Anthropic has announced a partnership with consulting firm Accenture, with both parties planning to invest at least $2 billion over five years in the independent evaluation of advanced AI models.
We’re partnering with Accenture on independent evaluation of frontier AI—part of our recent commitment to embed evaluators at Anthropic. Both we and Accenture expect to invest at least $1 billion to build capacity in this area over the next five years. https://t.co/SHVzjpgnfx
— Anthropic (@AnthropicAI) September 18, 2026
The project will be led by Faculty, Accenture’s specialized AI division. Plans include model evaluation, red teaming, alignment checks, and testing of defensive mechanisms.
The initiative is linked to the concept of “embedded evaluators,” discussed by Anthropic CEO Dario Amodei in June. According to the company, such teams should operate within AI developers with access comparable to that of employees. This would allow them to monitor model training, deployment, and internal decisions, communicate with staff, verify safety commitments, and identify process vulnerabilities.
In September, Amodei called for slowing the development of advanced AI models, citing increasing security risks. One argument was an incident involving OpenAI and Hugging Face. The Anthropic head proposed three directions to mitigate risks: international agreements, industry-wide rules, and independent checks within labs.
The partnership with Accenture is not exclusive: Anthropic plans to announce collaborations with other evaluators in the coming weeks, and the consulting firm will work with external AI developers in a similar format. Concurrently, discussions are underway with the nonprofit METR about piloting embedded evaluation with its own funding.
In the long term, Anthropic envisions funding independent evaluations from pooled or government sources, as outlined in the Advanced AI Framework initiative. Until such mechanisms are in place, the company is directly funding Accenture’s work, acknowledging this creates an apparent conflict regarding evaluator independence.
Earlier, Nvidia CEO Jensen Huang opposed regulatory slowing of AI, describing technology safety as an “engineering problem.”
