Anthropic let Accenture safety-test its AI from inside. Both expect to invest $1 billion.

Faculty staff will work inside Anthropic, "evaluating and red-teaming models, conducting alignment assessments, and testing model safeguards." Both companies expect to invest at least $1 billion in the project over the next five years.

The idea came from a blog post by Anthropic chief executive Dario Amodei. The arrangement is non-exclusive. More evaluators will be named in the coming weeks, and Anthropic is talking to METR and other nonprofits. Anthropic also says no standards yet exist for what evaluators may access or say.

Accenture's shares rose 8% after hours, TechCrunch reported. Some critics see the scheme as the industry policing itself. Anthropic says the evaluators make its accountability "more verifiable", while responsibility for the safety of its models stays with Anthropic.