California SB 942: the AI Transparency Act, section by section

California's AI Transparency Act requires a covered provider (anyone who produces a generative AI system that is publicly accessible in California) to embed a latent disclosure, in practice an imperceptible watermark, in the image, video and audio its system creates or alters, to offer a free disclosure verification tool, and to act when a licensee stops keeping the disclosure intact. It has been operative since 2 August 2026, and each day of violation is a separate $5,000 penalty.

The Act is California Business and Professions Code, Chapter 25 (commencing with Section 22757), enacted by SB 942 and amended by AB 853, then by SB 1000 and AB 2713 on 30 September 2026. SB 1000 is an urgency statute, so its changes applied the day it was signed. Unlike EU AI Act Article 50, SB 942 is specific about what the mark has to say.

§ 22757.1 defines a covered provider as a person that creates, codes or otherwise produces a generative artificial intelligence system that is publicly accessible within the geographic boundaries of California. Until SB 1000, the system also had to have over 1,000,000 monthly users. That threshold is gone.

Four things follow, and each catches somebody out:

• There is no size test. Not users, not revenue, not headcount. A small team with a niche product is covered as much as a large one.
• It is not a California-company test. Public accessibility within California is what matters, so a provider anywhere in the world can be covered.
• The 2,000,000 figure is not a provider threshold. The number people often quote belongs to AB 853's separate large online platform duties, below.
• The exemptions are narrow. SB 1000 leaves two: products that provide exclusively non-user-generated videogames, and, until 1 January 2029, systems designed primarily as assistive technology.

§ 22757.3 requires a latent disclosure, to the extent it is technically feasible, in image, video or audio content the system creates or alters. It can convey the fields directly or through a link to a permanent website. This is the part of the statute that decides whether a watermark is adequate, because a bare “this is AI” signal does not carry these fields.

From 1 January 2029 the disclosure must also say whether the system is designed primarily as assistive technology.

The statute also requires the latent disclosure to be compatible with the provider's disclosure verification tool, compliant or interoperable with widely recognized industry standards, and permanent or extraordinarily difficult to remove or tamper with.

No visible label any more

As first enacted, SB 942 also made a covered provider offer users the option of a visible disclosure identifying content as AI-generated. SB 1000 deleted that option. The latent disclosure is now the only disclosure the statute asks a provider to put in the content.

§ 22757.2 is the requirement most often missed, because it is not about the content at all — it is a product you must offer. A covered provider must make available a disclosure verification tool (the “AI detection tool” before SB 1000) that:

• is available at no cost to the user;
• lets a user assess whether image, video or audio content was created or altered by the provider's system;
• outputs any system provenance data detected in the content;
• outputs no personal information unless the person it relates to expressly consents;
• is publicly accessible (reasonable limits against security risks or misuse are allowed), and lets a user upload content or provide a URL;
• supports an API, so a user can invoke it without visiting the provider's website;
• collects user feedback on its performance.

There is a strict privacy limit attached: the provider may not collect, use, retain or sell personal information beyond what is strictly necessary, though users can opt in to be contacted. A tool that logs everything submitted to it is its own compliance problem.

A third-party tool is allowed

SB 1000 lets a provider meet this duty by directing users to a third-party disclosure verification tool. The tool must comply with § 22757.2, be compatible with the latent disclosures in the provider's content, and be clearly and conspicuously accessible through the provider's generative AI user interface.

Under § 22757.3(b), as amended by SB 1000, a covered provider that licenses its system to a third party must notify the licensee of its obligations under the Act.

If the provider knows an identifiable licensee has modified the system so that it no longer complies, the provider has 72 hours from discovery to either end the licence, or notify the licensee of the noncompliance and of its duty to report. The licensee then has 96 hours from that notice to fix the system or stop using it, and to report back. If it does not, the provider reports it to the Attorney General.

SB 1000 also says the provider has no duty to monitor or investigate its licensees. The clock starts when the provider knows. In practice this is an operational duty, not a legal one: once you know, you have three days to act, and most providers discover they have no mechanism for that.

AB 853 moved the provider duties to 2 August 2026 and extended the Act to three new parties. If you are not a covered provider, you may still be one of these.

The platform duty is worth reading carefully: it is a multi-standard detection obligation, not a single-vendor one. Under AB 2713, a platform need not act on provenance that is not compliant or interoperable with widely adopted specifications from an established standards-setting body — C2PA first — and a watermark is the layer that survives when a manifest has been stripped. The detail is on our AB 853 page.

1. 19 September 2024SB 942 signed and chaptered. Source
2. October 2025AB 853 signed. Provider duties moved from 1 January 2026 to 2 August 2026; platform, hosting and device duties added. Source
3. 2 August 2026Covered-provider duties operative: the latent disclosure, the free tool and the licensee duties.
4. 30 September 2026SB 1000 signed, in force immediately as an urgency statute: no user threshold, the disclosure verification tool (a third party's allowed), no visible-label option, a fifth latent-disclosure field, and the 72-hour licensee rule. AB 2713 signed the same day, rewriting the large online platform duties. Source
5. 1 January 2027Large online platform duties (as amended by AB 2713) and GenAI hosting platform duties operative.
6. 1 January 2028Capture device manufacturer duties operative.
7. 1 January 2029The assistive-technology exemption and the $50,000 penalty for falsely claiming it end; the latent disclosure must also say whether a system is designed primarily as assistive technology.

Under § 22757.4, a covered provider that violates the chapter is liable for a civil penalty of $5,000 per violation, and each day a violation continues is a discrete violation. Actions may be brought by the Attorney General, a county counsel or a city attorney.

That daily multiplier is the part worth internalising. This is not a one-off fine — an unresolved gap accrues. A licensee's violation is treated separately from the provider's own duties.

SB 1000 adds a second penalty until 1 January 2029: $50,000 per violation, per day, for falsely representing a generative AI system as designed primarily as assistive technology, the category that is exempt until then.

These two are routinely conflated, and the difference decides what your watermark has to carry.

A system built only for Article 50 — a bare “AI-generated” signal — does not satisfy SB 942, because it cannot name the provider or the version. A system built for SB 942 generally does satisfy Article 50(2). See our Article 50 page for the other side.

One more commonly-confused neighbour: AB 2013 is the training-data documentation law. Different subject, frequently searched alongside SB 942.

SB 942 is the regime where Verda's payload design matters most, because the statute enumerates fields rather than leaving them open.

On the AB 853 platform side, as amended by AB 2713, the requirement is to read provenance that follows widely adopted specifications. Verda writes signed C2PA manifests as well as the watermark, and is registered on the C2PA soft-binding algorithm list for image, audio and video.