Three former OpenAI researchers, Leopold Aschenbrenner, Tomek Korbak, and Jasmine Wang, published an open letter following their sudden dismissal last week. The authors argue that their termination may create a chilling effect on the company's culture, potentially discouraging current employees from speaking out on safety issues.
The researchers, who specialized in AI alignment and safety, claim their actions were within the scope of their professional duties. Aschenbrenner stated he was dismissed for excessive communication with third-party safety organizations, while Korbak cited his interactions with the non-profit research group METR as the reason for his firing. Wang denied allegations regarding unauthorized access to executive emails.
In their letter, the trio proposed three recommendations for OpenAI: allowing third-party safety audits, ensuring effective monitoring of frontier AI models, and supporting transparent communication between safety researchers and external institutions. They expressed concern that the company might use their departure as a pretext to limit future collaboration with organizations like METR.
An OpenAI spokesperson stated that an internal investigation concluded the three employees violated policies regarding the access and handling of sensitive information. Meanwhile, a leader within the company's research division issued an internal memo supporting the researchers' proposals, emphasizing that the dismissals were unrelated to the employees raising safety concerns or speaking publicly.
