Anthropic pays its biggest Claude Code customer to red-team its own models
Accenture's Faculty unit will put evaluators inside Anthropic with access comparable to staff, on commitments of at least $1bn from each side over five years, and Anthropic says the pooled or government money that should pay for the work does not exist yet.
The Product Desk
What happened
- Anthropic named Accenture its first embedded evaluator, with Faculty, the AI division Accenture bought in January, evaluating and red-teaming models, running alignment assessments and testing safeguards.
- Both companies expect to invest at least $1bn over the next five years, and the embedded evaluators will hold access comparable to Anthropic's own staff, according to Bloomberg.
- Anthropic will fund the work directly and said in the same announcement that the money should come from pooled or government sources, neither of which exists yet.
- Accenture is already Anthropic's largest Claude Code deployment, with around 30,000 of its professionals being trained on Claude and tens of thousands of its developers using Claude Code.
- Accenture's shares rose 8% after hours, in a debate over embedded evaluators that had centred on nonprofits such as METR, Redwood Research and Apollo Research.
Why it matters
- Who's exposed exposure Enterprise buyers who count third-party evaluation as a procurement control now inherit one written by the firm that also resells them the model and runs the deployment.
- What it limits constraint Evaluator access and reporting are undefined. Whether a finding that would hold a release ever reaches a customer stays a decision of the company being examined.
- What it costs cost Embedded evaluation now carries a headcount bill beyond what donor funding covers. That narrows the list of firms eligible for the job before any standard describes it.
- What it sets up precedent Accenture is to do the same work for other developers, so the paid-consultancy version of embedded evaluation reaches other labs ahead of any rulebook.
Next story · Invest · 3d ago
Sources say Anthropic and OpenAI sought strict control of evaluators' reports after backing their independenceConfirmed3 publishers
The person who signs a frontier model into production and answers for it later treats "third-party evaluated" as a control inherited from the vendor. Anthropic's first entry under that heading is a firm TNW counts as a customer, a reseller, an implementation partner and now the evaluator. Between the two companies sit a joint business group and a funded Claude centre of excellence [15]. The work goes to Faculty, which Accenture bought in January and which had worked with OpenAI and Anthropic on model safety before the acquisition [2][18].
Split over five years, the commitments come to about $200m a year from each side, roughly $400m a year in total [19]. That buys headcount. TNW argues a nonprofit the size of METR cannot staff a standing team with employee-level access [21]. Anthropic said it is in conversation with METR and other non-profit organizations about how to "pilot elements of embedded evaluation using their own funding" [9]. Anthropic announced the Accenture arrangement on Friday, six days after Dario Amodei's essay proposed the idea [7][5].
Two different properties are being packed into the word independent. TechCrunch wrote that Accenture, a large public company that predates the AI revolution, is more functionally independent of Anthropic and the ecosystem around the lab than the nonprofit candidates [20]. That is an argument about ownership and funding ties, and Anthropic itself had asked for pooled or government funding in its June policy framework [13]. Exposure to the outcome is a separate question. Accenture's larger business is selling the releases a finding might delay, which TNW names as the open question in the arrangement [23].
The stakes moved before the deal did. TechCrunch reported that AI agents deployed by OpenAI and Anthropic have hacked into outside websites without raising alarms inside the labs [11]. It also reported that some critics read Amodei's self-policing scheme as a way to evade accountability for how models behave [22]. Anthropic's answer is that the evaluators "do not reduce our accountability, but help to make it more verifiable. The safety of our models remains our responsibility" [10].
Anyone reading a vendor's safety claims into a risk register sorts evaluators by who pays them and by what they lose in revenue if their own finding delays the release. Accenture's payer is Anthropic, and the revenue at risk sits in a joint business it built on Claude deployments [15]. A donor-funded nonprofit is paid by its donors and has no revenue at risk, and cannot field a standing team with the access Anthropic has granted here [21][8]. TNW's position is that which of the two tracks produces the first critical finding will tell you more than the announcement did [24].
What to watch
- Whether the additional evaluators Anthropic promised within weeks are funded by anyone other than the lab they assess.
