Anthropic's first embedded evaluator is the consulting firm Accenture, but Anthropic is paying for the work
Faculty's team will sit inside Anthropic with access comparable to an employee's, but Anthropic says there are no standards yet for what an embedded evaluator should be shown.
Anthropic named the first outside team that will work inside it, on September 18, 2026. The job went to Accenture, and the work will be led by Faculty, Accenture's specialist AI business.
Each company expects to invest at least $1 billion in this over the next five years. Anthropic is funding Accenture's work directly, and the arrangement isn't exclusive on either side.
- Announced
- September 18, 2026
- Evaluator
- Faculty, Accenture's specialist AI business
- Money
- At least $1 billion each over five years
- Access
- Comparable to an employee's
- Next
- More evaluators in the coming weeks
What the evaluators get to see
An embedded evaluator sits inside the company rather than testing a finished model from outside. Anthropic's post says the access lets the team "watch models take shape in training" and speak directly to employees.
That access is meant to cover more than the models. Anthropic says an embedded evaluator can assess how the company operates and verify that it's keeping its safety commitments, then report incidents and give the public an account of what it saw.
What the team can do with what it finds surprised me, because Anthropic says that part isn't settled. There are no standards yet for what information embedded evaluators should have access to, or for how they report what they find.
The safety of our models remains our responsibility
Who pays for it
Anthropic does. The post says funding should come from "pooled or government sources" in the long run, the position Anthropic took in its Advanced AI Framework in June, and that neither of those exists today.
There is also no settled system for funding independent evaluation.
Anthropic says it's in dialogue with METR and other nonprofit evaluators about piloting parts of embedded evaluation on their own funding. Anthropic didn't give a date for that.
Who Faculty is
Accenture bought Faculty, and Accenture's release points at Faculty's work on the UK National Health Service's Early Warning System during the COVID-19 pandemic. Faculty already tests and evaluates models for AI labs as a business, and Accenture sells AI deployment to the enterprises that buy those models. Accenture says it has about 799,000 people and around 9,000 clients.
Julie Sweet, Accenture's chair and CEO, said safety "requires both deep technical expertise and a clear understanding of how AI is used in the real world". Marc Warner, Accenture's chief technology officer and Faculty's CEO, said the Anthropic work "is exactly the kind of work Faculty was built to do".
What happens next
Anthropic says more evaluators will be announced in the coming weeks, and that Accenture will work with other AI developers in the same way. The commitment itself came from Dario Amodei's pacing essay, which Anthropic's post names. Anthropic didn't name the next evaluator.