Anthropic is partnering with Accenture on independent evaluation of frontier AI. The partnership is a step toward the commitment, made in Anthropic's CEO essay "We Must Pace the Frontier," to place evaluators inside Anthropic.
Faculty, Accenture's specialist AI business, will lead the partnership. The work will include evaluating and red-teaming models, conducting alignment assessments, and testing model safeguards. Accenture helps businesses and governments deploy AI across many industries. Its understanding of how enterprises use AI in practice shapes its approach to safety, and it will bring that perspective to its evaluation of Anthropic's models.
Anthropic and Accenture each expect to invest at least $1 billion over the next five years to build capacity in this area.
Embedded evaluation is a new approach, and many details of how it will work are still being decided. Today's external evaluators work from outside AI companies. Embedded evaluators will work inside them, with access comparable to an employee's. This lets them watch models take shape during training, follow the decisions that govern how models are built and deployed, and speak directly to employees. From this position, they can assess how a company operates, verify that it is keeping its safety commitments, and identify blind spots. They can also report incidents and give the public a more informed account of benefits and risks.
Independent embedded evaluators do not reduce Anthropic's accountability. They make it more verifiable. The safety of Anthropic's models remains Anthropic's responsibility.
There are not yet any standards for what information embedded evaluators should be able to access or how they should report their findings. There is also no settled system for funding independent evaluation. In the long term, Anthropic thinks funding should come from pooled or government sources, as Anthropic's Advanced AI Framework called for in June. Neither exists today, so Anthropic plans to work with different evaluators under different funding arrangements.
Given how important and urgent this work is, Anthropic will fund Accenture's work directly. Anthropic is also in dialogue with METR and other nonprofit evaluators about piloting elements of embedded evaluation using their own funding. Anthropic believes that frontier AI ultimately needs an ecosystem of evaluators operating with shared standards.
Frontier labs are expected to work with several organizations at once. The partnership is non-exclusive. Anthropic will work with other evaluators, to be announced in the coming weeks, and Accenture will work with other AI developers in similar roles.
Anthropic will keep training and releasing frontier models, and it wants independent evaluators working alongside it as it does. Anthropic is sharing these early efforts now so that the public and other AI developers can see its process. It expects its approach to evolve as the field matures, and it will share more as the work begins and as more evaluators join.