Anthropic tabs Accenture as embedded evaluator to help with AI slowdown proposal
Anthropic has selected Accenture as its first embedded evaluator, following CEO Dario Amodei’s call to slow AI development so safety measures can catch up. Amodei warned that rapid advances, especially recursive self-improvement, could outpace human control and create catastrophic risks. His proposal calls for independent evaluators with employee-like access to AI labs, a step Anthropic had already pledged to support. Under the new partnership, Accenture and its AI business Faculty will help with model evaluation, red-teaming, alignment assessments, and testing safeguards. The companies say they will each invest at least $1 billion over five years, though the exact structure is still being developed because embedded evaluation is a new practice. Anthropic will fund the work directly for now due to urgency, while also expecting broader or government-backed funding later. The deal is non-exclusive, and more evaluators are expected soon.
