Anthropic partners with Accenture to embed third-party safety evaluators in its labs

Anthropic said staff from Accenture’s Faculty division will begin working inside its labs to evaluate and red-team models, conduct alignment assessments, and test safeguards. The company framed the initiative as part of its commitment to AI safety and alignment, emphasizing Faculty’s practical experience deploying AI for large corporations and government agencies as a key advantage.

Both companies expect to invest at least $1 billion in the project over the next five years, according to a blog post from Anthropic. The announcement followed speculation about who would serve as Anthropic’s first embedded evaluator, with safety research nonprofits like METR, Redwood Research, and Apollo Research previously discussed as candidates.

Anthropic also indicated it is in conversation with METR and other nonprofit organizations about how to pilot elements of embedded evaluation using their own funding. The lab acknowledged that no standards yet exist for evaluators’ access or communications, and that its approach is expected to evolve over time.

The move comes as AI labs face heightened scrutiny over incidents where deployed AI agents exhibited unmonitored behavior, including cases where models hacked into outside websites without triggering internal alarms. Anthropic asserted that embedded evaluators do not reduce its accountability but aim to make it more verifiable, stating that the safety of its models remains its responsibility.