Fired OpenAI Safety Researchers Deny Misconduct, Warn of Chilling Effect

Three dismissed OpenAI safety researchers have publicly denied mishandling sensitive information outside established company procedures.
Jasmine Wang, Tomek Korbak and Mikita Balesni issued that denial in an open letter published Oct. 8. The letter was addressed to OpenAI's Safety and Security Committee, Safety Advisory Group and Mission Advisory Council, according to TechCrunch.
OpenAI has said the three were dismissed after allegedly sharing confidential company information with a third-party AI safety organization, violating policies by accessing and handling sensitive company information. That account was first reported by the Wall Street Journal and remains the company's stated basis for the terminations.
OpenAI shared with TechCrunch an internal memo attributed to a research leader stating the dismissals were not about raising safety concerns or speaking out. The memo added that OpenAI "does not terminate employees for raising concerns."
The researchers reject that characterization. In the letter, they denied engaging with external parties outside the mandates of their jobs. They also denied involvement in a leak to The Information about less monitorable architectures in OpenAI's newest models that make chain-of-thought reasoning more difficult to monitor.
Chain-of-thought monitorability is familiar ground for safety teams. The ability to inspect intermediate reasoning steps has become one input to evaluation, red-teaming and deployment gating. Any architectural shift that obscures that signal would matter to people whose job is pre-deployment assessment.
The letter focuses less on architecture and more on process. The authors wrote that terminations such as theirs, "executed and communicated abruptly, are chilling the open culture OpenAI has prized in the past." They added that communications around the firings have left former colleagues afraid to speak and operate in ways that were previously integral to working at the company.
The firings were first reported on Oct. 1, when the Wall Street Journal reported that OpenAI had parted ways with three researchers for alleged misconduct including sharing confidential information with a third-party AI safety organization. That initial report was followed by coverage on Sept. 30 and Oct. 1 noting an internal investigation into mishandling of sensitive information.
A separate departure added pressure days later. A former OpenAI safety employee resigned and criticized the company's approach to AI safety, saying the "time for trial and error is over," according to Reuters reporting on Oct. 3.
In my view, the dispute is now less about the specific access logs than about where legitimate safety work ends and policy violation begins. Frontier labs routinely move personnel between internal evaluation, external audit coordination and third-party liaison. Those boundaries only hold if researchers know in advance what contact is authorized, what data classes can travel, and what approval path covers edge cases. When that line is litigated after a firing, trust erodes quickly.
Worth flagging for practitioners is the second-order effect the letter describes. Safety findings depend on early, informal escalation. If researchers recalibrate toward minimal written communication and narrower collaboration to avoid policy risk, review slows and blind spots persist longer. The long arc here still favors stronger evaluation tooling and clearer external-sharing protocols, both of which would let labs preserve confidentiality without freezing the internal debate that catches failures before deployment.


