The Anthropic Accenture partnership commits $2 billion over five years to embed independent safety evaluators inside Anthropic's AI labs.