Today’s outside AI evaluators look at a model from the outside. Anthropic and Accenture say they will change that. On September 18, 2026, the two companies announced a partnership that puts a team of evaluators inside Anthropic, working alongside its own staff and safety partners.
Each company said it expects to invest at least $1 billion over five years. The money goes toward AI safety, and Anthropic described it as building capacity for independent evaluation.
What the evaluators will do
The team’s job is to evaluate and red-team models, run alignment assessments and test model safeguards. Anthropic says its embedded evaluators will have access comparable to an employee’s. They will be able to watch models take shape in training, follow the decisions that govern how models are built and deployed, and talk directly with staff. From there they can judge how a company operates, check that it is keeping its safety commitments and spot blind spots. They can also report incidents and give the public a more informed account of benefits and risks.
Anthropic was careful on one point. Independent evaluators do not reduce its accountability, it said; they make that accountability easier to verify. “The safety of our models remains our responsibility.”
The work will be led by Faculty, the specialist AI business that is now part of Accenture. Accenture describes Faculty as a recognized expert in testing and evaluating models for some of the world’s leading AI labs. Its past work spans government, defense, healthcare and infrastructure, including the UK National Health Service’s Early Warning System during the COVID-19 pandemic.
“Embedded evaluation is an emerging area,” said Julie Sweet, Accenture’s chair and CEO, “and we look forward to partnering with Anthropic to help accelerate the development of embedded evaluators.” Marc Warner, Accenture’s chief technology officer and the CEO of Faculty, said the firm was founded on the belief that AI should be “safe by design, not safe by accident.”
Open questions
Anthropic admits the model is new and that many details of how it will operate are still being worked out. There are no standards yet for what information embedded evaluators should see or how they should report what they find. There is no settled system for paying for independent evaluation either.
In the long run Anthropic thinks the money should come from pooled or government sources, as it called for in its Advanced AI Framework in June. Because neither exists today, it will pay Accenture directly. It is also talking with METR and other nonprofit evaluators about piloting parts of embedded evaluation on their own funding.
The deal is not exclusive. Anthropic said it will work with other evaluators, to be announced in the coming weeks, and Accenture will take on similar roles with other AI developers. The commitment to embed evaluators traces back to an essay by Anthropic’s CEO, “We Must Pace the Frontier.”
Anthropic says it will keep training and releasing frontier models, and wants independent evaluators beside it while it does.
Sources
- Accenture and Anthropic Partner to Build Team of Embedded Evaluators at Anthropic
- Partnering with Accenture on embedded evaluation Anthropic
