OpenAI fires three safety researchers for "mishandling research information"

Three safety researchers at OpenAI were let go last week, according to an open letter they published denying their dismissal. Jasmine Wang, Tomek Korbak, and Mikita Balesni accused the company of firing them for violating policies and mishandling sensitive data. The researchers claim their termination marks a new culture within OpenAI that discourages workers from speaking up about safety concerns.

They argue that fear and unclear guidelines are stifling AI safety work and weakening accountability through third-party audits. The letter also denies the researchers' involvement in a leak to The Information about specific model architectures. OpenAI has not responded to the open letter, but did share an internal memo supporting the researchers and denying they were fired for raising safety concerns.