Synthetic performer disclosure is now California law. On 16 September 2026, at the Los Angeles headquarters of SAG-AFTRA, Governor Gavin Newsom signed Senate Bill 1050, which makes it an unlawful advertising practice to run an ad that prominently features an AI-generated performer without telling the audience. The rule takes effect on 1 January 2027, and it applies to audio as well as video.

The bill is short, and that is the point. It adds a single section — Article 10, commencing with Section 17610, of the Business and Professions Code — that defines a synthetic performer, says when one is “prominent” enough to trigger a label, prescribes the wording of the label, and then plugs the whole thing into California’s existing false-advertising machinery rather than inventing a new penalty regime.

This is the disclosure side of a fight the entertainment industry has been having in public all year — we covered the film summit AI declaration and TikTok’s AI likeness detection tool earlier. California is the second state to do this. New York’s law, signed on 11 December 2025 and effective 9 June 2026, got there first. But the two statutes are not copies of each other, and the differences matter for anyone buying media in both states.

This article works through the statutory text, the three-pronged prominence test, the five carve-outs, the takedown duty it places on broadcasters and platforms, and the practical compliance surface for advertisers who now face two overlapping state rules and a January deadline.

What the Synthetic Performer Law Actually Says

synthetic performer disclosure california sb 1050 ads b box with one round hole through its front face

The operative sentence is subdivision (b) of Section 17610: “It is unlawful for any person to create and cause to be published in an advertising medium an advertisement that prominently includes a synthetic performer without a clear and conspicuous disclosure that the advertisement includes a synthetic performer.”

Three things in that sentence do the work. The obligation falls on the person who creates and causes to be published — the advertiser and its agency, not the broadcaster. The trigger is prominently, which is defined. And the remedy is a disclosure, not a prohibition: nothing in the law stops you from using a synthetic performer at all.

The bill took 216 days to travel from introduction on 12 February 2026 to signature on 16 September, and a further 107 days pass before it bites.

SB 1050 milestones, in days from introduction on 12 February 2026
Amended in Senate, 23 Apr 70 days
Amended in Assembly, 11 Jun 119 days
Passed both houses, 31 Aug 200 days
Signed, 16 Sep 216 days
Operative, 1 Jan 2027 323 days
Bars are each milestone as a share of the 323 days from introduction to the operative date.

The definition is narrower than “AI in advertising”

A synthetic performer is “a digital figure, voice, or representation 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 who is not recognizable as any identifiable natural person.”

Every clause narrows it. It must be created at least partly with generative AI. It must create a realistic impression of a human performance — a stylised cartoon spokesanimal is not a synthetic performer. And it must not be recognisable as a real person, because a recognisable real person is already covered by California’s digital replica and right-of-publicity statutes.

“In whole or in part” is the broad edge

The one phrase that widens the definition is “in whole or in part.” A human-shot performance that has been substantially regenerated, a real actor’s voice replaced by a generated one, or a composite built from generated and captured elements can all land inside the definition even though a person was on set. The question is whether the finished asset creates the realistic impression of a human performance that is not any identifiable person.

Voice alone counts

California covers “a digital figure, voice, or representation.” A radio spot or a podcast read voiced entirely by a generated voice is squarely within the statute. This is the single largest practical divergence from New York, and it is discussed in detail below.

The Legislature wrote its own First Amendment defence

Section 1 of the bill contains findings that read as a pre-emptive answer to a commercial-speech challenge: consumers “have a substantial interest in knowing when an advertisement uses a synthetic performer in a manner that could reasonably be interpreted as a human performance,” and the required disclosure “is a purely factual commercial disclosure consistent with constitutional protections for commercial speech.” That is the Zauderer framing, stated on the face of the bill.

When a Synthetic Performer Is "Prominent" Enough to Trigger a Label

synthetic performer disclosure california sb 1050 ads c lantern box with one ring handle

Not every generated element in an ad requires a disclosure. The statute defines “prominently” with three alternative prongs, and meeting any one of them is enough.

ProngStatutory wordingWhat it captures in practice
(A) DemonstrationIn the foreground and demonstrating or illustrating the product or serviceThe generated person holding, using or wearing the thing being sold
(B) NarrationProviding or voicing the on- or off-camera narration or commercial messageGenerated voiceover, generated presenter, generated audio-only read
(C) ReactionIllustrating or reacting to the on- or off-camera narration or commercial messageThe generated “customer” nodding, smiling or responding to a human voiceover

Prong C is the one teams will miss

The first two prongs are intuitive. The third is not. A synthetic performer who never speaks, never touches the product, and simply reacts on camera to a human narrator is still prominent under (C). A great deal of generated stock footage — the smiling office worker, the relieved homeowner, the delighted diner — exists precisely to illustrate a voiceover, and that is the conduct the third prong names.

