CA AB3211

Requires generative AI providers to apply difficult-to-remove provenance data to synthetic content identifying it as AI-generated, include the provider's name, and make this data difficult to remove or disassociate from the content.

Mandates generative AI providers create publicly accessible provenance detection tools allowing users to assess whether content was created or altered by AI systems and conduct adversarial testing before releasing new provenance data methods.

Prohibits providers and distributors from making available systems designed to remove provenance data from synthetic content in ways that would deceive consumers about content origin or history.

Requires newly manufactured recording devices sold in California to offer users the option to apply difficult-to-remove provenance data to nonsynthetic content and requires software updates for devices manufactured before July 1, 2026.

Requires large online platforms to label synthetic content with detected provenance data, label content with unknown provenance, produce annual transparency reports on deceptive synthetic content moderation beginning July 1, 2026, and establishes administrative penalties up to $100,000 for intentional violations and $25,000 for unintentional violations.