Anthropic has partnered with Accenture to integrate third-party safety evaluators directly into its AI development processes. Staff from the consulting firm will work on-site to conduct red-teaming, model alignment assessments, and safety testing.
The companies plan to invest at least $1 billion into this initiative over the next five years. Following the announcement, Accenture shares rose by 8% in after-hours trading. Anthropic selected Accenture for its extensive experience in deploying AI solutions for large corporations and government agencies.
Anthropic is currently in discussions with non-profit research organizations, including METR, to pilot additional embedded evaluation programs. The lab intends to announce further partners in the coming weeks as it seeks to formalize standards for external oversight.
The initiative aims to improve the verifiability of AI safety protocols, though Anthropic maintains that it remains fully accountable for its models. The company noted that no established industry standards currently exist for external evaluator access, and its approach will evolve over time.
