Anthropic has signed a partnership with Accenture to embed independent safety evaluators inside the AI company, a move backed by at least $2 billion in combined investment over five years. The deal, announced September 18, 2026, gives Accenture's specialist AI business Faculty access to Anthropic's model training, deployment decisions, and internal teams - a level of visibility that goes well beyond today's standard external audits.
"Embedded evaluation is new, and many of the details about how it will operate are still being worked out," Anthropic said. Unlike external evaluators who test models after they ship, embedded teams will watch models take shape during training, follow the decisions that govern how those models are built and deployed, and speak directly to employees.
The partnership fulfills a commitment Anthropic CEO Dario Amodei made in his essay "We Must Pace the Frontier," where he argued for embedding evaluators within frontier AI labs. Anthropic said the safety of its models remains its responsibility. "Independent embedded evaluators do not reduce our accountability, but help to make it more verifiable," the company said.
What the evaluators will do
Faculty will conduct model evaluations and red-teaming, perform alignment assessments, and test model safeguards. Accenture brings practical knowledge from helping enterprises deploy AI across industries, and Anthropic said that perspective will shape how the evaluators approach safety. The goal is to assess how the company operates, verify that it keeps its safety commitments, and identify blind spots that internal teams might miss.
Embedded evaluators can also report incidents and give the public a more informed account of benefits and risks. But the mechanics of that reporting remain unsettled. "There are, as yet, no standards for what information embedded evaluators should have access to, or how they should report what they find," Anthropic said.
Funding and the push for standards
Anthropic will fund Accenture's work directly, a pragmatic choice given that no pooled or government funding mechanism exists yet. The company's June Advanced AI Framework called for pooled funding sources long-term. In the meantime, Anthropic is also in dialogue with nonprofit evaluators like METR to pilot elements of embedded evaluation using their own funding.
The partnership is non-exclusive. Anthropic plans to announce additional evaluators in the coming weeks, and Accenture will work with other AI developers in similar capacities. Anthropic said it expects frontier labs to work with several organizations at once, building toward an ecosystem of evaluators operating with shared standards.
Why this matters for executives and strategy leaders
The embedded evaluation model signals a shift in how frontier AI companies plan to demonstrate safety to regulators, enterprise customers, and the public. For CIOs and IT leaders evaluating AI vendors, the presence of independent evaluators with internal access could become a factor in procurement decisions - especially as standards for what those evaluators examine and disclose begin to solidify. Professionals tracking governance and compliance frameworks should watch how this experiment shapes the broader conversation about third-party AI auditing. Those building internal expertise can explore structured learning paths like AI Safety Engineering Courses and AI IT Strategy Training to prepare for a landscape where embedded evaluation may become an industry norm.
Your membership also unlocks:
