Fired OpenAI Researchers Publish Open Letter Warning of Chilling Effect on Safety

Three recently terminated OpenAI researchers—Mykyta Balesni, Tomek Korbak, and Jasmine Wang—published an open letter warning that their sudden dismissals create a chilling effect on safety research and external collaboration. While OpenAI claimed the terminations resulted from sensitive information handling policy violations, the researchers maintained their actions fell within their normal safety evaluation duties. This incident highlights growing friction within frontier AI companies between strict internal data security policies and open, transparent safety collaboration. It raises broader concerns in the AI governance community about whether researchers can effectively evaluate model risks alongside independent third parties without fear of retaliation. OpenAI management stated in an internal memo that the dismissals were unrelated to raising safety concerns, citing specific policy breaches such as unauthorized executive email access. In response, the researchers proposed three recommendations for OpenAI: allow external safety auditors inside the company, monitor frontier models continuously, and foster transparent communication with independent safety organizations.