Jasmine Wang, Tomek Korbak, and Mikita Balesni, three former OpenAI safety researchers, published an open letter challenging the company's claims regarding their recent termination. The researchers were dismissed last week over allegations of mishandling sensitive information and sharing confidential data with a third-party AI safety organization.
In their letter to the company's safety committees, the researchers denied leaking information about model architectures to The Information or engaging with external parties outside their job mandates. They argued that their collaboration with outside experts was a standard practice necessary for addressing AI safety risks, rather than a policy violation.
Jasmine Wang specifically addressed her dismissal, stating she was accused of accessing an executive's email. She explained that the access was originally granted for recruiting purposes and that she had requested its removal from IT. She noted that she reported the accidental opening of a sensitive email to the executive involved within minutes.
The researchers warned that their firing creates a chilling effect on OpenAI's internal culture, discouraging employees from openly discussing safety concerns. They called on the company to maintain its commitments to transparency, third-party auditing, and open dialogue with the broader safety ecosystem.
OpenAI has not provided specific details regarding the policies allegedly violated. However, an internal memo shared by the company stated that the terminations were not retaliatory and that the firm continues to encourage employees to raise safety concerns.
