Anthropic names Accenture as first embedded evaluator

Anthropic said it has selected Accenture as its first embedded evaluator as part of a broader effort to slow AI development and strengthen safety checks. The move follows a proposal from Anthropic CEO Dario Amodei, who outlined a threestep approach aimed at giving independent evaluators employeelike access to AI labs so they can review models, run redteam tests, and assess safeguards. According to the companies, the work will include model evaluation, alignment assessments, and safety testing. The partnership is nonexclusive, and Anthropic said it expects to add more evaluators in the coming weeks. Anthropic and Accenture each expect to invest at least $1 billion over the next five years. Because there is no established funding system for independent evaluation, Anthropic said it will pay for Accenture’s work directly for now.