Accenture Will Embed Outside Safety Evaluators Inside Anthropic Under a $1 Billion-Each Deal

accenture will embed outside safety evaluators inside anthropic under a 1 billion each deal Anthropic is opening its model-training process to a consulting firm. Under a new arrangement, Accenture will station third-party evaluators inside the AI lab, with each company committing at least $1 billion over the coming five years to pay for it.

Anthropic is opening its model-training process to a consulting firm. Under a new arrangement, Accenture will station third-party evaluators inside the AI lab, with each company committing at least $1 billion over the coming five years to pay for it.

It is a sizeable sum for a job that has yet to be defined. By Anthropic’s own admission, the embedded-evaluator concept is untested, and there are no agreed standards yet covering its scope, the level of access evaluators receive, or the way they should raise concerns.

The agreement marks the first concrete step under a proposal Anthropic CEO Dario Amodei put forward last week aimed at slowing down the pace of AI development. His three-part plan calls on every “frontier AI” company to grant “ongoing, employee-like access” to embedded external evaluators, whose remit is to verify that safety practices and commitments are being honoured, report incidents, and judge how well models are aligned.

The role Accenture’s evaluators will play at Anthropic

Anthropic has been notably precise about the access involved. According to the company, Accenture’s evaluators will “watch models take shape in training, follow the decisions that govern how those models are built and deployed and speak directly to employees.”

The lab’s argument for choosing Accenture leans on the firm’s track record of helping businesses and governments roll out AI across a range of industries. Anthropic added that Accenture understands how enterprises actually use AI in practice, which it says offers a view of Anthropic’s own models that the lab cannot generate on its own.

Still, Accenture’s core business is consulting, and its client roster now includes the very company whose models it is meant to assess. Anthropic did not speak to that tension directly. It did note that Accenture will also evaluate other AI companies, meaning the relationship is not an exclusive one.

More evaluators are on the way

Anthropic said additional companies beyond Accenture will be brought in to evaluate its frontier models, and it intends to name them in the coming weeks. For now, the only firm figures in the arrangement are financial: at least $1 billion from each side, a five-year term, and a rulebook still waiting to be written.