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

Anthropic and Accenture said each company will commit at least $1 billion over five years to independent evaluation of advanced AI systems. Reuters reported that Accenture's Faculty unit will test models for vulnerabilities, alignment and safeguard performance, including red-team exercises.

Anthropic described an embedded-evaluation approach that gives independent evaluators access comparable to employees while preserving their autonomy. The initiative arrives as developers and policymakers place greater emphasis on demonstrating that increasingly capable systems can withstand adversarial testing before broad deployment.