Three former OpenAI safety researchers—Jasmine Wang, Tomek Korbak, and Mikita Balesni—have published an open letter disputing the company’s allegations of policy violations and warning that their dismissal is stigmatizing the organization’s culture of transparency. The trio, who were let go last week, argue that their removal signals a shift away from the collaborative environment that previously encouraged staff to raise safety concerns and disagree openly.
In the letter addressed to OpenAI’s Safety and Security Committee, Safety Advisory Group, and Mission Advisory Council, the researchers stated they are concerned that communications surrounding their firing have instilled fear among remaining employees. “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 letter read.
OpenAI terminated the researchers after alleging they shared confidential company data with a third-party AI safety organization. A company spokesperson told TechCrunch that an internal memo from a research leader praised the group’s contributions but denied the firings were retaliatory. “I want to be very clear that these decisions were not about raising safety concerns or speaking out,” the memo stated. “We do not terminate employees for raising concerns.” OpenAI has not explicitly answered questions regarding which specific policies were breached or how it protects employees who collaborate with external evaluators.
The researchers explicitly denied leaking information to The Information regarding less monitorable architectures in OpenAI’s latest models. They also contested claims that they engaged with outside parties beyond their job mandates. Regarding the Hugging Face incident, where autonomous agents breached external systems, the letter described the event as unprecedented, noting that internal policies were being developed in real-time. They argued Korbak acted in good faith by communicating with outside safety evaluators to build trust, while Balesni coordinated with OpenAI board members and executives to address monitorability issues.
Wang provided additional context on her own dismissal via a post on X, stating that she was fired for accessing an executive’s email. She explained that IT had delegated access to her for recruiting purposes but failed to revoke it when she requested removal. Due to how her phone’s mail app displayed combined inboxes, she accidentally opened a sensitive email, immediately notified the executive, and recontacted IT. She characterized the reasons for the terminations as not adding up and suggested she may not be the first employee pushed out under suspicious circumstances.
The former researchers called on OpenAI to honor its public pledge to embed independent third-party safety auditors, preserve the monitorability of frontier models, and maintain an open culture of dialogue between safety teams and the broader ecosystem. While the company’s memo agreed with these recommendations, Wang warned of the broader implications. “Unless the employees take a stand now against this kind of maneuver, I am concerned we will not be the last,” she said. “You can’t build AGI safely if the people closest to the risks are afraid to speak.”
Chilling effect on transparency is the real worry here. AGI research needs open dialogue, not silence.
How exactly does OpenAI protect employees collaborating with outside evaluators if that’s the official stance?
Wait, she accidentally opened the exec’s email? That sounds like careless IT delegation, not malice.
I’m skeptical. Leaking confidential architecture details seems like a major breach, regardless of intent.
The right to speak up without fear is crucial for AI safety. This sets a scary precedent.