Anthropic has granted Accenture employee-level access to its artificial intelligence model development process, making the consulting firm its first embedded safety evaluator, according to a report published by Channel Insider. Specialists from Faculty, an Accenture team, will work directly alongside Anthropic staff to assess models, test protective measures, conduct alignment reviews, and red-team systems as they're being built. The arrangement is part of a broader partnership in which both companies expect to commit at least $1 billion over five years.
Anthropic said the embedded evaluators will receive access comparable to what employees have, letting them observe models while they're training, track decisions about how systems are constructed and released, and communicate directly with staff members. That level of access is designed to help evaluators confirm safety commitments, spot blind spots, and flag incidents, the company stated. Accenture noted that Faculty brings experience from AI projects spanning government, defense, healthcare, and infrastructure sectors. The arrangement is non-exclusive, and Anthropic is in discussions with METR and other nonprofit evaluators, expecting frontier labs to collaborate with multiple evaluation organizations rather than depend on one outside reviewer. Accenture CEO Julie Sweet said the work demands both technical knowledge and insight into how organizations actually deploy AI.
According to the report, CNBC described the partnership as the first tangible implementation of Anthropic CEO Dario Amodei's proposal for frontier AI companies to provide outside evaluators continuous, employee-level access. Anthropic separately stated that outside evaluation doesn't transfer responsibility for model safety away from the company itself. The company acknowledged the model is still evolving, noting there aren't established standards for what embedded evaluators should access, how findings should be documented, or how independent evaluations should be funded—Anthropic will fund Accenture's work directly for now.
While the Anthropic engagement involves specialized frontier-model work that doesn't immediately create a managed service provider offering, the underlying tasks point toward areas channel firms could increasingly face as customers put more AI into production, the report finds. If enterprises adopt similar evaluation practices, consultancies, security partners, and MSPs could see demand for AI red teaming, governance reviews, safeguard testing, and ongoing monitoring. MSPs already managing customer security, cloud environments, or infrastructure may be especially positioned to extend those relationships into AI oversight as businesses seek help assessing how deployed systems behave and whether controls continue functioning. Accenture's latest Pulse of Change survey found that 82% of more than 3,000 executives were increasing AI investment, while only 23% reported achieving widespread, sustained business value—a gap the report suggests evaluation services could help address.
For MSPs and other channel firms, the standards around embedded evaluation will be worth tracking. If embedded evaluation becomes more standardized and repeatable, AI assurance could develop into another layer of managed security, governance, and consulting work alongside the systems partners already help customers deploy. Until clear access, reporting, funding, and independence rules emerge, however, the opportunity remains an early-stage model rather than a ready-made MSP service. The channel may soon face a choice between building evaluation capabilities early or waiting for clearer guardrails—a timing decision that could determine who captures the first wave of enterprise AI oversight contracts.