Anthropic develops tools to confirm AI models' identities

The work targets provenance and trust in model-served responses.