Accenture Gets Inside Access to Test Anthropic's AI Models

Anthropic named Accenture its first embedded evaluator on Sept. 18. Staff from Accenture's Faculty AI division will work inside Anthropic with employee-level access to red-team models — deliberately stress-test them for flaws — run alignment checks to see if behavior matches intent, and test safeguards. It moves safety testing from outside checks via the API to ongoing internal review.