On 9 June 2026, New York's synthetic performer disclosure rules take effect. The statute, S.8420-A / A.8887-B, was signed by Governor Kathy Hochul on 11 December 2025 and amends New York General Business Law Section 396-b. It defines a synthetic performer as a digitally created asset created, reproduced, or modified by computer using generative AI or a software algorithm, intended to give the impression that the asset is a human performer not recognisable as any identifiable natural person. Any paid ad served to a person in New York that features such a performer must carry a label. The penalty stack starts at $1,000 and climbs to $5,000 per subsequent violation, and it lands on whoever produces or creates the ad with actual knowledge of the synthetic performer's inclusion. Platforms and media outlets are exempt.
That last detail is the one most creative directors have not yet metabolised. The platform is not the safety net here. The buyer that commissions the work, and the shop that makes it, can both find themselves inside the definition of who produced or created the ad, depending on facts and on who had actual knowledge. For any brand running a US media plan from outside the state, geography offers no cover either. If the impression hits a New Yorker, the obligation attaches.
Two carve-outs are worth holding in mind. The law exempts advertisements for expressive works such as films, television, streaming, and video games where the synthetic performer's use is consistent with the underlying work. It also exempts uses where AI is deployed solely for language translation of a human performer. Outside those lanes, treat the next forty-eight hours as the moment AI creative compliance moves from a policy memo to an operational discipline.
The law lands on creative operations, not legal
The instinct inside most holding companies will be to route this to general counsel, get a one-pager out to clients, and move on. That misreads where the work actually sits. A disclosure obligation is not satisfied by a contract clause. It is satisfied by a label, applied to the right asset, in the right channel, every time the asset runs.
That means the spine of compliance here is the asset itself. Was it generated by a model, in part or in whole. Which model. Which prompt. Which version of the asset is live in which placement. Who signed off, and what did they know when they signed. None of that is a legal question. All of it is a creative-ops question.
Brands that have spent the last three years building proper digital asset management, with metadata fields that travel with the file from generation through trafficking, will find the new regime mostly procedural. They tag at the point of creation, the tag persists through resizing and reformatting, the trafficking system reads the tag and either renders the disclosure or refuses to push the asset live. Brands that still treat asset management as a folder structure on a shared drive will be writing cheques.
The five thousand dollar gap is a metadata gap
The escalation from $1,000 to $5,000 per subsequent violation matters more than the headline number suggests. The first violation is a slap. The second is a signal that the system that produced the first violation is still producing them. Regulators tend to read the second violation as evidence of structural failure, and they price it accordingly.
A single AI-generated hero asset can be reformatted into dozens of placements across social, display, video, and connected TV. If the disclosure obligation is enforced manually, every reformat is a chance to drop the label. Every dropped label is a fresh violation. The fines compound in a way that looks linear from the outside and exponential from the inside.
The fix is unglamorous. Every generative output needs a provenance record from the moment it is created. That record needs to be machine readable, it needs to ride with the file through every downstream system, and it needs to be the trigger for the disclosure render. It is also what establishes, or rebuts, the actual knowledge question if a regulator comes asking. This is not a creative problem. It is a pipeline problem, and the work to solve it is the same work that makes generative production economically viable at scale in the first place.
Brand and agency now share a definition problem
The other detail that will reshape conversations between brand and agency this month is the contour of the liability. The statute attaches to whoever produces or creates the ad and has actual knowledge of the synthetic performer's inclusion. That phrase can capture the brand, the agency, or both, depending on how the work was scoped, who held the prompt, and who approved the asset.
That changes the procurement conversation. A brand evaluating an agency for AI-heavy production is no longer just asking whether the work is good and the rate is right. It is asking whether the agency has the operational maturity to keep the brand out of a regulator's letter, and whether the chain of custody around generative outputs is clean enough to answer the actual knowledge question without ambiguity. Agencies that can demonstrate that chain will price the capability into their fees. Agencies that cannot will quietly lose pitches without ever being told why.
Inside the brand, the same logic applies to in-house studios. The case for bringing generative production in-house has rested on speed and cost. From 9 June it also rests on control. An in-house team that owns the generation, the tagging, and the trafficking has fewer hand-offs where a disclosure can be lost and fewer ambiguities about who knew what. That is a real argument for insourcing, and it will accelerate a trend that was already underway.
Disclosure is a creative constraint, not just a legal one
The temptation will be to treat the label as a sticker. Smallest legible type, bottom corner, satisfy the rule, move on. That underestimates how disclosure interacts with creative effect.
A label on a synthetic performer changes how the viewer reads the performer. It collapses a particular kind of suspension of disbelief. For some categories that does not matter. For luxury, beauty, and any campaign that relies on aspirational identification with a human face, it matters a great deal. The creative brief has to absorb the disclosure as part of the composition, not bolt it on at the end. The brands that work this out first will produce AI-led work that holds up under the label. The brands that do not will quietly retreat to traditional production for their hero assets and use generative tools only for the supporting cast.
That retreat is not a failure of the technology. It is a reasonable response to a constraint, and it will sharpen the question of where generative production actually earns its keep. The answer is increasingly going to be in volume, variation, and localisation, rather than in the single hero spot. That is a healthier place for generative tools to sit anyway.
What to watch through Q3
Three things will tell us how this plays out.
The first is enforcement posture. New York could ease in with warnings or it could make an example of an early violator. Either signal will travel fast through brand legal teams and reset budgets for compliance tooling. How regulators interpret actual knowledge in their first cases will matter as much as the fine itself.
The second is contagion. Other states have been active on synthetic media and deepfake legislation, and if New York's rule survives its first ninety days without a constitutional challenge that sticks, expect comparable advertising statutes to follow, each defining synthetic performer slightly differently. The operational answer to a patchwork of definitions is a single internal standard set to the strictest one and applied everywhere.
The third is what the platforms do. Meta, Google, and The Trade Desk are not on the hook for the disclosure under New York's rule, but they have every commercial reason to build tooling that helps their buyers comply. A platform that offers automated disclosure rendering as part of its ad server becomes meaningfully stickier. A platform that does not will leak buyers to one that does.
The through line on all three is the same. Disclosure is becoming an operational layer of the ad stack, and the brands that wire it cleanly into their asset pipeline now will spend the rest of the year shipping work. The brands that treat 9 June as a legal memo and not a production redesign will spend the rest of the year writing cheques to Albany.