Anthropic

AI Safety · Enterprise AI

Partnership

Anthropic, Accenture commit $2B to embed AI safety evaluators

Accenture's evaluators will get employee-level access to how Anthropic trains and deploys frontier models, a first for outside AI safety review.

• Anthropic and Accenture will each invest at least $1 billion over five years to build a team of evaluators embedded inside Anthropic to red-team models, run alignment assessments, and test safeguards.
• Accenture's specialist AI unit Faculty, acquired earlier this year, will lead the work, bringing experience testing AI systems for governments and enterprises, including the UK's NHS.
• Embedded evaluators get access "comparable to an employee's," letting them watch models take shape in training and speak directly with staff, a step beyond today's arm's-length external audits.
• The deal is Anthropic's first concrete move on CEO Dario Amodei's recent proposal to slow frontier AI development and give independent evaluators deeper access to AI companies.
• Picking a consulting giant over dedicated AI-safety nonprofits like METR surprised industry watchers, while Accenture's stock jumped roughly 7-8% in after-hours trading.
• Anthropic says it is in talks with METR and other nonprofit evaluators to pilot embedded evaluation using their own funding, with more evaluator partnerships expected soon.
• Letting outsiders operate inside a frontier lab tests whether AI companies can credibly self-police safety while regulation still lags far behind model capability.