Three fired OpenAI safety researchers warn that fear and unclear rules could obstruct safety work. Tomek Korbak, Jasmine Wang and Mikita Balesni published a joint letter to the company’s safety and oversight bodies on October 8.
They call for continued access for independent evaluators, preservation of monitorable model architectures and clear procedures for working with outside safety organizations. Former colleagues must remain able to raise risks openly, they argue.
The authors deny misconduct. OpenAI denies retaliation and cites policy violations. The grounds for their dismissals remain disputed.
The three worry that collaboration with METR could be restricted. Their letter does not establish that this collaboration has ended. Their requests concern oversight of future systems and the handling of safety concerns.