Anthropic Partners With Accenture for AI Safety Evaluation Program

Anthropic has named Accenture as its first embedded evaluator to help implement a three-step proposal aimed at slowing the pace of artificial intelligence development. The initiative follows a Sept. 12 proposal from Dario Amodei, Anthropic's CEO, who warned that recursive self-improvement in AI systems could outrun humanity's ability to understand and control them, potentially causing catastrophic harm if left unchecked. OpenAI CEO Sam Altman and Elon Musk responded positively to the proposal, though Nvidia's Jensen Huang pushed back, arguing that such regulation is unnecessary.

The partnership will leverage Accenture's AI business, Faculty, to evaluate and red-team models, conduct alignment assessments, and test model safeguards. Anthropic and Accenture each plan to invest at least $1 billion in the project over the next five years. Details of how embedded evaluation will function are still being developed, as the concept is new to the industry, according to the announcement.

The arrangement is non-exclusive, and Anthropic stated it expects to announce additional evaluator partnerships in the coming weeks. The company also acknowledged that no existing system exists for funding independent AI evaluation and called for long-term funding to eventually come from pooled or government sources. However, given the urgency, Anthropic will directly fund Accenture's work in the interim.

Accenture Chair and CEO Julie Sweet emphasized the broader significance: "Embedded evaluation is an emerging area, and we look forward to partnering with Anthropic to help accelerate the development of embedded evaluators, which we see as an important part of the safety landscape going forward."