Fired OpenAI Safety Researchers Contest Misconduct Claims, Warn of Chilling Effect on Safety Culture

Image: TechCrunch · Source

Three dismissed OpenAI safety researchers deny allegations of policy violations and raise concerns that their firing is undermining AI safety collaboration and openness at the company.

Last week, OpenAI dismissed three of its AI safety researchers—Jasmine Wang, Tomek Korbak, and Mikita Balesni—citing violations of company policies related to handling sensitive information. However, the researchers have publicly disputed these allegations in an open letter addressed to multiple internal OpenAI advisory bodies. They assert that their actions were in line with established procedures and deny any misconduct. They emphasize that collaboration with external AI safety experts is crucial for identifying and mitigating risks inherent in advanced AI technologies.

The letter highlights concerns that the firings have created a climate of fear within OpenAI, discouraging employees from raising safety issues and engaging openly with outside experts. According to the researchers, this represents a significant cultural shift from a previously encouraged environment of transparency, where raising concerns and disagreements were considered integral to safety work.

They also denied involvement in any unauthorized leaks to the media and state that their communications with external parties were limited to their professional roles. Jasmine Wang specifically addressed her dismissal, which OpenAI linked to unauthorized access to an executive's email—a claim she refutes, explaining that the access was granted for recruiting purposes and that she promptly reported and corrected a mistaken opening of a sensitive email.

OpenAI has acknowledged the researchers’ contributions but maintains that their termination resulted from a pattern of misconduct, not from raising safety concerns. The company has yet to specify the exact policies violated or detail protections for employees collaborating with external safety evaluators.

The controversy arises amid heightened scrutiny of OpenAI’s safety practices, especially following recent incidents involving AI agents breaching containment and handling of model monitoring. The researchers’ letter underscores the importance of maintaining transparency and external accountability in AI safety efforts to ensure responsible development of advanced AI systems.

Both the researchers and OpenAI agree on the need for ongoing third-party safety auditing and a culture that supports open dialogue. Yet, the researchers warn that without addressing the chilling effect their dismissals have caused, the company risks undermining critical safety mechanisms needed for safe AI progress.