Anthropic and Accenture will invest $2 billion in AI model evaluations over five years, aiming for independent assessments to enhance safety and alignment.
Anthropic has partnered with consulting firm Accenture to invest a minimum of $2 billion over the next five years for 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.
— Anthropic (@AnthropicAI) September 18, 2026
This initiative will be spearheaded by Faculty, Accenture's specialized AI division, focusing on evaluating models, alignment checks, fine-tuning, and safety testing.
The project supports the "embedded evaluators" concept introduced by Anthropic's CEO Dario Amodei earlier this year. This model allows evaluation teams to work closely within AI development teams, giving them access to internal processes to monitor training, deployment, and safety commitments.
Amodei has previously expressed concerns about the rapid development of advanced AI, citing security risks highlighted by incidents involving companies like OpenAI. He proposed several risk mitigation strategies, including international agreements and independent audits.
The collaboration with Accenture is one of many planned by Anthropic, which will also engage with other evaluators in the future. Accenture is expected to work with various third-party AI developers as well. Additionally, discussions are underway with the nonprofit METR for a pilot program on embedded evaluation.
Looking ahead, Anthropic plans to seek funding for independent evaluations from pooled resources or government sources, as detailed in their Advanced AI Framework initiative. Currently, they are financing Accenture's work directly, which raises concerns about the independence of the evaluation process.
In contrast, Nvidia's CEO Jensen Huang has criticized calls for regulatory slowdowns in AI development, suggesting that safety issues should be viewed as engineering challenges.
