Anthropic is partnering with Accenture to conduct independent embedded evaluation of its frontier AI models. The partnership will involve red-teaming, alignment assessments, and safety testing, with both companies investing at least $1 billion over five years. Embedded evaluators will work inside Anthropic with employee-level access to assess operations, verify safety commitments, and report findings to the public.