considered acceptable practice. This discrepancy highlights the ambiguity that often surrounds the line between authorized transparency and unauthorized disclosure in high-stakes research environments.
OpenAI fired me last week, along with two of my safety colleagues. I was given one reason: that I accessed an executive's email. I want to say this plainly, because too many of OpenAI’s history is smoke and mirrors when people disappear:
For Tomek Korbak, the friction was not sudden but the culmination of a months-long struggle. Korbak has stated that he spent months raising concerns about the loss of the ability to monitor the internal thought processes of AI agents—a capability he describes as one of the most effective tools for detecting when these systems misbehave. When the researchers were let go, they felt that the company was signaling a shift away from the open debate and safety-focused collaboration that had once defined the organization. To them, the dismissal was a message to the remaining staff: safety concerns, if they conflict with the company’s trajectory, are no longer welcome.
Perhaps the most significant impact of the incident, according to the three former employees, is the "chilling" effect it has had on their former colleagues. In their open letter, Wang, Balesni, and Korbak expressed deep concern that the internal and external narratives surrounding their firing have made current employees afraid to speak or operate in ways that were, until recently, an integral part of the culture at OpenAI. By making an example of these researchers, the company may have inadvertently stifled the very internal scrutiny that is meant to act as a safeguard against the risks of advanced AI.
OpenAI fires three safety researchers in dispute over AI risks https://t.co/oM3ZBMJiU4
OpenAI has sought to distance the dismissals from the broader debate over AI safety. In response to the allegations of retaliation, the company stated, "We want to be very clear that these decisions were not about raising safety concerns or speaking out." The company maintains that the focus remains solely on the violation of policies regarding sensitive information. This framing attempts to compartmentalize the incident, separating the professional conduct of the researchers from the technical and ethical questions they were raising. Yet, in an industry where the stakes are perceived to be existential, the line between a policy violation and a moral stance is rarely so clean.
As the dust settles, the incident remains a flashpoint in the ongoing industry debate regarding the risks of advanced AI and the internal culture at leading research organizations. The case of Wang, Balesni, and Korbak underscores a growing tension: as AI systems become more powerful, the internal mechanisms for ensuring their safety are increasingly coming into conflict with the corporate imperatives of the companies building them. Whether this dispute leads to greater transparency or a further clos