Anthropic Partners with Accenture for Independent Embedded AI Evaluation

Anthropic is partnering with Accenture to conduct independent embedded evaluation of its frontier AI models, a commitment outlined in its CEO’s essay “We Must Pace the Frontier.” Accenture’s specialist AI business, Faculty, will lead this initiative to enhance model safety and alignment.

The partnership will focus on evaluating and red-teaming models, conducting alignment assessments, and testing model safeguards. Both Anthropic and Accenture expect to invest at least $1 billion each over the next five years to build capacity and advance this critical area of AI safety.

Embedded evaluators will operate inside AI companies with access comparable to employees, observing models from training through deployment and speaking directly with staff. This access allows them to assess company operations, verify safety commitments, identify blind spots, and report incidents, making accountability more verifiable.

Anthropic will directly fund Accenture's work, acknowledging the absence of established standards for information access or a settled funding system for independent evaluation. The company is also in dialogue with non-profit evaluators like METR to pilot elements of embedded evaluation, aiming for an ecosystem with shared standards and long-term pooled or government funding.