Anthropic said September 18 that it and Accenture each expect to invest at least $1 billion over the next five years in capacity for independent evaluation of frontier AI.
The work is intended to build a team of embedded evaluators at Anthropic. That team will evaluate and red-team models, conduct alignment assessments and test model safeguards, according to the companies’ announcements.
Faculty, Accenture’s specialist AI business, will lead the partnership. Its experience includes work in government, defense, healthcare and infrastructure, including the UK National Health Service’s Early Warning System during COVID-19. Accenture did not state how many evaluators the team will include.
The embedded evaluators will receive workplace-equivalent access at Anthropic. That gives them the ability to observe models during training, follow decisions about how they are built and deployed, and speak directly with employees, Anthropic said in its blog post.
Anthropic is funding Accenture’s work in the interim. It also said it is discussing pilots with METR and other nonprofit evaluators that would use their own funding.
The arrangement is non-exclusive. Accenture can work with other AI developers, while Anthropic said it plans to work with additional evaluators in the coming weeks.
Accenture Chair and CEO Julie Sweet said AI safety needs technical expertise alongside an understanding of how AI is used in practice. Dr. Marc Warner, Accenture’s CTO and Faculty CEO, said the partnership fits Faculty’s founding mission.
The announcement follows Anthropic’s commitment to give outside evaluators employee-level access, as Go Big News reported on September 14. Anthropic also said the partnership advances a proposal in CEO Dario Amodei’s essay, “We Must Pace the Frontier.”
Anthropic announced the partnership in a September 18 blog post and a post on X.