Background crowds are outside it

Conversely, a synthetic performer in the background of a street scene meets none of the three prongs. The statute’s construction clause confirms this by stating that nothing in the section restricts the creation, distribution or exhibition of synthetic content generally. The label attaches to a role, not to a technology.

There is no “actual knowledge” escape hatch

New York’s statute applies only where the advertiser has actual knowledge that a synthetic performer appears. California’s does not contain that qualifier. An advertiser that licenses a generated asset without asking, or accepts an agency deliverable without a provenance warranty, is exposed in California in a way it would not be in New York. That single omission is why contract language, not creative review, is the real compliance lever.

The Exact Disclosure Wording the Statute Prescribes

synthetic performer disclosure california sb 1050 ads d crate box with one raised band

Subdivision (c) is unusually prescriptive for a disclosure statute: the disclosure “shall use wording substantially similar to ‘this performance features a synthetic performer’ or ‘no human performer is depicted.'”

Two safe harbours, not a format spec

The two quoted phrases function as safe harbours. “Substantially similar” leaves room for house style, but it does not leave room for euphemism. “Made with AI,” “digitally enhanced,” or “created using advanced technology” are not substantially similar to either model phrase, because neither of them tells the viewer that the person is not a person.

“Clear and conspicuous” is defined, and it is a reasonable-consumer test

The statute defines a clear and conspicuous disclosure as one “that is difficult to miss, easily understandable, and presented in a manner that a reasonable consumer would notice, read, and comprehend, taking into account the medium, format, and context in which the advertisement appears.”

There is no font size, no duration and no placement rule. Instead, the standard flexes with the medium — which means a two-frame super at the end of a fifteen-second vertical video is judged against what a reasonable consumer scrolling a phone would actually notice, not against a checklist.

Audio needs an audio disclosure

Because the definition is medium-sensitive and California covers voice, a purely visual disclosure on an audio-only ad cannot work: there is nothing to see. A spoken tag is the only realistic route for radio, streaming audio and podcast reads, and it consumes inventory in a format where every second is priced.

The disclosure is factual, not a warning

Nothing in the statute requires the disclosure to be pejorative, nor to explain how the asset was made. It is a statement of fact about the performance. Teams that draft it as a legal disclaimer will produce something longer and less conspicuous than the statute’s own model phrases.

Who Has to Take a Synthetic Performer Ad Down

synthetic performer disclosure california sb 1050 ads e whistle tube with a round chamber

The most operationally significant part of the law is subdivision (e), which reaches past the advertiser to the broadcaster, the platform and the ad network.

The duty is court-order-gated

An advertising medium must stop carrying an ad containing a synthetic performer only when two conditions are both met: a court of competent jurisdiction has issued an order finding the ad violates the chapter or enjoining the creator from publishing it — including an injunction under Code of Civil Procedure section 527 — and the medium has been served with the order plus “information reasonably sufficient to identify the advertisement.”

What the medium must then do

On receipt, the advertising medium must, “as soon as commercially reasonable and technically feasible,” remove, disable access to or cease further dissemination of the ad in California, and stop accepting payment for further dissemination. Both duties, not one: the money has to stop as well as the pixels.

Who counts as an advertising medium

The definition is broad — broadcast stations, cable operators, multichannel video programming distributors, online platforms, streaming services, digital advertising networks and publishers. It carves out only entities “not capable of removing, disabling access to, or ceasing further dissemination of, an advertisement,” which is a narrow technical exception rather than a general intermediary shield.

This is not a notice-and-takedown regime

No private complaint, regulator letter or union demand triggers the duty. Without a served court order, an advertising medium has no statutory obligation under Section 17610 at all. That design keeps platforms out of the business of adjudicating whether a given face is synthetic, and it puts the burden of proof where the litigation is.

The Five Carve-Outs Every Synthetic Performer Review Should Check

synthetic performer disclosure california sb 1050 ads f periscope tube with a box on top

Subdivision (d) lists five things the section must not be construed to do. Two of them are substantive exemptions that will decide most borderline cases.

