Three former OpenAI safety researchers say they were fired after raising concerns that the company was not doing enough to monitor the behaviour of AI agents, while OpenAI says they were let go for breaching company policy by mishandling information.
Safety researchers Mikita Balesni, Tomek Korbak and Jasmine Wang made the accusations on Thursday, a week after their dismissals. In social media posts and an open letter, they warned that their firings would discourage other employees from speaking up about risks associated with frontier AI.
"I believe we were fired for prioritizing safety over the near-term interests of OpenAI as a corporation," Balesni wrote on X. Korbak said he believed he was terminated for raising concerns that OpenAI was "losing the ability to monitor what AI agents think, one of our best tools for catching when they misbehave".
The three said their dismissals would create a culture of fear that undermines AI safety work and weakens third-party accountability. They denied violating company policy, stating that their engagement with external safety experts was consistent with the mandate of their roles.
OpenAI said in a statement that it had uncovered "a significant breach of trust" by the former researchers that went beyond what was described in their letter, and that the decisions "were not about raising safety concerns or speaking out". The company said it actively encourages spirited debate on safety and research internally.
"We want to be very clear that these decisions were not about raising safety concerns or speaking out," OpenAI said. "Safety and research debates happen every day at OpenAI, often spirited and highly critical. We actively encourage these discussions and consider them essential to making the right decisions. We cannot do the work in front of us without it."
The trio's open letter defended what they called a once-open culture at OpenAI, warning that if people closest to the risks cannot work in high-trust collaboration with each other and third parties, AI cannot be developed safely.
Analysis
Why This Matters
- The firings could deter OpenAI employees and staff at other AI labs from raising internal safety concerns, potentially allowing risks to go unaddressed.
- The incident highlights ongoing tensions between rapid commercial deployment of AI and the need for rigorous independent oversight.
- If the former employees pursue legal action or whistleblower complaints, it may set precedents for protections for AI safety researchers.
Background
OpenAI, the developer of ChatGPT and other advanced AI systems, has faced internal conflicts over safety culture before. In 2024 and 2025, several senior safety researchers left the company, citing concerns that commercial pressures were being prioritised over caution. The company has since restructured its safety oversight processes, but critics argue that internal whistleblowers still face retaliation. The three former employees were part of a team focused on monitoring the internal reasoning of AI agents, a technique used to detect potentially harmful behaviour before deployment.
Key Perspectives
Former employees (Balesni, Korbak, Wang): They argue they were fired specifically for advocating for stronger safety measures and for consulting external experts, which they say was within their job description. They warn that abrupt terminations will chill internal debate and make it harder for independent safety organisations to hold AI companies accountable.
OpenAI: The company insists the dismissals were for serious misconduct related to mishandling confidential information, not for raising safety concerns. It maintains that internal debate is encouraged and essential, and that these actions were an exception to its normal culture of openness.
Critics and sceptics: Some observers question the timing and opacity of the terminations, noting that OpenAI has not provided specific details of the alleged breach. Others point out that without independent verification, it is difficult to assess whether the firings were justified or punitive.
What to Watch
- Whether the former employees pursue legal claims or regulatory complaints, such as with the US National Labor Relations Board or California whistleblower protections.
- Any further departures from OpenAI's safety team or public statements from current employees.
- How OpenAI's board and external safety committees respond to the allegations, including any review of termination procedures.
- Potential congressional or regulatory attention on whistleblower protections in the AI industry.