What the letter is
Three safety and alignment researchers fired by OpenAI last week published an open letter on Thursday addressed to the company’s Safety and Security Committee, Safety Advisory Group and Mission Advisory Council. It is signed by Tomek Korbak, Jasmine Wang and Mikita Balesni, and titled “OpenAI cannot make AI safe on its own”.
OpenAI has said the three violated its policies by accessing and handling sensitive company information, and a spokesperson told TechCrunch that an investigation found a pattern of misconduct going beyond sharing information with an outside evaluation group. The company has not said which policies.
What they deny
The letter denies that the three were the source of a leak to The Information about new, less monitorable architectures. “We do not know who it was,” it says. “We had no reason to leak it; quite the contrary, the article undermined our own work on cross-company limits on the development of unmonitorable architectures.”
It denies that any of them engaged with outside parties beyond their job mandates. On Korbak, it says the Hugging Face incident investigation was “without precedent and internal policies were being developed in real time”. On Balesni, that he coordinated with board members and executives and “took care to remove sensitive details from materials before sharing them”. On Wang, that her access to an executive’s email was delegated to her for recruiting, that she asked for it to be removed and IT did not do so.
The argument about everyone else
The substance of the letter is not the three firings. It is what the firings signalled to the people who still work there.
“If conduct that was considered normal last month now constitutes grounds for sudden dismissal, everyone at OpenAI is left guessing where the line is,” the letter says. It argues that terminations “executed and communicated so abruptly, are chilling the open culture OpenAI has prized in the past”, and that the freedom to work with outside experts without fear “is itself an essential safety mechanism”. It adds: “AI is not a normal technology, and OpenAI is not a normal company.”
Three requests
First, that OpenAI honour last month’s public commitment to embed third-party safety auditors, and that “our firing should not be used as a pretext for stepping away from those partnerships”. The letter says the authors are concerned the dismissals could be used to justify ending the company’s work with METR or narrowing external auditors’ access.
Second, that OpenAI preserve the monitorability of frontier models. The letter says monitorability is degrading, and quotes public statements from inside the company that chain-of-thought monitorability is “fragile and unfortunately trending in a negative direction”. Its position is that as long as safety depends on monitorability, no frontier lab should ship architectures that reduce it.
Third, that the company keep an open channel between its safety researchers and the wider safety community. The letter asks to be circulated internally.
OpenAI’s position
OpenAI has not formally responded to the letter. It shared an internal memo from a research leader which praises the three and states: “I want to be very clear that these decisions were not about raising safety concerns or speaking out.” The memo says the company does not terminate employees for raising concerns, and that it agrees with the recommendations.
What to watch
Whether METR’s access to OpenAI models changes, and whether the company restates in public the commitment it made in September to give independent evaluators ongoing, employee-like access. That is the one request in the letter with a verifiable answer.