Construction clauseEffect
(d)(1) Synthetic content generallyDoes not restrict creating, distributing or exhibiting synthetic content
(d)(2) Editorial contentDoes not regulate the expressive or informational content of an ad, only the factual disclosure
(d)(3) Other remediesDoes not affect rights or remedies under Sections 17200, 17500 or any other deception law
(d)(4) Expressive worksDoes not apply to ads for films, TV, streaming content, documentaries or video games, if the use in the ad matches the use in the work
(d)(5) Translation and accessibilityDoes not apply where generative AI is used solely for language translation of a human performer or other accessibility features

The expressive-works carve-out has a condition attached

A trailer for a film that contains a synthetic performer is exempt — but only “provided that the use of a synthetic performer in the advertisement or promotional material is consistent with its use in the expressive work.” Generate a character for the trailer who does not appear in the film, and the exemption stops applying. That consistency test is the part worth writing into marketing sign-off.

Translation and dubbing are explicitly protected

Using generative AI solely to translate a human performer’s delivery into another language, or to add accessibility features, does not create a synthetic performer. The word doing the work is “solely.” An AI dub that also regenerates the performer’s face or replaces the voice with a wholly generated one is no longer solely translation.

Nothing here displaces existing false-advertising law

Clause (d)(3) is a reminder rather than an exemption: an ad that discloses its synthetic performer correctly can still be deceptive for entirely different reasons, and Sections 17200 and 17500 remain fully available.

How California and New York Differ on Synthetic Performer Disclosure

Both states now require a disclosure. Anyone running national campaigns has to satisfy the stricter rule on each axis, because the two statutes are strict on different things.

FactorCalifornia SB 1050New York GBL 396-b
Signed16 September 202611 December 2025
Effective1 January 20279 June 2026
Audio-only adsCoveredExcluded
Non-AI synthetic humansOutside it — requires generative AIInside it — covers other software-created assets
Knowledge requirementNone statedActual knowledge
Prescribed wordingTwo model phrases in the statute“Conspicuous disclosure”, undefined
Stated penaltyNone — routes into Sections 17500 and 17200$1,000 first, $5,000 each subsequent
Duty on the carrierCourt-ordered takedown plus stop paymentDisseminator protection

The composite rule is stricter than either statute

Run the two side by side and a single national creative standard emerges: disclose on audio as well as video, do not rely on not knowing, use California’s model wording, and keep the disclosure conspicuous in the format it ships in. That composite satisfies both states, and it is simpler to operate than a per-state matrix.

New York’s penalty arithmetic scales badly

New York’s figures are per violation, and a programmatic campaign generates a great many placements. Taking the statute’s own numbers — $1,000 for a first violation, $5,000 for each subsequent one — the exposure curve is steep.

New York statutory exposure by number of non-compliant placements
10 placements $46,000
50 placements $246,000
100 placements $496,000
250 placements $1,246,000
Arithmetic: one placement at $1,000 plus the remainder at $5,000 each.

California’s open-ended remedy may bite harder

California names no figure, which is not the same as naming a small one. A violation of Section 17610 is a violation of Section 17500 — a misdemeanour — and is enforceable through the Unfair Competition Law in Section 17200, which brings injunctive relief, restitution and civil penalties assessed per violation in an action brought by public prosecutors. An injunction that pulls a flight mid-campaign is usually more expensive than any penalty schedule.

Where the Synthetic Performer Rule Sits in California's AI Stack

SB 1050 is the fourth layer of a stack California has been building since 2024, and reading it in isolation overstates how much is new.

LawWhat it governsOperative
AB 2602Contract consent for digital replicas of real performers1 January 2025
AB 1836Digital replicas of deceased performers1 January 2025
SB 942 as amended by AB 853Provenance and detection tooling for generative AI providers2 August 2026
SB 1050Synthetic performer disclosure in advertising1 January 2027

The replica laws and this one cover opposite cases

AB 2602 governs digital replicas of identifiable people and works through consent in contracts. Section 17610 governs generated people who are not identifiable and works through disclosure to the audience. Between them there is very little gap: if the face is recognisable you need consent, and if it is not you need a label.

Provenance tooling makes the label auditable

California’s AI Transparency Act, as amended, requires covered generative AI providers to attach provenance signals to what their systems produce. That metadata is what turns “we did not know it was generated” into a checkable claim — and it is one reason California felt able to omit New York’s actual-knowledge qualifier.

