New York's synthetic performer law requires a business that produces an ad with an AI-generated person to disclose it conspicuously, once it knows the person is synthetic. That rule, General Business Law section 396-b, has applied since 9 June 2026. California adds its own on 1 January 2027, with model wording and audio ads in scope.
For a brand running AI presenter ads to US audiences, compliance comes down to a line in the ad, a named person who adds it, and a record that it ran. Start with the line itself. Read what follows as guidance from a studio that makes these ads. It is not legal advice; have counsel confirm how each rule applies to a specific campaign.
What does New York's section 396-b require?
The duty sits in subdivision 3 of General Business Law section 396-b. It applies to "any person engaged in the business of dealing in any property or service who for any commercial purpose produces or creates an advertisement". That person "shall conspicuously disclose in such advertisement that a synthetic performer is in such advertisement, where such person has actual knowledge".
The law came from Senate bill S8420-A, signed on 11 December 2025 as Chapter 617. Its own text sets the start: "the one hundred eightieth day after it shall have become a law". Count 180 days from 11 December 2025 and you land on 9 June 2026.
A synthetic performer, in the statute's words, is "a digitally created asset created, reproduced, or modified by computer, using generative artificial intelligence or a software algorithm". It must give "the impression that the asset is engaging in an audiovisual and/or visual performance of a human performer who is not recognizable as any identifiable natural performer". That is the invented presenter: someone who looks human and is nobody in particular.
Two details decide most cases. The definition says "visual" as well as "audiovisual", so check still image ads with an AI person alongside video. And the knowledge condition protects a business that did not know; a brand that briefed, approved or paid for an AI presenter knows.
The penalty is fixed in the text. A first violation carries a civil penalty of USD 1,000. Every later one costs USD 5,000.
Which ads fall outside New York's rule?
Section 396-b lists its own exemptions. Each one is narrow:
- Audio ads. Subdivision 7 exempts "audio advertisements". A podcast spot with a synthetic voice and no picture is outside the New York duty.
- Translation only. Also exempt are ads "where the use of artificial intelligence solely involves the language translation of a human performer". A real presenter dubbed into Spanish with AI stays outside the rule; an AI presenter speaking Spanish does not.
- Expressive works. Ads for a film, TV or streaming programme, documentary, video game or similar audiovisual work are exempt "provided that the use of a synthetic performer in the advertisement or promotional material is consistent with its use in the expressive work".
- The media. "Nothing in this section shall apply to any medium used for advertising." The channel that carries the ad has no duty under the section.
An amending bill is pending. Assembly bill A11683, introduced on 2 September 2026, would require the disclosure on screen "for the entire duration" the synthetic performer appears. It would also delete the words "not recognizable as any identifiable natural performer" from the definition and exclude only digital replicas authorised by the right holder or by law, which would bring unauthorised replicas of real people under the disclosure rule. It sits in committee and is not law. If you want one creative standard that already meets the stricter version, keep the line on screen for as long as the AI person is visible.
California's SB 1050 gives model wording and covers audio from 1 January 2027
California SB 1050 was approved by the Governor on 16 September 2026 as Chapter 246 and adds section 17610 to the Business and Professions Code. It carries no urgency clause, so the California Constitution sets the date: a statute from a regular session takes effect "on January 1 next following a 90-day period from the date of enactment". Ninety days from 16 September ends in mid-December. The start is 1 January 2027.
The section targets "an advertisement that prominently includes a synthetic performer without a clear and conspicuous disclosure that the advertisement includes a synthetic performer". The definitions carry the detail:
- Synthetic performer: an asset "created in whole or in part using generative artificial intelligence that creates the realistic impression of the audio, audiovisual, or visual performance of a human performer".
- Advertisement: "any audio, video, or audiovisual message". Audio is in scope, the widest gap between the two states.
- Prominently: includes a performer "in the foreground and demonstrating or illustrating the product or service", as well as narration and reacting to the ad's message.
- Clear and conspicuous: "difficult to miss, easily understandable, and presented in a manner that a reasonable consumer would notice, read, and comprehend".
California also sets the words. The disclosure uses "wording substantially similar to "this performance features a synthetic performer" or "no human performer is depicted."" Pick one of those lines and use it in both states. New York sets no wording, and the California line says exactly what New York asks the ad to disclose.
The "prominently" test reads like a description of AI UGC ads: a generated presenter in the foreground, holding the product and talking to camera. A background figure in a lifestyle scene is a harder call. Disclose it too.
Enforcement runs through false advertising law. SB 1050 makes a breach "a violation of Section 17500", enforceable under the unfair competition chapter that begins at section 17200. Section 17500 treats false advertising as a misdemeanor, punishable by up to six months in county jail, a fine of up to USD 2,500, or both. Section 17536 adds a civil penalty of up to USD 2,500 for each violation. California exempts expressive works and AI that "solely involves the language translation of a human performer or other accessibility features", and platforms must take an ad down once a court has found it in breach.
A replica of a real person needs consent, whatever the label says
New York's definition covers only people "not recognizable as any identifiable natural performer". A digital twin of a real creator, employee or actor falls under older consent and contract rules in both states, and section 396-b states that it leaves New York's publicity rights in Civil Rights Law sections 50, 50-f and 51 in place. If you brief AI avatars and spokespeople modelled on a real employee or creator, this is the category that applies.
In New York:
- Civil Rights Law section 50 bars using "the name, portrait, picture, likeness, or voice of any living person" for advertising "without having first obtained the written consent".
- Section 50-f requires prior consent from the rights holder before a deceased personality, meaning someone domiciled in New York at death, appears in advertising. Chapter 616, signed on 11 December 2025, tightened its rule for a deceased performer's digital replica: use "with knowledge" that the replica was not authorised, in an audiovisual work, sound recording or live musical performance, now needs that prior consent, and a disclaimer no longer avoids liability. Damages are "the greater of two thousand dollars or the compensatory damages suffered".
- General Obligations Law section 5-302 makes a contract clause for a living performer's digital replica "void and unenforceable" when it lacks a specific description of the uses and the performer had no lawyer or union representation. It covers contracts from 1 January 2025.
In California, AB 2602, now Labor Code section 927, makes vague replica clauses unenforceable for new performances fixed on or after 1 January 2025 unless the performer had a lawyer or union. AB 1836, codified in Civil Code section 3344.1, requires consent for a digital replica of a deceased personality in audiovisual works or sound recordings, with damages of "the greater of ten thousand dollars ($10,000) or the actual damages".
A disclosure line does not cure missing consent. Before a replica is generated, put the signed consent, the description of uses, the markets and the end date on file. The usage rights framework for AI ads sets out how to record those terms per asset.
