Wiki · Governance · Last reviewed June 25, 2026

California AI Transparency Act

The California AI Transparency Act turns synthetic-media provenance, disclosure, and detection into duties for covered generative AI providers and, through later amendments, certain platforms and capture-device manufacturers.

Category: Governance Updated: June 25, 2026 Tags: California, synthetic media, provenance, watermarking, disclosure

Definition

The California AI Transparency Act is Chapter 25 of Division 8 of the California Business and Professions Code, created by SB 942 in 2024 and amended by AB 853 in 2025. The statute is aimed at synthetic media identification rather than general model safety. It asks whether generated or altered image, video, and audio content can carry provenance signals that users and platforms can inspect.

SB 942 was approved on September 19, 2024. AB 853, approved on October 13, 2025, delayed the operative date from January 1, 2026 to August 2, 2026 and added platform, hosting, and capture-device provisions with later start dates.

Scope

The central regulated actor is the covered provider: a person that creates, codes, or otherwise produces a generative AI system with more than 1,000,000 monthly visitors or users and that is publicly accessible in California. The statute defines GenAI broadly, but the provider disclosure and detection-tool duties focus on image, video, audio, or combinations of those media.

Covered providers must make a no-cost AI detection tool available. The tool must let users assess whether covered media was created or altered by that provider's GenAI system, output detected system provenance data, avoid outputting personal provenance data, support uploads or URLs, and support API invocation. Providers may limit access for demonstrable security or integrity risks.

The law distinguishes manifest and latent disclosures. A manifest disclosure is meant to be perceived by a person; a latent disclosure is present but not manifest. Providers must offer users an option to include a manifest disclosure in covered media and must include latent disclosures when technically feasible and reasonable, with information such as provider name, system name and version, creation or alteration time, and a unique identifier.

AB 853 added later layers. Beginning January 1, 2027, large online platforms must detect certain standards-compliant provenance data, surface available system provenance information, and avoid knowingly stripping system provenance data or digital signatures when technically feasible. GenAI hosting platforms receive a parallel 2027 duty for hosted systems. Beginning January 1, 2028, covered capture-device manufacturers must support latent disclosures in captured content, subject to technical feasibility and widely adopted specifications.

Agent Context

Agent systems make the law operational. A media agent can generate a video, edit a voice track, resize an image, upload a file, repost it, or hand it to another service. Each step can preserve, damage, replace, or strip provenance signals.

An agent workflow should log the source model, media type, transformation chain, disclosure state, export format, platform destination, and whether provenance survives compression, screenshotting, clipping, or reposting.

Governance Use

For a provider, the governance record should document whether the system crosses the statutory user threshold, which outputs are covered media, where the detection tool lives, whether an API is available, what provenance fields are emitted, and what access limits exist for abuse or security reasons.

For a platform, the record should identify whether it qualifies as a large online platform, what standards-compliant provenance data it detects, how users inspect it, and what transformations might remove metadata or digital signatures. For a hosting platform, the record should show how hosted models are checked before release.

For procurement, California's law can be turned into a vendor questionnaire: latent disclosures, detector behavior, export formats, license controls, and metadata survival in common publishing workflows.

Limits

The Act is not a universal AI law. It does not cover every model, user, jurisdiction, platform, or content type. It excludes products and services that exclusively provide non-user-generated video game, television, streaming, movie, or interactive experiences.

Provenance is evidence about origin and modification, not proof that content is true, consensual, lawful, harmless, or fairly contextualized. A human-made file can still mislead; an AI-labelled file can still be benign satire, illustration, accessibility support, or routine production work.

The statute inherits technical fragility. Metadata can be stripped, watermarks can be attacked, signatures can be lost in conversion, detection tools can miss altered outputs, and visible labels can be ignored or misunderstood.

Review Record

Source Discipline

Use California Legislative Information for bill text and chapter history. Use Governor of California releases for signing context. Treat law-firm alerts, vendor guides, and trade commentary as secondary explanation, not authority for statutory deadlines, thresholds, definitions, or penalties.

Spiralist Reading

Spiralism reads the California AI Transparency Act as a discipline for the synthetic public record. It cannot decide whether an image should persuade us or whether a platform acted responsibly. It does make one demand legible: generated media should not drift through civic life without a trace of the machine, tool, or institution that helped make it.

Sources

Return to Wiki