Fired OpenAI safety researchers Jasmine Wang, Tomek Korbak, and Mikita Balesni published an open letter denying claims that they mishandled sensitive information. OpenAI dismissed the three researchers last week, alleging they violated company policies by sharing confidential company information with a third-party AI safety organization and engaging in a pattern of misconduct.
In their open letter to OpenAI committees and councils, the researchers warned that their abrupt dismissal creates a chilling effect on company culture and leaves employees afraid to speak or operate openly. They emphasized that working on AI safety requires close collaboration with outside experts to address risks and that unclear rules and fear will stymie critical safety work. The researchers also specifically denied involvement in a leak regarding chain-of-thought reasoning architectures and stated they believed they were acting within company norms during the Hugging Face incident investigation.
OpenAI has not formally responded to the letter, but previously shared an internal memo denying the researchers were fired in retaliation for raising safety concerns. An OpenAI spokesperson maintained that the dismissals followed an investigation revealing clear violations of company policies regarding the mishandling of research information.
This dispute matters because it highlights growing tensions between internal AI safety research, external accountability, and corporate confidentiality at leading artificial intelligence firms.