More than one hundred artificial intelligence experts and evaluators, coordinated by the AI Evaluator Forum, issued a joint call for frontier AI labs to grant third-party evaluators greater authority and protections to independently oversee development of the most advanced models, according to an open letter whose signatories include Geoffrey Hinton and members from Johns Hopkins University, Stanford University, and METR. The letter demands that foundation model providers ensure third-party evaluators have "scientific objectivity, transparency, independence, and robust protections," that evaluation work be conducted independently of the companies involved, that technical transparency increase, and that evaluators be "protected from retaliation by the companies they are embedded in." AI Evaluator Forum President Conrad Stosz said the goal is to demonstrate common ground on fundamental principles and establish "baseline principles" and "greater standardization" for independent evaluators, without advocating a single approach. The renewed focus stems from a proposal by Anthropic CEO Dario Amodei to grant certain evaluators employee-like access to inspect and audit advanced foundation models and development processes, which Stosz said appears to involve far more extensive access, including company computers, candid employee conversations, and sensitive internal data and unreleased systems, citing the unreleased OpenAI model used in the recent Hugging Face attack incident. Amodei's proposal drew public support from OpenAI CEO Sam Altman, xAI's Elon Musk, and Microsoft CEO Satya Nadella, though implementation questions remain.