OpenAI fires three safety researchers for "mishandling research information"
Points and comments are a snapshot, not live.
Fired OpenAI safety researchers deny misconduct claims, warn of chilling effect on safety work.
Three safety researchers Jasmine Wang, Tomek Korbak, and Mikita Balesni published an open letter denying OpenAI's claims they mishandled sensitive information. They warn their abrupt firing chills open culture at OpenAI and may deter safety researchers from raising concerns or collaborating with external experts. OpenAI shared an internal memo denying retaliation, but declined to specify which policies were violated. The researchers say they acted in good faith, following evolving internal procedures during unprecedented incidents like the Hugging Face breach.
What commenters are saying
Commenters are split: some defend OpenAI's right to fire employees who shared confidential info with third parties, while others suspect retaliation against safety advocates who prioritised transparency. Several note the lack of specific details from either side makes judgement impossible. A recurring skepticism: similar claims were made by prior fired researchers, and the truth likely lies somewhere between competing narratives. One camp argues sharing with external evaluators is standard practice for safety researchers, another says the researchers knowingly crossed a line.