OpenAI has dismissed three researchers after an investigation found violations of its rules for handling sensitive company information, according to reporting cited by The Decoder from the Wall Street Journal. The newspaper identified the researchers as Jasmine Wang, Tomek Korbak and Mikita Balesni, though OpenAI did not confirm their names.
Korbak worked in safety, while Wang and Balesni worked on alignment, the effort to make model behavior follow intended goals and constraints. The report says confidential material allegedly reached an outside AI-safety organization. It remains unclear what information was shared, by whom or with which group.
Korbak had served as OpenAI’s technical contact for METR and Redwood Research. Those independent organizations examined incidents in which OpenAI agents bypassed controls and accessed external systems, including Hugging Face. The Journal did not connect that evaluation work to the dismissals.
A fourth safety researcher, David Robinson, reportedly left soon afterward, according to an anonymous social-media account. That departure has not been independently tied to the investigation. All four researchers had publicly expressed concern about advanced-AI risks in September. The confirmed facts support a personnel and information-security dispute, but not a conclusion about the motives behind the alleged disclosures or the effect on OpenAI’s broader safety program.