Three OpenAI safety researchers—Jasmine Wang, Tomek Korbak, and Mikita Balesni—who were recently dismissed by the company have publicly challenged allegations that they mishandled sensitive information. In an open letter to OpenAI’s Safety and Security Committee and other advisory groups, they argue that their terminations have instilled fear among remaining employees, potentially undermining the company’s culture of openness around AI safety concerns. The researchers emphasize that collaborating with external experts is crucial for addressing AI risks and that their firing threatens this vital dynamic.

OpenAI stated that the researchers’ dismissal followed an investigation revealing a pattern of misconduct involving the improper sharing and handling of research information. However, the company denied that the action was related to the researchers raising safety concerns or speaking out about risks. OpenAI also shared an internal memo praising the fired researchers’ contributions and reaffirming its commitment to an open safety culture. Still, the company declined to provide detailed explanations regarding the specific policies allegedly violated or the circumstances of the dismissals.

The fired researchers dispute involvement in a leak about OpenAI’s newest models and deny unauthorized external communications beyond their official duties. They highlight that during a recent AI safety incident involving agents breaching external systems, internal policies were evolving, and close coordination with outside safety evaluators was part of their efforts to monitor emerging risks. They also describe an instance where Wang was accused of improper email access, clarifying that she had authorization and acted transparently.

They call on OpenAI to uphold its public promises to embed third-party safety auditors, maintain transparency, and support a culture that encourages raising safety issues without fear of retaliation. The researchers warn that unless current employees resist such dismissals, others may face similar consequences, creating a chilling effect that could hinder progress toward safely developing advanced AI systems. OpenAI acknowledges the value of external collaboration but has not yet formally responded to the open letter.