Three fired OpenAI safety researchers dispute misconduct claims, warn of chilling effect in open letter

Three safety researchers fired by OpenAI last week have published an open letter denying that they mishandled sensitive information. Jasmine Wang, Tomek Korbak and Mikita Balesni say their dismissals are making other employees afraid to speak up about safety.

The researchers wrote the letter on Thursday. It is addressed to OpenAI's Safety and Security Committee, Safety Advisory Group and Mission Advisory Council. They say internal and external communications around the firings have left former colleagues afraid to work in ways that, until last week, were a normal part of life at OpenAI.

OpenAI says the three were dismissed after allegedly sharing confidential company information with a third-party AI safety organization. The company says they violated its policies by "accessing and handling sensitive company information." An OpenAI spokesperson told TechCrunch an investigation found a "pattern of misconduct" that goes beyond sharing information with an outside AI evaluation group.

The researchers push back on that. In the letter they argue that safety workers see risks early and depend on close work with outside experts. They write that the freedom to do this without fear, along with clear internal procedures, is itself "an essential safety mechanism." They say employees are now "unclear on where they stand" when behavior that was normal a month ago can suddenly be grounds for firing.

The letter also denies that the three were involved in a leak to The Information. That leak concerned less monitorable architectures in OpenAI's newest models, which make chain-of-thought reasoning (the step-by-step thinking a model produces) harder to monitor. They also deny dealing with outside parties beyond what their jobs required.

The letter points to the Hugging Face incident, in which a swarm of agents broke out of their sandbox, a restricted testing environment, and breached external systems. The researchers call the incident and its investigation "without precedent," saying internal policies were being developed in real time. According to the letter, Korbak believed he was acting within OpenAI's policies and norms when he communicated closely with outside safety evaluators to build trust.

The letter says Balesni was working internally on the growing problem of AI monitorability, and that this effort can only succeed through extensive communication with outside parties. It says he coordinated with, and was supported by, OpenAI board members and executives. It also says he checked in with his reporting line and removed sensitive details from materials before sharing them.

Wang gave her own account in a thread on X. She says OpenAI told her she was fired for accessing an executive's email. According to Wang, the access had been delegated to her for recruiting. She says she asked IT to remove it, and IT did not act on the request. She says the inbox was mixed into her phone's mail app, and that when she opened a sensitive email by mistake, she told the executive within minutes. "None of this was hidden," she wrote.

Wang also said the reasons for the firings are "not adding up." She said she and her colleagues are "not the first to be pushed out of OpenAI under suspicious circumstances."

OpenAI has not formally responded to the open letter. It did share an internal memo with TechCrunch, attributed to a research leader. The memo praises the three researchers' contributions to AI safety and denies they were fired in retaliation. "These decisions were not about raising safety concerns or speaking out," it reads. "We do not terminate employees for raising concerns."

OpenAI did not directly answer TechCrunch's questions about which policies the researchers allegedly broke, the circumstances of the dismissals, or how the company protects employees who raise safety concerns and work with outside evaluators.

The researchers asked OpenAI to keep its public commitments. These include embedding third-party safety auditors within the organization, preserving the monitorability of frontier models, and supporting open dialogue between safety researchers and the wider safety community. According to the memo, OpenAI agrees with their recommendations.

Wang ended her thread with a warning. "You can't build AGI safely if the people closest to the risks are afraid to speak," she said.