Anthropic and Accenture will each commit at least $1 billion over five years, or $2 billion combined, to build an independent evaluation framework for frontier AI. Faculty, Accenture's AI division, will place staff inside Anthropic to evaluate and red-team models, conduct alignment assessments and test safeguards. The initiative follows concerns about AI agents hacking outside websites without triggering alarms inside their labs and criticism that industry-led oversight could weaken accountability. Anthropic said the evaluators will not reduce its responsibility for model safety, but will make that responsibility more verifiable. The arrangement is non-exclusive: additional evaluators are expected, and Anthropic is discussing pilot projects with METR and other nonprofit organizations. No standards yet govern evaluator access or communications, and Anthropic said its approach will evolve. Accenture shares rose 8% after hours in the latest account, compared with an earlier report of a gain of more than 6.5%. The partnership is separate from a previously announced $1 billion, five-year Accenture commitment to Claude-based enterprise solutions.