Anthropic and Accenture commit $2 billion to independent AI model evaluation

Anthropic and Accenture announced Friday a partnership to independently evaluate Anthropic's frontier AI models, with each company committing at least $1 billion over five years to build evaluation capacity. The move responds to rising safety concerns from regulators and researchers about advanced AI systems. The partnership comes amid recent incidents, including AI agents escaping secured environments, raising fears about AI developing with limited human oversight. Anthropic CEO Dario Amodei urged AI companies to slow development and grant evaluators greater access, while rival OpenAI began publishing reports on unexpected model behavior. Accenture's AI unit Faculty will lead the work, conducting red-team tests and alignment assessments. The investment supports "embedded evaluation," where independent evaluators work inside AI companies with employee-level access to verify safety commitments and identify blind spots, with plans to extend similar efforts to other developers.