Anthropic Taps Accenture as Its First Embedded AI Safety Evaluator in $1 Billion Deal
Accenture’s Faculty team will work inside Anthropic to red-team models, assess alignment and test safeguards—the lab’s first embedded evaluation partnership with an outside firm. The companies plan to invest at least $1 billion over five years.
Anthropic announced in September 2026 that Accenture will become the first outside firm to work inside the AI lab, scrutinizing its models and staff as part of a new embedded evaluation program.
Staff from Faculty — an AI division Accenture acquired in January — will work within Anthropic to evaluate and red-team models, conduct alignment assessments, and test model safeguards. Both companies expect to invest at least $1 billion in the project over the next five years.
The selection of Accenture surprised many AI watchers and moved markets: Accenture shares rose 8% after hours following the announcement. Discussion around embedded evaluators had largely centered on AI safety research organizations such as METR, Redwood Research, and Apollo Research. Anthropic said it is still in conversations with METR and other nonprofits about piloting elements of embedded evaluation using their own funding, and that additional evaluators will be announced in the coming weeks.
Anthropic pointed to Accenture’s practical experience deploying AI for large corporations and government agencies as a key advantage. The lab also noted that, as a large public company predating the AI era, Accenture is more functionally independent of Anthropic and the broader AI lab ecosystem than specialist safety organizations might be.
The move follows recent incidents in which AI agents deployed by OpenAI and Anthropic hacked into outside websites without triggering internal alarms — developments that have raised the stakes for external model evaluation. Anthropic acknowledged that no standards yet exist for evaluators’ access or communications, and said its approach will evolve over time.
Critics have argued that Anthropic CEO Dario Amodei’s embedded evaluator framework amounts to industry self-policing that could limit accountability. Anthropic pushed back, stating that the evaluators “do not reduce our accountability, but help to make it more verifiable,” and that model safety remains the company’s own responsibility.