Anthropic picks Accenture for third-party AI safety evaluations

This move represents the first step in the CEO’s plan to slow down AI development.

Anthropic's partnership with Accenture is one of the first steps of a plan to slow down the pace of AI development, as proposed by its CEO , Dario Amodei, last week. In Amodei's three-step approach, the proposal called for all "frontier AI" companies to give "ongoing, employee-like access" to embedded third-party evaluators that are tasked with ensuring safety practices and commitments are being followed, reporting incidents and assessing AI model alignment. More specifically, 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," according to Anthropic.

Anthropic explained that Accenture had experience helping businesses and governments deploy AI across several industries. The AI company also said that Accenture has experience with how enterprises use AI and can offer more perspective when it comes to Anthropic's own models. However, Anthropic noted that the idea of embedded evaluators is new and hasn't yet set standards when it comes to scope, what access the evaluators get nor the procedure of reporting concerns.

To get a better sense, Anthropic said it will be working with others companies beyond Accenture when it comes to evaluating its frontier AI models. The company said it will announce other evaluators it in the coming weeks, while also expressing that Accenture will work with AI companies besides Anthropic.