AI experts demand independence for safety evaluators

More than 100 AI researchers and evaluators published a public letter Friday warning that third-party safety evaluators lack the independence, resources, and legal protections needed to credibly assess the risks posed by frontier AI models. The letter, organized by the AI Evaluator Forum, calls on frontier AI companies to guarantee evaluators scientific objectivity, transparency, and protection against interference. Signatories include AI researcher Geoffrey Hinton and representatives from Johns Hopkins University, Stanford University, and nonprofit evaluator METR. The letter sets minimum conditions, including that evaluators must not be owned by the companies they assess, must not receive payment contingent on findings, and must be shielded from retaliatory litigation. It also demands evaluators receive access equivalent to senior internal employees, including review of unreleased systems.