California's AI Transparency Act (SB 942): What Changes for Images

California's AI Transparency Act became operative on 2 August 2026 under AB 853. SB 1000 changed the provider rules immediately when it was signed on 30 September 2026. AB 2713, signed the same day, revised the platform rules that start in 2027. This guide summarizes the chaptered laws as reviewed on 12 November 2026; it is not legal advice.

Who the provider rules cover

SB 1000 removed the one-million-monthly-user threshold. A covered provider is now a person that creates, codes or otherwise produces a generative AI system publicly accessible in California. The law has specific exclusions, including a temporary exclusion before 1 January 2029 for systems designed primarily as assistive technology for people with disabilities.

Latent disclosures and verification

To the extent technically feasible, providers must include a latent disclosure in images, video and audio their systems create or alter, except for minor modifications defined by the law. The disclosure identifies the provider, system and version, creation or alteration time, a unique identifier, and whether the content was created or altered. It must meet the law's durability, compatibility and standards requirements.

Providers must make a disclosure verification tool available at no cost. It must support checks on uploaded content or a URL and access through technology such as an API, subject to the law's access and privacy rules. A qualifying third-party tool can satisfy this requirement. SB 1000 removed this chapter's earlier requirement to offer a visible manifest-disclosure option; other disclosure rules can still apply.

What happens if a licensee disables compliance

If a provider knows an identifiable licensee modified a licensed system so it no longer complies, the provider must terminate authorization or send a noncompliance notice within 72 hours of discovery. A notified licensee has 96 hours after receiving the notice to bring the system into compliance or stop using or making it available, and report its action. The provider must report specified failures to respond or comply to the Attorney General.

Platform and device requirements

• From 1 January 2027, covered large online platforms must detect and surface qualifying provenance information. AB 2713 also prohibits knowingly stripping qualifying system provenance data or digital signatures from uploads, distributed content or downloads, to the extent technically feasible. These duties do not require platforms to retain, display or provide downloads of personal information.
• From 1 January 2027, GenAI hosting platforms must not knowingly make available systems that do not place the disclosures required by the Act.
• From 1 January 2028, the capture-device rules apply to devices first produced for sale in California on or after that date. Manufacturers must offer latent-disclosure functionality and enable it by default, within the law's technical-feasibility and standards limits. This is not a blanket requirement to retrofit every existing device.

Enforcement

The Act provides for public civil enforcement by the Attorney General, a city attorney or county counsel. Its general penalty is $5,000 per violation, with each day of continuing violations by specified regulated entities counted separately. False claims that a system primarily serves as assistive technology are subject to a separate rule and higher penalty. Consult the statute for the applicable provision.

What this means for image users

Personal metadata such as GPS coordinates is different from system provenance that identifies how content was created. The Act assigns different duties to providers, licensees, hosting platforms, large online platforms and device manufacturers. It does not establish a blanket permission to remove AI-origin or copyright information. Check the rules that apply to your role, location and intended publication, along with the generator's terms and the platform's disclosure requirements.