Anthropic taps Accenture as first embedded AI safety evaluator

Anthropic has selected Accenture, via its AI division Faculty, as its first embedded safety evaluator, with staff set to scrutinize the company's models and internal practices. The consulting firm will conduct red-teaming, alignment assessments, and model safeguard testing, with both companies investing at least $1 billion over five years. The announcement, which followed CEO Dario Amodei's proposal for third-party evaluators inside AI labs, surprised observers and drove Accenture's shares up 8% after hours. Anthropic highlighted Accenture's practical AI deployment experience and functional independence as key advantages, while noting no standards yet exist for evaluator access or communications. Anthropic plans to announce more evaluators soon and is in talks with non-profits like METR about piloting elements using their own funding. Critics view the self-policing scheme as a way to evade accountability, but Anthropic insists the evaluators make its responsibility more verifiable, not less.