Anthropic and Accenture will each invest at least $1 billion over five years in independent evaluation of frontier AI models, according to an announcement by the companies. Faculty, Accenture's AI unit, will embed evaluators inside Anthropic with access comparable to an employee's.
The evaluators will conduct red-teaming, run alignment assessments and test safeguards on Anthropic's models. Anthropic is paying Accenture directly for the work. The partnership is non-exclusive; Anthropic says it will name more evaluators in coming weeks, and Accenture will work with other AI developers.
Frontier AI evaluation remains nascent and contested. No regulatory framework requires it, though the U.S. National Institute of Standards and Technology has published voluntary AI risk management guidance. Accenture brings scale to the work; the consulting firm has 738,000 employees globally and operates evaluation practices across financial services and other regulated sectors.
The five-year horizon reflects both parties' expectation of extended engagement. Anthropic has grown rapidly since securing $5 billion from Google in January 2025 and raising $2 billion from Amazon and others in a Series D round that valued the company at $60 billion. The company operates Claude, a large language model competing with OpenAI's GPT-4 and Google's Gemini.
The announcement treats embedded evaluation as distinct from customer penetration testing or bug bounties, which remain separate. Faculty evaluators will report findings to Anthropic's leadership and safety teams; the deal does not establish public reporting requirements or independent verification of assessment outcomes.
A $1 billion five-year spend by each party means Accenture is committing roughly 0.13 percent of its projected annual revenue to this single engagement, assuming flat revenue of approximately $70 billion annually. Anthropic's equivalent commitment, as a much smaller company, represents a more substantial operational proportion. The document to watch is any published evaluation methodology Accenture shares with the broader industry; if Faculty does not release a framework for others to cite by the end of 2027, the partnership may remain opaque to competitors and regulators assessing whether independent evaluation is becoming a real constraint or a cosmetic exercise.
