Accenture Evaluators to Embed Inside Anthropic for AI Safety Tests
On Friday, September 18, 2026, Anthropic published a blog post and, with Accenture, issued a joint press release the same day. The subject was embedded evaluation—independent scrutiny of Anthropic’s frontier artificial intelligence models from a seat inside the company. Anthropic, the San Francisco public benefit corporation behind the Claude models, was naming its first partner for that work. Accenture, the global technology and consulting firm, was committing a dedicated team to work alongside Anthropic’s own people. Anthropic and Accenture each expected to invest at least $1 billion over five years in embedded evaluation and AI safety capacity.
Julie Sweet, chair and CEO of Accenture, was on the page: “Accenture is bringing together a dedicated team with deep AI, security and industry expertise to work alongside Anthropic.”
She set the full case out once: “Safety requires both deep technical expertise and a clear understanding of how AI is used in the real world. Embedded evaluation is an emerging area and we look forward to partnering with Anthropic to help accelerate the development of embedded evaluators, which we see as an important part of the safety landscape going forward.”
Faculty, Accenture’s specialist AI unit, would lead the work. The brief covered red-teaming models, conducting alignment assessments, and testing model safeguards. Accenture already helped businesses and governments deploy AI across many industries, and that practical grasp of how the technology is used would inform the safety approach brought to Anthropic’s systems.
Accenture completed its acquisition of Faculty on 16 March 2026, ABC Money reported. TechCrunch placed the same purchase in January. The UK company arrived with applied work already running through public systems. Faculty had developed the UK National Health Service’s Early Warning System during the COVID-19 pandemic. Its portfolio spanned government, defense, healthcare and infrastructure, and it had tested and evaluated models for some of the world’s leading AI labs while building complex systems meant to be safe and ethical by design.
Dr. Marc Warner, chief technology officer of Accenture and CEO of Faculty, stated the founding line in full: “Faculty, which is now part of Accenture, was founded on the belief that AI should be safe by design, not safe by accident. That conviction has guided our work with governments, defense, and healthcare systems across the world. Joining forces with Anthropic as embedded evaluators is exactly the kind of work Faculty was built to do—and as part of Accenture, we have the platform to do it at a scale that can genuinely move the industry.”
Ashley Capoot of CNBC framed the announcement as the first concrete step toward implementing Dario Amodei’s proposal to slow the pace of artificial intelligence development. Amodei had published a three-step plan; its opening plank granted third-party evaluators employee-level access inside the lab to verify safety practices and report incidents. Anthropic said it had unilaterally committed to that first step. The partnership with Accenture put the commitment into practice.
The novelty was the access. Embedded evaluators would work inside Anthropic with standing comparable to an employee’s—watching models take shape in training, following the decisions that govern how those models are built and deployed, and speaking directly with staff. From that seat they could verify that the company kept its safety commitments, identify blind spots, and file incident reports.
Anthropic’s blog post drew the line in plain terms: “[Embedded evaluators] do not reduce our accountability, but help to make it more verifiable. The safety of our models remains our responsibility.” Given the importance and urgency of the work, Anthropic would fund Accenture’s work directly.
Accenture planned to train approximately 30,000 professionals on Claude. That figure belonged to a commercial partnership that ran parallel to the safety work and stayed distinct from it. Accenture would become a premier AI partner for Claude Code, Anthropic’s coding assistant. The companies would co-invest in a Claude Center of Excellence within Accenture. Claude was available through AWS Bedrock, Google Cloud Vertex AI, and Microsoft Azure. The commercial terms covered workforce skills, engineering support, and the path from pilot projects into production.
Tim Fernholz of TechCrunch wrote that the choice of Accenture surprised many AI watchers. Conversation about embedded evaluators had pointed toward safety research organizations; a technology consulting giant was not the first name most expected. Markets moved the same day. Accenture shares rose 6.65 percent to $196.08 in the 24 hours around the announcement, ABC Money reported. TechCrunch cited an 8 percent after-hours jump. Trading volume was nearly four times the 20-day average.
The partnership was non-exclusive. Anthropic said it expected frontier labs to work with several organizations at once, and that further evaluators would be announced in the coming weeks. Accenture remained free to offer similar embedded-evaluation services to other AI developers.
No standards yet governed what information embedded evaluators should see or how they should report findings, and no settled system existed for funding the work. “Long-term, we think funding should come from pooled or government sources, as we called for in our Advanced AI Framework in June,” Anthropic stated. “As neither exists today, we plan to work with different evaluators under different funding arrangements.”
Faculty’s evaluators would take the first seats inside Anthropic. The next names were still unannounced.
“We’re sharing these early efforts now so people and other AI developers can see our process. We expect our approach to evolve as the field matures, and we’ll share more as our work begins and as we bring on additional evaluators.”
