California AB 2602 and AB 1836, both signed in September 2024 and in force since January 1, 2025, are the most detailed performer protections in the United States for AI digital replicas. For fashion brands, which frequently work with Los Angeles-based models, photographers, and creative talent, these laws have immediate practical implications that most brands haven't yet integrated into their standard agreements. And they are only the start of what Sacramento has been building.

What AB 2602 covers

AB 2602 is a contract law. It does not ban digital replicas. It makes certain contract provisions unenforceable, which in practice means your standard model release may not give you the AI rights you think it does.

Under the statute, a "digital replica" is a computer-generated, highly realistic electronic representation that is readily identifiable as the voice or visual likeness of a real individual. A contract provision allowing the creation and use of a digital replica in place of work the individual would otherwise have performed in person is unenforceable if both of the following are true:

  • The provision does not include a reasonably specific description of the intended uses of the replica, and
  • The individual was not represented by legal counsel or a labor union when the deal was negotiated.

The law was drafted with the entertainment industry in mind, it was a SAG-AFTRA priority during the 2023 strikes, but its text covers contracts for the performance of personal or professional services generally. A model booked for a campaign shoot is performing personal services. If your release contains a boilerplate grant like "brand may use AI to modify or extend the images," with no specific description of the intended AI uses, and the model signed without a lawyer or union representation, you should assume that grant is unenforceable in California.

The practical upshot for brands:

Be specific. Spell out the AI uses you actually intend: body or garment alteration, likeness extension across additional looks, synthetic reshoots. A vague omnibus AI clause is exactly what the statute targets.

Paper the representation prong. Encourage talent to have the agreement reviewed by counsel or their union, and document that they did. It is the difference between an enforceable replica clause and an unenforceable one.

Pay for what you're getting. The statute doesn't set compensation terms, but if AI extends a model's appearance across more looks than they were photographed in, negotiating real compensation for that use is both good relations and good evidence that the consent was informed.

One caution: I have seen claims circulating that California regulators have "interpreted" AB 2602 to automatically cover routine AI retouching in fashion imagery. I have not found any official guidance saying that. What is true is simpler and still important: if the AI output is a highly realistic, readily identifiable rendering of your model doing work she didn't actually perform, you are in digital replica territory and your contract needs to say so specifically.

What AB 1836 adds: deceased personalities

AB 1836 amends California's post-mortem right of publicity, Civil Code section 3344.1, to cover AI digital replicas of deceased personalities. Producing, distributing, or making available a digital replica of a deceased personality's voice or likeness in an expressive audiovisual work or sound recording without consent from the rights holder creates liability of at least $10,000, or actual damages if greater. The underlying post-mortem right runs for 70 years after death.

For fashion brands, this has specific implications:

Archival imagery reactivation. If you're thinking about using AI to recreate past campaigns featuring deceased talent in a contemporary context, get estate consent first.

AI-generated icons. Using AI to resurrect the likeness of a deceased fashion icon in campaign film or video requires clearance with whoever controls the rights. The statutory floor of $10,000 per violation is a floor, not the ceiling, and the reputational damage of an unauthorised digital resurrection would cost more than the statute ever could.

SB 942 and AB 853: the watermarking layer arrives August 2026

The California AI Transparency Act, SB 942, takes a different approach: instead of regulating you, it regulates your tools. Covered providers, generative AI systems with more than a million monthly users accessible in California, must offer a free public AI detection tool, give users the option of a visible AI disclosure, and embed latent, machine-readable provenance data in AI-generated images, video, and audio. Penalties run $5,000 per violation, and each day of non-compliance is a separate violation.

The timeline moved: AB 853, signed in October 2025, pushed the operative date from January 1, 2026 to August 2, 2026, deliberately aligning with the EU AI Act's Article 50 deadline. AB 853 also extends the framework: large online platforms must surface provenance data starting January 1, 2027, and capture-device makers follow after that.

Why designers should care even though the obligations fall on AI companies: provenance metadata is becoming infrastructure. The images your tools generate will carry machine-readable origin data, platforms will read it, and consumers will see it. Brands that treat provenance as part of their asset pipeline now will not be surprised later. I walk through the mechanics in our provenance guide, and our disclosure generator covers the visible-label side.

AB 2013 and SB 53: the transparency stack

Two more pieces complete the picture. AB 2013, effective January 1, 2026, requires developers of generative AI systems to publicly post documentation about their training data. If you have ever wondered whether your lookbooks are in a training set, this is the first law that forces developers to say something about it. I cover it in detail in my AB 2013 explainer, and it pairs with the bigger question in Your Designs Are Probably in an AI Training Dataset.

SB 53, the Transparency in Frontier Artificial Intelligence Act, signed September 29, 2025, imposes safety and transparency duties on the largest frontier model developers. It will not change your model releases. It matters because it confirms the trajectory: California is regulating the whole AI supply chain, and the compliance culture it creates will reach every tool you use.

The overlap with New York and the FTC

California's replica laws, New York's rules, and the FTC's requirements are the three frameworks most likely to touch a fashion campaign involving real or synthetic people. They are compatible but not identical:

The California laws govern consent and contract enforceability for digital replicas. New York now comes at it from two directions: the Fashion Workers Act requires clear written consent for digital replicas of models, and the synthetic performer disclosure law, effective June 9, 2026, requires conspicuous disclosure when ads shown to New York audiences feature AI-generated synthetic performers. I break that down in my New York explainer. The FTC framework governs disclosure to consumers nationwide.

Consent under California law does not satisfy New York's disclosure duty, and neither satisfies the FTC. The practical solution: draft model agreements and publication workflows that clear all three at once, specific replica consent language for California, replica consent plus ad disclosure for New York, and consumer-facing labels that meet the federal standard. That is achievable in one document and one workflow, but it requires deliberate drafting rather than adapting a standard release. For everything the other 48 states are doing, the tracker is current.

This article is editorial analysis, not legal advice. For questions about your specific situation, consult a qualified attorney.