Anthropic embeds Accenture as its first safety evaluator — about $1 billion each over five years
Faculty staff get employee-level access to red-team, align and test safeguards. Anthropic will fund Accenture’s work directly. The deal is the first concrete step in Amodei’s slowdown plan. Non-exclusive; METR talks continue.
Feature1 minEasy
Priya Shah1 min read
- Anthropic
- Accenture
- Faculty
- Evaluators
- Safety

Photo: CNBC / Chance Yeh / Getty ImagesSource page
Anthropic said on Friday it will embed staff from Faculty, Accenture’s specialist AI unit, inside the company to evaluate and red-team frontier models, run alignment assessments and test safeguards. Each side expects to invest at least $1 billion over five years in building that capacity, Anthropic wrote. Reuters put the combined figure as a $2 billion commitment; Accenture shares rose 7 percent in extended trading. Because there is still no pooled or government pot for independent evaluation — the funding Anthropic called for in its June Advanced AI Framework — the lab will pay for Accenture’s work directly. Ashley Capoot at CNBC called the deal the first concrete step in Dario Amodei’s three-part plan to slow the frontier. Embedded evaluators were the piece he said Anthropic would do unilaterally.
The point of “embedded,” in Anthropic’s note, is employee-level access: watching models in training, following build-and-deploy decisions, speaking to staff, verifying safety commitments, reporting incidents. The company said that does not reduce its own accountability. There are no agreed standards yet for what evaluators see or how they publish. The partnership is non-exclusive. Anthropic said it is in dialogue with METR and other nonprofit evaluators on their own funding, will name more partners in the coming weeks, and expects Accenture to do similar work for other developers. Reuters noted OpenAI’s Wednesday turn to regular misalignment reports. This is a staffing and money announcement, not a published eval.


