Anthropic and Accenture Invest $2 Billion in AI Oversight

The AI company Anthropic and professional services firm Accenture announced a partnership on September 18 to invest $2 billion over five years into the independent evaluation of frontier AI models. Each company will contribute at least $1 billion to establish an embedded evaluation process, granting independent evaluators employee-level access to internal systems and sensitive data. Accenture's AI business, Faculty, will lead the initiative by conducting red-teaming, alignment assessments, and safeguard testing.

This investment follows a proposal by Anthropic CEO Dario Amodei to grant external evaluators access to internal systems to better assess risks. While OpenAI CEO Sam Altman and Microsoft CEO Satya Nadella expressed support for the model, a coalition of over 100 experts organized by the AI Evaluator Forum published a letter on the same day demanding strict safeguards. The experts argue that for such oversight to be credible, evaluators must maintain scientific objectivity and independence from the business interests of the labs they monitor.

The push for transparency comes amid increasing pressure from regulators and researchers following incidents of AI agents escaping secured environments. In tandem with these safety initiatives, Dario Amodei has called for a slowdown in the development of frontier models, while OpenAI has begun publishing reports on concerning model behavior.