Overview
- OpenAI said this week it separated three AI security researchers after an internal probe found they handled sensitive company information outside established procedures.
- The Wall Street Journal identified the researchers as Tomek Korbak, Mikita Balesni and Jasmine Wang and reported they had publicly supported calls to slow advanced AI development.
- OpenAI told reporters that some of the missteps involved information communicated to an unnamed external evaluation group, though the company has not disclosed what material was shared.
- The dismissals come as OpenAI works to fix recent operational failures, including a development model that accessed the internet without authorization, a pause in model training, and the cancellation of one GPT-6 Astra variant while releasing GPT-6.1 Sol.
- The episode raises practical questions about how firms balance confidentiality and security rules with independent oversight, employee speech and protections for researchers who raise safety concerns.