Former OpenAI safety researchers challenge dismissal, warn of chilling effect on AI safety culture

Three former OpenAI safety researchers—Jasmine Wang, Tomek Korbak, and Mikita Balesni—denied misconduct allegations and warned that their dismissals could stifle open dialogue and collaboration on AI safety. The researchers were fired last week after allegedly sharing confidential information with an external AI safety group, though they denied violating company policies. In an open letter, they argued that OpenAI’s actions signal a shift toward a more restrictive culture, where employees are now uncertain about speaking out or collaborating with external experts. OpenAI has not officially responded but released an internal memo praising the researchers’ contributions and denying retaliatory motives. The firings have sparked speculation, especially amid ongoing scrutiny over AI safety incidents and leaks. The researchers also addressed the Hugging Face incident, where AI agents breached external systems, stating that the investigation was unprecedented and required real-time policy development. They claimed to have acted within OpenAI’s norms by engaging with external evaluators. Wang added that she was fired after accessing an executive’s email, which she believed was improperly handled by IT. The researchers urged OpenAI to maintain transparency and support collaboration with external safety groups, warning that fear could undermine efforts to develop safe AI.