IMPLICATOR
Anthropic named Accenture's Faculty unit its first embedded evaluator on Friday, giving the outside team access comparable to Anthropic employees. Each company expects to invest at least $1 billion over the next five years, while Anthropic will pay for Accenture's evaluation work directly. The selection puts an existing commercial partner inside the AI lab on the same day researchers called for evaluators not to have other significant commercial business with the companies they examine.
What Changed
- Anthropic named Faculty, Accenture's AI unit, its first embedded evaluator, with access comparable to Anthropic employees.
- Anthropic and Accenture each expect to invest at least $1 billion over five years, and Anthropic will fund Accenture's work directly.
- Anthropic says long-term funding should come from pooled or government sources, neither of which exists yet.
- More than 100 AI researchers, including Geoffrey Hinton, asked the same day for evaluators with no significant commercial business with the labs they check.
AI-generated summary, reviewed by an editor. More on our AI guidelines.
Inside the lab
Faculty, the AI business Accenture acquired, will evaluate and red-team Anthropic models, conduct alignment assessments and test safeguards. Its staff will be able to watch models take shape during training, follow decisions about how they are built and deployed, and speak directly with employees.
Accenture said the dedicated team will combine AI, security and industry expertise. Faculty has evaluated models for leading AI labs. Its prior projects include an early-warning system for the UK National Health Service during the Covid-19 pandemic.
Anthropic said the evaluators could use their access to check whether the company follows its safety commitments, identify blind spots and report incidents. This is the first concrete step on the embedded-evaluator commitment in Dario Amodei's Sept. 12 essay.
FREE AI BRIEFING · WEEKDAYS
Track who gets to check the frontier labs.
Get the AI stories shaping the day, with concise context from San Francisco. The briefing takes about five minutes and arrives at 4:45 a.m. Pacific, 7:45 a.m. Eastern.
Free. No hype. Unsubscribe anytime.
There are no standards yet for what information embedded evaluators should receive or how they should report their findings. The companies did not disclose when the work will begin or how large the Accenture team will be. They also gave no fee, success measure or procedure for handling a finding that conflicts with Anthropic's account.
Who pays
Anthropic said it will fund Accenture's work. In the same announcement, it said pooled or government funding should support independent evaluation over the long term. Neither funding mechanism exists today.
The expected $1 billion investment by each company is meant to build capacity over five years, not a disclosed contract price. Accenture shares rose 7% in extended trading on Friday after the announcement.
The agreement is nonexclusive. Anthropic plans to name more evaluators in the coming weeks, while Accenture can take similar assignments from other AI developers. Anthropic is also talking with METR and other nonprofit evaluators about pilots that would use the nonprofits' own funding.
Know someone who'd find this useful? ✉️ Email it to a friend in one click, or they can subscribe free here.
Anthropic said embedded evaluators do not reduce its responsibility for model safety. Their role, it said, is to make that responsibility more verifiable.
An existing customer
Accenture and Anthropic formed the Accenture Anthropic Business Group under a multiyear partnership announced in December 2025. About 30,000 Accenture professionals were set to receive Claude training under that agreement, and Claude Code was being rolled out to tens of thousands of Accenture developers. In March 2026, Accenture launched Cyber.AI, a security product powered by Claude.
Faculty had worked with Anthropic and OpenAI on model safety before Accenture bought the company. Earlier in 2026, Accenture projected that its work with key AI partners including Anthropic, OpenAI and Palantir would more than double over the following year.
More than 100 AI researchers, including Geoffrey Hinton, signed a public letter on Friday, Sept. 18, calling for embedded evaluators to be "meaningfully independent." Such evaluators "should not be owned or governed by frontier AI companies, should not have other significant commercial business with them, and should not accept any form of payment or other reward contingent on the evaluator's findings," the letter said.
Frequently Asked Questions
What will Accenture's evaluators do inside Anthropic?
Faculty, Accenture's AI unit, will evaluate and red-team Anthropic models, conduct alignment assessments and test safeguards. Its staff can watch models take shape in training, follow decisions on how they are built and deployed, speak with employees and report incidents.
Who is paying for the evaluation?
Anthropic will fund Accenture's work directly. Each company expects to invest at least $1 billion over five years to build capacity. Anthropic said pooled or government funding should support independent evaluation long term, but neither mechanism exists today.
Is the arrangement exclusive?
No. Anthropic plans to name more evaluators in the coming weeks, Accenture can do similar work for other AI developers, and Anthropic is talking with METR and other nonprofits about pilots funded by the nonprofits themselves.
What business ties do Accenture and Anthropic already have?
A December 2025 partnership formed the Accenture Anthropic Business Group, with about 30,000 Accenture professionals set to receive Claude training. In March 2026 Accenture launched Cyber.AI, a security product powered by Claude.
What remains undisclosed?
There are no standards yet for what evaluators can access or how they report findings. The companies did not disclose a start date, team size, fee, success measure or procedure for a finding that conflicts with Anthropic's account.
AI-generated summary, reviewed by an editor. More on our AI guidelines.
Marcus Schuler
Editor-in-Chief and founder of Implicator.ai. Former ARD correspondent and senior broadcast journalist with 10+ years covering tech. Writes daily briefings on policy and market developments. Based in San Francisco. E-mail: editor@implicator.ai
