Anthropic Taps Accenture to Evaluate Its AI Models for Safety

Accenture is adding a new service to its consulting roster: evaluating Anthropic's frontier AI models for safety compliance. The two companies announced a partnership that will place Accenture staff inside Anthropic as third-party evaluators, with each side investing at least $1 billion into the joint effort over the next five years.

The arrangement is an early piece of a plan Anthropic CEO Dario Amodei laid out last week to slow the pace of AI development. Under his three-step proposal, frontier AI companies would give embedded outside evaluators "ongoing, employee-like access" to verify safety practices and commitments, report incidents, and assess how well models align with intended behavior. Anthropic said Accenture's evaluators will watch models take shape during training, follow the decisions governing how those models are built and deployed, and speak directly with employees.

Anthropic pointed to Accenture's experience helping businesses and governments deploy AI across industries, saying that background gives the firm useful perspective on how enterprises use the technology. Still, Anthropic acknowledged that embedded evaluators are a new idea, with no established standards yet for their scope, access, or how they report concerns. The company said it will bring in other evaluators beyond Accenture in the coming weeks, and that Accenture will also work with AI companies besides Anthropic.

Source: Engadget