Faculty, the AI unit of Accenture, will embed safety evaluators within Anthropic to red-team models and run alignment assessments. This first embedded evaluator role provides access to systems comparable to that of an employee, with Anthropic paying for the work directly. Both companies commit at least $1 billion each over five years to build evaluation capacity for frontier AI models, though the non-exclusive deal allows Accenture to work with other developers and Anthropic will name additional evaluators in the coming weeks.

The partnership aims to implement a proposal by Anthropic CEO Dario Amodei to slow AI development through third-party oversight. More than 100 experts in the AI Evaluator Forum recently called for strict independence standards, including editorial freedom and protections from retaliation for evaluators. Critics argue that Accenture's business ties, including the existence of an Accenture Anthropic Business Group, undermine the independence of these safety checks.