Audit requirements are arriving alongside

Newsom has also signed measures establishing independent verification and third-party audit expectations for AI systems. None of that governs advertising directly, but it shapes the evidentiary environment in which a synthetic performer dispute would be litigated.

What Advertisers Should Do Before 1 January 2027

The deadline is a little over three months away, and most of the work is procurement and paperwork rather than creative.

Add a provenance warranty to every content contract

Because California omits the actual-knowledge qualifier, “the agency did not tell us” is not a defence. Every production, stock-footage, voice and post-production agreement needs a warranty stating whether generative AI was used to create or substantially alter any human performance, with an indemnity attached.

Build a synthetic performer register

Maintain a per-asset record: was generative AI used, on which element, does the result read as a realistic human performance, is the person identifiable, and which prominence prong applies. That register is the artefact an enforcement action will ask for, and it is far cheaper to keep contemporaneously than to reconstruct.

Decide the disclosure treatment per format, now

A disclosure that is conspicuous in a thirty-second TV spot may not be in a six-second bumper or a nine-second audio pre-roll. Settle the treatment for each format your media plan actually buys — supered text, spoken tag, persistent caption — and get it into the creative template before briefs go out.

Do not let translation quietly become replacement

Localisation pipelines drift. A workflow that started as AI translation of a human read can become a fully generated voice after a vendor change, and the (d)(5) exemption evaporates with it. Re-confirm the exemption at each localisation cycle rather than at the original sign-off.

Treat the expressive-works exemption as conditional

Entertainment marketers relying on (d)(4) should confirm that the synthetic performer in the trailer is used consistently with the finished work. Where a promo uses generated material that never appears in the film, plan for a disclosure.

Align on one national standard

Two states with different rules is an argument for one internal standard that clears both, not for a state-by-state creative matrix. For most brands, adopting California’s model wording across all audio and video assets in the United States is cheaper than maintaining the split.

What the Synthetic Performer Debate Is Really About

The political framing at the signing was about work, not consumer confusion, and the two are not the same argument.

The union case

SAG-AFTRA advocated for the bill and hosted the signing. National Executive Director Duncan Crabtree-Ireland described it as an important step in protecting consumers from deceptive AI practices and ensuring AI is not used at the expense of human creativity. A disclosure requirement does not stop an advertiser from replacing a human performer, but it does price the choice: audiences who dislike generated presenters can now identify them.

The consumer case

Newsom’s framing was simpler — “Californians deserve to know when the person selling them something isn’t a person at all.” Senator Angelique Ashby, the bill’s author, put it as requiring “that synthetic figures be labeled as such, thus protecting consumers from false advertising.”

The legal case is deliberately modest

The bill’s findings claim only that the disclosure is purely factual and commercial. That modesty is strategic. A statute that compelled an opinion, or that banned the practice, would face a much harder constitutional test than one that requires an advertiser to state a fact about its own ad.

Expect more states, not fewer

Two states in ten months, with a third of the American advertising market between them, is the pattern that usually precedes either a wave of copies or a federal pre-emption fight. The composite compliance standard described above is the hedge against both.

Frequently Asked Questions About Synthetic Performer Disclosure

Does the synthetic performer rule apply to an ad made outside California?

The statute reaches advertisements made available to consumers in California, and the advertising medium definition is framed around distribution “to consumers in this state.” Where the ad was produced is not the test.

Is a generated voice alone enough to require a disclosure?

Yes. California’s definition covers “a digital figure, voice, or representation,” and the narration prong of the prominence test is met by voicing the commercial message. This is the main respect in which California is broader than New York.

What if only part of the performance was generated?

The definition says “created in whole or in part using generative artificial intelligence.” Partial generation can still produce a synthetic performer if the finished asset creates the realistic impression of a human performance by no identifiable person.

Does an influencer video need the disclosure?

If the creator appears as themselves, no — they are an identifiable natural person. If the video features a generated presenter that meets one of the prominence prongs, the requirement applies like any other ad.

Who can sue over a missing synthetic performer disclosure?

The section routes enforcement into Sections 17500 and 17200 rather than creating a private right of action of its own, and subdivision (e) contemplates a court order obtained by someone with standing under those provisions.

When does a platform have to pull the ad?

Only after it has been served with a court order finding a violation or enjoining the creator, together with enough information to identify the ad. Then it must act as soon as commercially reasonable and technically feasible, and stop taking payment.

Teams working through this alongside a wider AI governance programme may find our AI strategy service useful, and our AI models and tools hub tracks the generative systems most likely to produce the assets this law now covers.

References