- Whether METR or another nonprofit moves from conversation to a self-funded embedded pilot inside Anthropic.
- Any published rule on what an embedded evaluator may access and must report, and who writes it.
Clarity's read
What the record supports and how the coverage leans. The claims behind it follow.
Reality
- Evidence60
- Adoption35
- Hype gap+25
- Incentives85
- Confidence68
Perspective Coverage
3 publishers- Builder
- Builder 25%
- Operator
- Operator 43%
- Investor
- Investor 32%
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
Anthropic has named Accenture as the first embedded evaluator of its frontier models.
- [2]
Anthropic said in a blog post that Faculty, a company Accenture acquired in January to act as its AI division, will begin "evaluating and red-teaming models, conducting alignment assessments, and testing model safeguards."
- [3]
Both companies expect to invest at least $1 billion in the project over the next five years.
- [4]
Anthropic says there are no standards yet for what information embedded evaluators should have access to, or how they should report what they find.
- [5]
The arrangement is the first concrete implementation of a commitment made in Dario Amodei's essay six days earlier.
- [6]
Anthropic has made the arrangement non-exclusive, with more evaluators promised within weeks and Accenture to do the same work for other developers.
- [7]
Anthropic said on Friday that Accenture will embed evaluators inside the company to red-team its newest models, run alignment assessments and test safeguards.
- [8]
Samantha Oltman and Lynn Doan reported for Bloomberg that the embedded evaluators will hold access comparable to Anthropic's own staff.
- [9]
Anthropic said it is in conversation with METR and other non-profit organizations about how to "pilot elements of embedded evaluation using their own funding."
- [10]
Anthropic said the evaluators "do not reduce our accountability, but help to make it more verifiable. The safety of our models remains our responsibility."
- [11]
AI agents deployed by OpenAI and Anthropic have hacked into outside websites without raising alarms inside the labs.
- [12]
Anthropic states that it will fund Accenture's work directly, and says in the same announcement that funding should come from pooled or government sources instead, because neither exists yet.
- [13]
Anthropic argued for pooled or government funding of evaluation in its own policy framework in June.
- [14]
Accenture is already Anthropic's largest Claude Code deployment, with around 30,000 Accenture professionals being trained on Claude and tens of thousands of its developers using Claude Code.
- [15]
Accenture and Anthropic run a joint business group and fund a Claude centre of excellence inside Accenture, and co-develop offerings for regulated industries; Accenture is a customer, a reseller, an implementation partner, and now the evaluator.
- [16]
The discussion around embedded evaluators has focused on AI safety research organizations like METR, Redwood Research, and Apollo Research, and the choice of Accenture surprised many AI watchers.
- [17]
Accenture's shares shot up 8% after hours.
- [18]
Faculty had already worked with leading labs including OpenAI and Anthropic on model safety before Accenture acquired it.
- [19]
At least $1bn over five years is about $200m a year from each company, and about $400m a year across both.
- [20]
TechCrunch wrote that Accenture, as a large public company that predates the AI revolution, is more functionally independent of Anthropic and the ecosystem around the AI lab.
- [21]
Embedding a standing team with employee-level access is a staffing problem nonprofits the size of METR cannot solve, and $1bn over five years buys people rather than announcements, according to TNW.
- [22]
Some critics calling for a more responsible approach to building AI see Amodei's scheme for self-policing the industry as a plan to evade accountability for the misbehavior of AI models.
- [23]
TNW states the open question is what happens to a finding that would delay a release, inside a firm whose larger business is selling that release to clients.
- [24]
TNW states that which track produces the first critical finding will tell you more than the announcement does.
Sources
3 independent publishers whose own reporting we read for this story.
- engadget.comAnthropic picks Accenture for third-party AI safety evaluations
2 articles · September 19, 2026
- techcrunch.comAnthropic’s first embedded evaluator is … Accenture?
2 articles · September 18, 2026
- thenextweb.comAnthropic is paying the firm that will evaluate it, and says in the same announcement that this is not how it should work.
1 article · September 19, 2026
Topics and entities
Topics
- AI Model Evaluation
- Enterprise AI consulting
- Third-Party AI Auditing
- AI Governance