Hinton joins open letter calling for independent AI risk assessments

On September 18, 2026, a coalition of oversight organizations released an open letter advocating for standardized conditions for third-party AI risk assessments. The letter, supported by over 50 AI experts including Geoffrey Hinton and Stuart Russell, urges companies like Anthropic and OpenAI to ensure independent evaluators have access to internal resources and can operate without interference. The letter emphasizes the need for transparency, independence, and protection for assessors, who should be allowed to communicate directly with company boards and publish findings without censorship. It also calls for diverse expertise to cover risks such as biosecurity and AI control. Microsoft CEO Satya Nadella endorsed the idea of third-party access to AI firms, joining a growing conversation about industry accountability. The letter highlights the importance of safeguarding assessors from retaliation and granting them equal access to tools and data as internal staff. It argues that AI labs must introduce external evaluators with varied backgrounds to comprehensively address potential risks. Anthropic and OpenAI CEOs previously pledged to allow third-party access to internal information, but the letter seeks to formalize these commitments into enforceable standards. The open letter reflects rising concerns about AI safety and the role of independent oversight in mitigating risks.