California AI Transparency Act.

Assembly Member Wicks' California AI Transparency Act establishes new requirements for generative AI systems, online platforms, and device manufacturers to provide transparency in AI-generated content through mandatory disclosures and detection tools.

The legislation requires large online platforms with over 2 million monthly users to label content containing machine-readable provenance data, indicating the AI system used and any available digital signatures. These platforms must allow users to inspect provenance information while being prohibited from removing system provenance data or retaining personal provenance data from shared content. Companies operating generative AI systems with over 1 million monthly users must provide free detection tools for assessing whether content was created or modified by their systems.

Beginning January 2028, manufacturers of cameras, phones, and other capture devices must include options for embedding latent disclosures in recorded content, containing details like device information and creation timestamps. The bill mandates that these features be enabled by default while giving users the ability to opt out. Manufacturers must also make secure hardware-based provenance data available to third-party applications. The law explicitly prohibits the development of tools designed to remove these disclosures.

Violations incur civil penalties of $5,000 per day, enforceable through actions by the Attorney General, city attorneys, or county counsel. The legislation includes provisions for injunctive relief and attorney fees in cases involving third-party violations.