They argue that their dismissals and unclear rules for external collaboration risk chilling internal dissent and cooperation with independent safety organizations. The authors recommend preserving independent evaluator access, maintaining monitorability of frontier models, and clarifying procedures for communicating with outside experts. The letter also sets out the authors’ accounts of their dismissals and related episodes; these are claims by the authors, not independently established findings.
Raw SignalOct 10, 20261 min read1 source
OpenAI Cannot Make AI Safe on Its Own (6 minute read)
In one minute
Three former OpenAI safety researchers—Tomek Korbak, Jasmine Wang, and Mikita Balesni—published a letter addressed to the company’s Safety and Security Committee, Safety Advisory Group, and Mission Advisory Council.