Three former OpenAI safety researchers have criticized the company's firing process, claiming it creates a culture of fear and undermines internal safety efforts.

OpenAI recently terminated three safety researchers, sparking an open letter from the group that claims the dismissals are causing fear among remaining staff. The employees—Tomek Korbak, Jasmine Wang, and Mikita Balesni—were fired following an investigation into a security incident involving Hugging Face. OpenAI maintains that the terminations resulted from a breach of company policy regarding sensitive information, rather than the researchers' internal safety advocacy.

Circumstances of the Terminations

Tomek Korbak stated he was fired after a meeting with the head of the safety department, during which he was told the company no longer trusted him. Korbak, who served as a technical contact for the external safety lab METR, claims he was not provided with written documentation regarding his alleged misconduct. Jasmine Wang reported that her termination followed an incident where she accidentally accessed an executive's email inbox, an issue she says she reported immediately after it occurred.

Researcher Demands and Company Response

In their open letter, the three researchers called on OpenAI to honor commitments to embed external safety auditors like METR, maintain the monitorability of frontier AI models, and clarify rules for working with outside safety groups. OpenAI responded by stating that an internal investigation found the employees violated policies on handling sensitive information. The company denied that the firings were related to the employees raising safety concerns and confirmed it is working on contracts with external safety auditors.