Accenture’s AI unit Faculty will place evaluators inside Anthropic with employee-like access to training and release decisions. Each company expects to invest at least $1 billion over five years, but Anthropic pays for the work and Accenture is also a major Claude partner.
New to this? Read it in simple words
- On September 18, Anthropic and Accenture announced a deal on AI safety testing. A team from Accenture will work inside Anthropic and test its AI models.
- Anthropic says the team will get access similar to employees. Each company expects to invest at least $1 billion over five years.
- Anthropic pays for the work, and Accenture is also a major Anthropic business partner. This raises a question about how independent the testers are.
- The announcements do not give the team’s size or its right to publish findings. Anthropic admits there are no standards yet for how testers should report.
- Embedded evaluator
- An outside tester who works inside a company, with close access to its work.
- Alignment
- Work to make an AI’s goals and actions match what people want.
Outside evaluators with employee-like access
On September 18, Anthropic and Accenture announced that Accenture will build a team of embedded evaluators at Anthropic. The team will work alongside Anthropic’s own staff.
Faculty, Accenture’s specialist AI unit, will lead the work. It will cover red-teaming, which means trying hard to make a model fail, as well as alignment checks and safeguard tests. TechCrunch reports that Accenture bought Faculty in January.
Anthropic says the evaluators will get access similar to employees. They can watch models being trained, follow build and release decisions, talk directly to staff, and report incidents.
Each company expects to invest at least $1 billion over five years. Anthropic pays for the work.
The money is an expectation, not a fixed budget
Each company expects to invest at least $1 billion over five years. The announcements describe this as an expectation, not a binding pledge. TechCrunch reports that Accenture shares rose 8 percent in after-hours trading on September 18.
Anthropic will pay for Accenture’s work directly. It says long-term funding should come from pooled or government sources. It also admits that there are no standards yet for what evaluators can see or how they should report.
The deal is not exclusive. Anthropic says more evaluators will be named within weeks, and that it is talking with METR and other nonprofit groups. It presents the deal as a promise kept from Amodei’s pacing essay.
Independence needs more than a job title
Accenture is also a major business partner of Anthropic. In December 2025, the two companies created the Accenture Anthropic Business Group, with plans to train about 30,000 staff on Claude.
That creates a clear independence question, because Anthropic pays the evaluator and Accenture sells services built on Claude. TechCrunch notes that critics see the plan as a form of self-policing.
Neither announcement gives the team’s size, its reporting lines, or its right to publish findings. The Opus 5.5 launch post names METR and Frontier Design as its pre-release testers, not Accenture. The first public evaluator report will show how much this access really means.
Sources
Every fact in this story comes from the sources below. Open them to check our work.
- 1
- 2Primary source · September 18, 2026Accenture and Anthropic Partner to Build Team of Embedded Evaluators at Anthropic Accenture Newsroom
- 3
- 4Primary source · December 9, 2025Accenture and Anthropic Launch Multi-Year Partnership to Drive Enterprise AI Innovation and Value Across Industries Accenture Newsroom
- 5
We read both company announcements and Accenture’s 2025 partnership release, then compared them with TechCrunch. The investment figures are company expectations, not contracts. We found no public details about the team’s size, reporting lines, or publication rights.