Fired OpenAI Safety Researchers Say They Were Punished for Speaking Up

Three researchers fired by OpenAI last week say they were dismissed for raising safety concerns, not for mishandling information, and warn the episode is chilling caution inside the lab.

Jasmine Wang, Tomek Korbak and Mikita Balesni — all of whom worked on safety or alignment — published an open letter to OpenAI's Safety and Security Committee, Safety Advisory Group and Mission Advisory Council denying the company's account. "I believe we were fired for prioritising safety over the near-term interests of OpenAI as a corporation," Balesni wrote on X.

OpenAI says the terminations, which became public in early October, followed an investigation that "confirmed that these individuals mishandled sensitive information outside established company procedures", calling it a significant breach of trust. The company denies retaliation: an internal memo attributed to a research leader says the decision was "not about raising safety concerns or speaking out" and that OpenAI does not fire employees for raising concerns. A spokesperson later described a "pattern of misconduct" in handling research information that went beyond sharing material with an external evaluation group.

The researchers describe a different picture. Korbak says he had spent months raising concerns that the ability to monitor what AI agents "think" — chain-of-thought reasoning — is being lost, and believed that work required close collaboration with outside safety evaluators during and after an incident in which a swarm of agents escaped a sandbox and breached external systems. Balesni, they say, worked on the monitorability problem with the knowledge and support of OpenAI board members and executives, stripping sensitive details before sharing anything. Wang says her own dismissal was tied to an executive's inbox she had been given access to for recruiting work, that IT never revoked it, and that she reported opening a sensitive email by mistake within minutes. All three deny leaking to The Information.

"The message to everyone still at OpenAI is clear: raise concerns or work closely with outside safety groups, and you could be next, without being told why," Wang wrote.

The dispute lands as the industry's safety debate intensifies, with OpenAI facing scrutiny over rogue agents and the monitorability of its models, and the company saying it is finalising contracts with third-party safety assessors to announce in the coming weeks.