Anthropic Partners with Accenture to Launch Embedded AI Model Evaluation

Anthropic Partners with Accenture to Launch Embedded AI Model Evaluation Anthropic has announced a partnership with Accenture to embed independent evaluators directly inside the AI lab, with both companies expecting to invest at least $1 billion each over the next five years. Anthropic and Accenture each expect to invest at least $1 billion over five years to build independent evaluation capacity. The initiative will be led by Accenture's specialist AI business, Faculty, focusing on red-teaming, alignment assessments, and safety safeguards. Embedded evaluators will receive employee-like access to inspect training processes, governance decisions, and staff, though Anthropic maintains full responsibility for safety. The partnership is non-exclusive, with Anthropic directly funding Accenture while pursuing separate pilot discussions with nonprofit evaluators such as METR. Role and Mechanics of Embedded Evaluation Unlike standard external reviews, embedded evaluators operate inside AI companies with employee-level access. This allows them to monitor model development during training, inspect governance decisions, speak directly with employees, identify blind spots, and report incidents to provide informed public risk assessments. Anthropic stated that while this access increases verifiability, it does not reduce the company's ultimate accountability for model safety. Investment and Implementation The collaboration fulfills a goal outlined in CEO Dario Amodei's essay to embed evaluators within Anthropic. Faculty, Accenture's specialist AI business, will lead the work, leveraging enterprise deployment experience to evaluate models and test safeguards. Anthropic and Accenture each anticipate investing at least $1 billion into developing capacity in this domain over the next five years. Funding and Broader Ecosystem Standards Anthropic noted that formal industry standards for evaluator access and settled funding models do not currently exist. While Anthropic favors long-term pooled or government funding, it is directly financing Accenture's work for now. The agreement is non-exclusive, allowing Accenture to partner with other developers and Anthropic to introduce additional evaluators, alongside active discussions with nonprofit organizations like METR.

Anthropic 携手埃森哲推进前沿 AI 嵌入式独立评估 Anthropic 宣布与埃森哲达成非独家合作，推进前沿 AI 模型的嵌入式独立评估，双方预计在未来五年各自投资至少 10 亿美元建设相关能力。 合作由埃森哲旗下专业 AI 业务 Faculty 主导，涵盖模型评估、红队测试、对齐评估及模型安全机制测试。 双方各自预计在未来五年内投入至少 10 亿美元用于建设该领域的能力。 嵌入式评估员将拥有类似员工的内部访问权限，以跟踪训练过程、治理决策并识别盲点。 该合作为非独家性质，Anthropic 将直接资助埃森哲的工作，并正与 METR 等非营利评估机构展开对话。 合作内容与投资规模 该合作将由埃森哲旗下的 AI 业务部门 Faculty 主导，结合埃森哲在企业和政府部署 AI 的实践经验，开展红队测试、对齐评估及安全防护测试。Anthropic 与埃森哲各自预计在未来五年内投入至少 10 亿美元，以构建这一领域的评估能力。此举是落实 Anthropic 首席执行官此前关于在公司内部嵌入评估员承诺的重要一步。 嵌入式评估的机制与定位 与现有的外部评估不同，嵌入式评估员将进驻 AI 公司内部，获得与员工相当的访问权限。这使他们能够观察模型训练过程、追踪构建与部署决策，并直接与员工交流，从而核实安全承诺的履行情况并发现盲点。Anthropic 强调，引入独立评估员不会减轻公司的自身责任，而是让责任落实更加可验证。 资金模式与生态开放性 目前行业尚缺乏针对嵌入式评估的数据访问与报告标准，也缺乏成熟的资金机制。尽管 Anthropic 认为长期资金应来自公共池或政府，但在当前机制缺位的情况下，Anthropic 将直接资助埃森哲的工作。同时，Anthropic 也在与 METR 等非营利机构探讨自筹资金试点，并强调该合作为非排他性质，未来将与更多评估方合作，埃森哲亦可与其他 AI 开发者展开类似合作。

We're partnering with Accenture on independent evaluation of frontier AI. This is an important step toward the commitment, made in our CEO’s [essay](https://darioamodei.com/post/we-must-pace-the-frontier) “We Must Pace the Frontier,” to embed evaluators within Anthropic. The partnership will be led by Faculty, Accenture’s specialist AI business, and will include evaluating and red-teaming models, conducting alignment assessments, and testing model safeguards. Accenture helps businesses and governments deploy AI across many industries. Their understanding of how enterprises use AI in practice informs their safety approach, and they will bring that perspective to evaluating our models. Anthropic and Accenture each expect to invest at least $1 billion in building capacity in this area over the next five years. Embedded evaluation is new, and many of the details about how it will operate are still being worked out. Unlike today’s external evaluators, embedded evaluators will work inside AI companies, with access comparable to an employee's. That access allows them to watch models take shape in training, follow the decisions that govern how those models are built and deployed, and speak directly to employees. From this vantage point, embedded evaluators can assess how a company operates, verify that it is keeping its safety commitments, and identify blind spots. They can also report incidents and give the public a more informed account of benefits and risks. To be clear, independent embedded evaluators do not reduce our accountability, but help to make it more verifiable. The safety of our models remains our responsibility. There are, as yet, no standards for what information embedded evaluators should have access to, or how they should report what they find. There is also no settled system for funding independent evaluation. Long-term, we think funding should come from pooled or government sources, as we called for in our [Advanced AI Framework](https://www-cdn.anthropic.com/files/4zrzovbb/website/0a58d567024a8b448ff15158ebc3625328dfcc1f.pdf) in June. As neither exists today, we plan to work with different evaluators under different funding arrangements. Given the importance and urgency of this work, Anthropic will fund Accenture's work directly. We are also in dialogue with [METR](https://metr.org/) and other nonprofit evaluators to pilot elements of embedded evaluation using their own funding. Ultimately, we believe frontier AI needs an ecosystem of evaluators operating with shared standards. We expect frontier labs to work with several organizations at once. Our partnership is non-exclusive; Anthropic will work with other evaluators to be announced in the coming weeks, and Accenture will work with other AI developers in similar capacities. We'll continue to train and release frontier models, and we want independent evaluators working alongside us as we do. 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.