Internet Lawyer Blog
California now has two distinct transparency regimes for generative artificial intelligence. Assembly Bill 2013 requires public documentation about training data, while Senate Bill 942—the California AI Transparency Act—addresses the provenance of AI-generated images, video, and audio. The laws overlap in subject matter, but their coverage, disclosures, and implementation dates are different.
AB 2013 has applied since January 1, 2026. SB 942, as amended by AB 853, became operative on August 2, 2026. Developers should analyze each law independently and document why a particular product, release, or business role is covered or excluded.
What Does AB 2013 Require?
AB 2013 added Civil Code sections 3110 and 3111. It applies to a developer that designs, codes, produces, or substantially modifies an AI system or service for use by members of the public. A developer’s affiliates and a hospital’s medical staff members are excluded from “members of the public” for purposes of this definition. A substantial modification includes a new version, release, or update that materially changes functionality or performance, including through retraining or fine-tuning. The statute also treats testing, validation, and fine-tuning by the developer as training.
For a generative AI system or service released on or after January 1, 2022, a covered developer must post training-data documentation on its website. Existing covered systems required a posting by January 1, 2026. For later releases and substantial modifications, the posting must occur before the product is made publicly available to Californians, whether access is paid or free.
The documentation must include a high-level dataset summary. Required subjects include dataset sources or owners; how the datasets support the product’s intended purpose; approximate numbers and types of data points; labeling characteristics; whether the data include material protected by copyright, trademark, or patent, or are entirely in the public domain; and whether they include personal information or aggregate consumer information. Developers must also address purchasing or licensing, cleaning and processing, collection periods, first-use dates, and the use of synthetic data. Because the statute says the listed information is not exhaustive, a mechanically completed checklist may not always be sufficient.
Are Any AB 2013 Systems Excluded?
The disclosure duty does not apply to a system or service whose sole purpose is security and integrity, or whose sole purpose is operating aircraft in the national airspace. It also excludes a system or service developed for national security, military, or defense purposes that is available only to a federal entity. The repeated use of “sole purpose” makes a broad reading of these exclusions risky for multipurpose products.
What Does SB 942 Require in 2026?
SB 942 currently applies to a person that creates, codes, or otherwise produces a generative AI system with more than one million monthly visitors or users that is publicly accessible in California. Although the statutory definition of a generative AI system includes text generation, the principal content duties apply to images, video, audio, and combinations of those media.
The Act does not apply to a product, service, website, or application that provides exclusively non-user-generated video game, television, streaming, movie, or interactive experiences. Because this exclusion depends on exclusivity, a mixed-purpose service should not assume that it falls outside the Act.
A covered provider must make a free AI detection tool publicly accessible, subject to reasonable security limitations. The tool must accept an upload or a URL, support an application programming interface, assess whether covered media was created or altered by the provider’s system, and output detected system-provenance data without outputting personal-provenance data. Providers must collect efficacy feedback and observe statutory limits on collecting and retaining user information and submitted content.
Providers must offer users the option to include a manifest disclosure identifying covered content as AI-generated. The disclosure must be clear, conspicuous, suitable for the medium, understandable to a reasonable person, and—when technically feasible—permanent or extraordinarily difficult to remove. This is an obligation to offer the option; the present statute does not say that every user must select it.
Separately, a covered provider must include a latent disclosure in covered AI-generated content created by its system. To the extent technically feasible and reasonable, it must convey the provider’s name, the system name and version, the creation or alteration date and time, and a unique identifier. The latent disclosure must be detectable by the provider’s tool, consistent with widely accepted industry standards, and difficult to remove when technically feasible.
How Does SB 942 Address Licensing and Penalties?
A provider that licenses its system must contractually require the licensee to preserve the system’s latent-disclosure capability. If the provider learns that a licensee disabled that capability, it must revoke the license within 96 hours, and the licensee must stop using the system. A violation can carry a $5,000 civil penalty, enforceable by the Attorney General, a city attorney, or county counsel. Each day of noncompliance is treated as a separate violation.
Could SB 1000 Change These Requirements?
Yes. As of September 4, 2026, SB 1000 has passed the Legislature and is enrolled but has not become law. If signed in its present urgency form, it would take effect immediately and materially revise the California AI Transparency Act. Among other changes, it would remove the one-million-user threshold, replace the “AI detection tool” framework with a “disclosure verification tool,” eliminate the manifest-disclosure option, and revise latent-disclosure, privacy, and licensee procedures. Businesses should follow the enacted statute while preparing for the possibility of an immediate change.
What Should Generative AI Companies Do Now?
- Classify every product separately. Determine whether the company is a developer under AB 2013, a covered provider under SB 942, or both.
- Build a defensible data inventory. Map training, testing, validation, and fine-tuning datasets to every AB 2013 disclosure category.
- Create a release gate. Do not publicly launch a covered release or substantial modification in California before its documentation is approved and posted.
- Test provenance controls. Verify that manifest options, latent disclosures, detection tools, URLs, and APIs function across supported media and distribution paths.
- Minimize verification data. Configure the detection workflow to avoid prohibited output or retention of personal-provenance data and unnecessary content.
- Update licensing terms. Preserve disclosure functionality through contracts, technical controls, notice procedures, and a rapid revocation process.
- Monitor pending and phased rules. Track SB 1000 and the January 1, 2027 duties that AB 853 created for large online platforms and GenAI hosting platforms.
What Is the Practical Takeaway?
California AI transparency compliance is not limited to adding a visible watermark. AB 2013 requires a repeatable process for publishing training-data information before covered releases. SB 942 requires product, privacy, provenance, and licensing controls for qualifying providers. The most reliable approach is to connect legal classification, dataset governance, engineering, release management, and contract administration in one documented compliance program.
Our law firm provides legal counsel regarding artificial intelligence, internet law, data privacy, cybersecurity, technology transactions, and regulatory compliance. Businesses developing or offering generative AI systems should consult qualified counsel regarding their products, users, disclosures, and distribution arrangements. This article is provided for informational purposes only and does not constitute legal advice.