California's AI Disclosure Law Will Make Ads Say When the Performer Isn't Human

Starting January 1, 2027, any ad shown to Californians that prominently includes an AI-generated synthetic performer must disclose it clearly, under Senate Bill 1050, which Governor Gavin Newsom signed on September 16. The law covers digital humans and AI voices that look and sound like real people but belong to no one. California is the second state to put this AI disclosure rule into law, after New York, and its version also reaches radio spots and podcast reads.

Synthetic performers were selling products well before either state had a legal category for them. A brand could commission a warm, trustworthy spokesperson who had never been born and run her across streaming and broadcast in the same week. The advertising statutes on the books were drafted when every face in a commercial belonged to somebody, so the practice ran ahead of the vocabulary.

What SB 1050 requires

The bill, authored by Sen. Angelique Ashby of Sacramento and sponsored by SAG-AFTRA, defines a synthetic performer as "a digital figure, voice, or representation created, in whole or in part, using generative artificial intelligence." In plain terms, a realistic human who does not exist. Disclosure is triggered when that performer is prominent, meaning in the foreground demonstrating the product or carrying the message as narrator. The label must be "clear and conspicuous," which the statute defines as difficult to miss, and its wording must be substantially similar to a short line such as "no human performer is depicted."

Enforcement runs through machinery California already trusts. A violation counts as false advertising under Business and Professions Code section 17500 and can be pursued under the state's Unfair Competition Law, the statutes behind much of California's consumer class-action work. Private plaintiffs can bring those claims, though they must prove actual damages, according to an analysis by the law firm Manatt. Once a court finds an ad unlawful, the broadcasters and platforms served with the order must stop running it in California.

Two states, two versions of the same rule

New York moved first. Governor Kathy Hochul signed that state's law, S8420, in December 2025, and it took effect on June 9, 2026. SAG-AFTRA backed both bills, and both signings took place at union offices. The two laws share a purpose and differ in ways any third state will want to study.

Each difference reflects a choice about where the risk sits. New York asks whether the advertiser knew, while California asks whether the ad discloses, so "the agency built it" is unlikely to work as a defense there. The Association of National Advertisers and the 4As urged a veto, arguing that the label applies even when no consumer is misled about anything material. Santa Clara University law professor Eric Goldman has questioned whether the takedown duty fits with federal Section 230. Those are fair questions for the courts.

Rules written for a slower world

I spent part of my early career at Napster, when the rules for recorded music had been built around plastic discs and radio playlists. The music moved to a new medium faster than any contract could be redrafted. The people who came out ahead learned the new terms while they were still being written. I call that pattern rules written for a slower world: institutions governing something that moves faster than their own rule-making cycle. That cycle runs on deliberation by design, since hearings and amendments are how a democracy makes rules that hold up.

What I respect about SB 1050 is the question it chose to govern. The law leaves the rendering software and the model names alone and attaches instead to one question a consumer will still be asking in 2035: is the person talking to me a person? A rule anchored to a question that stable keeps its meaning as the tools change underneath it. The California Legislature took this up on its own initiative, with Sen. Ashby carrying it, and wrote a rule built to outlast the current generation of software.

Tools change every quarter. The question of who is speaking to you stays the same.

What this means for the people in the ad

For a working voice actor in Los Angeles, the bookings that pay the rent between bigger jobs are often a regional radio spot or a podcast read. Those are exactly the audio formats New York's law leaves out and California's law reaches. The disclosure makes the substitution visible, so a buyer choosing between a human voice and a synthetic one makes that choice in public. Duncan Crabtree-Ireland, SAG-AFTRA's National Executive Director, called the law "an important step in the broader effort to protect consumers from deceptive AI practices."

For most families, the law will arrive in the car. Some morning next year, a California parent driving to school with a podcast on will hear a short line before an ad: no human performer is depicted. A child in the back seat may ask what that means. The answer, that the friendly voice selling something was never a person, is a conversation more families are going to have, and a plain label makes it easier to start.

"Californians deserve to know when the person selling them something isn't a person at all."

Gov. Gavin Newsom, on signing SB 1050, September 16, 2026

What leaders can do before January

For enterprise marketing leaders, the practical work is an inventory: which current and planned campaigns use a synthetic presenter or an AI voice, and who signs off on the label. Because California's law has no knowledge qualifier, that inventory has to reach into agency and vendor work as well. Teams running national campaigns may find it simpler to apply California's wording everywhere. It is the same discipline leaders had to build for the EU's own AI chatbot disclosure rule: once one jurisdiction sets the stricter bar, most teams just build to that bar everywhere.

For public officials in other states, two working models now sit side by side. A third state can learn from both before drafting a line. Practitioners who have watched these tools enter real ad pipelines can help legislative staff move quickly at the front end. More on how institutions are adapting their rulebooks lives on my AI governance page.

I advise the California State University system's AI Working Group. The question I bring into that work is the one SB 1050 answered well: which part of this will still be true in 5 years, and can we write the rule around it? Take it into your next leadership meeting, whether you run a marketing team or a statehouse office. Which of your own rules was written for a slower world, and what lasting question should it be anchored to? If you are working through that, I'm easy to find.

Sources: Office of Governor Gavin Newsom, September 16, 2026 · Davis+Gilbert, September 18, 2026 · Manatt, Phelps & Phillips client alert · MediaPost, September 17, 2026 · Associated Press, June 9, 2026 · New York State Senate, S8420.

Here is what makes Alex a credible voice on this topic: He advises the California State University system's AI Working Group, work that puts him inside the policy process on questions like this one, not just commenting on it from outside.

Want to talk through what this means for your team? Book a conversation →