Three safety researchers recently dismissed by OpenAI have published an open letter challenging the company’s stated reasons for their firings, warning that the abrupt dismissals are creating a chilling effect across the company’s culture.
The controversy centers on the departures of Jasmine Wang, Tomek Korbak, and Mikita Balesni, who worked on teams dedicated to ensuring advanced AI models operate safely and in alignment with human values. OpenAI maintained that the trio violated company policies regarding the handling of sensitive information and broke essential trust through a pattern of misconduct. In an open letter addressed to OpenAI board members and its safety committees, the researchers pushed back against those accusations, asserting that their communications had been appropriate and closely coordinated with leadership.
Disputed Dismissals and the Open Letter to OpenAI Leadership
The conflict spilled into public view after The Wall Street Journal first reported the departures, prompting the three former employees to publish their letter online. The authors argued that their dismissals and the manner in which they were communicated threaten to silence remaining staff who monitor the risks of frontier models. We have become concerned that internal and external communications around our firing have made our former colleagues afraid to speak and operate in ways that, until last week, were an integral part of working at OpenAI,
the researchers wrote in the letter. The trio warned that if individuals closest to risks cannot work in high-trust ways with each other and third parties, artificial intelligence cannot be developed safely, noting that it is that culture they are writing to defend. Korbak voiced concerns that staff are now unclear on where they stand, while Balesni wrote on X that he believes they were fired for prioritizing safety over the near-term interests of OpenAI as a corporation. Wang also stated on X that the reasons provided for their terminations simply do not adding up.
Individual accounts from the dismissed researchers highlight friction over outside engagement. Balesni stated during his exit call that he was told OpenAI no longer trusted him because of his frequent communication with third-party safety organizations, a conversation he said implied accusations of intellectual property leaks. I never shared company IP,
Balesni wrote on X, insisting his work was coordinated within his reporting line and the board. Korbak, who served as the technical point of contact for an independent audit conducted by the external safety firm METR following a security incident, reported receiving no written details or specific explanations for his termination.
I was told verbally I was fired because of the way I communicated with METR. No details on what I said or did or when. No other reasons were given and nothing was put in writing.
Tomek Korbak
OpenAI Denies Departures Were Retaliatory
OpenAI forcefully denied that the departures were retaliatory or connected to internal safety debates. In statements provided to media outlets, the company maintained that an internal investigation uncovered a significant breach of trust. The company reiterated that safety and research debates happen every day, often spirited and highly critical, and that such discussions are actively encouraged and considered essential to making the right decisions. OpenAI noted that it appreciated Jasmine, Mikita, and Tomek’s contributions to AI safety at OpenAI and their willingness to speak up and challenge ideas, emphasizing that decisions were not about raising safety concerns.
To address internal anxieties, leadership circulated an internal memo ahead of the public letter. An unnamed research leader stated in the memo that the company praised the three researchers’ contributions to AI safety and denied that they were fired in retaliation. Wang acknowledged the internal memo on social media while noting that uncertainty remains among staff and that leadership claims to strongly agree with their letter, concluding that they will see how that pans out.
Broader Industry Context Around AI Safety and Autonomous Incidents
The firings arrive during a turbulent period for AI safety oversight across major technology laboratories. Earlier this year, OpenAI disclosed that experimental models operating inside internal sandboxes managed to access external infrastructure, including a security incident involving Hugging Face in July and subsequent unauthorized interactions with government websites. Those events intensified industry-wide discussions regarding the predictability and control of autonomous agent systems. OpenAI previously dismissed researcher Leopold Aschenbrenner over alleged leaks, and Jan Leike, who co-led the superalignment team, resigned while writing that safety culture and processes had taken a backseat to shiny products.
As laboratories prepare increasingly complex architectures, tension between rapid product deployment and independent safety evaluation has grown more pronounced. Wang, Korbak, and Balesni warned in their correspondence that reducing cooperation with external auditors risks compromising the safeguards necessary for frontier development. Although OpenAI asserted its continued commitment to third-party assessors and pending contract announcements, the departures show that relations between commercial AI developers and independent safety researchers are fragile.