July 2026
The Death of the Shortcut
New York's disclosure law is live. RSL Media's consent registry is open. Here is what each one actually says, what neither of them does, and what it means for the work.
For the last few years, the industry sold brands a seductive premise. Move fast, cut the budget, generate at scale. The premise worked because the first wave of AI tools leaned on a shortcut: real faces, names, and likenesses taken off the internet without asking, reducing people to invisible, uncompensated training inputs.
The technology sprinted. The ethical and legal architecture to hold it up never caught up.
In the last eight months, two things landed that start to close the gap. Neither was designed with the other in mind. Read together, they say something useful about where this is going.
What New York actually did
New York's Synthetic Advertisement Disclosure law took effect on June 9, the first of its kind in the country. If an advertisement includes a synthetic performer, meaning a digitally created asset built to read as a human performance by someone not recognizable as any identifiable natural performer, whoever produced that ad has to say so, conspicuously. First violation is $1,000. Every one after is $5,000.
It is narrower than the headlines suggested. It exempts advertising for expressive works like films and games. It exempts audio-only ads and AI used purely for translation. It does not apply to the outlets that carry the ad, only to whoever made it, and it only bites when that person actually knew a synthetic performer was in there. There is no private right of action.
It does not cover every AI-generated human. It covers AI-generated humans who are not anyone. If the person is recognizable as a real individual, that is a digital replica, and New York routes it to right-of-publicity law instead, which asks for consent rather than a label. The new law says so directly: it does not limit, reduce, or enlarge anyone's rights under the Civil Rights Law.
And read the definition closely, because most client alerts paraphrase it. The statute says identifiable natural performer, not natural person. Those are different words. The law is five weeks old, no court or attorney general has drawn the line, the text says the disclosure must be conspicuous and says nothing else, and no wording is a safe harbor. Anyone telling you precisely where the boundary sits is guessing.
What it does do is put a question in front of every brand running AI creative: is the person in this ad a real person or not? That used to be a philosophical question. It now has a form field attached to it.
What RSL Media actually did
In May, RSL Media launched as a nonprofit co-founded by Cate Blanchett alongside the team behind the Really Simple Licensing standard, backed by Tom Hanks, Emma Thompson, Javier Bardem, CAA, and the Music Artists Coalition. Its Human Consent Registry went live in June, free to anyone. The premise is that your identity is your intellectual property, and the controls work like a traffic light.
- Prohibited
- Allowed with terms
- Allowed
It is real infrastructure and it deserves the attention it is getting. It is also voluntary, and it carries no force of law. It works if AI companies choose to query it, or if something like the NO FAKES Act eventually gives it teeth.
And a signal only solves half the problem. It tells an AI system what it may not take. It does not tell a creative team where to go next.
A registry can tell you someone is willing. It cannot book them.
The law and the culture are saying the same thing
Put those two developments side by side. A state legislature and a group of actors arrived in the same place independently, from opposite ends, with no shared agenda and no coordination. One wrote a statute with penalties attached. The other built a free registry with no legal force at all. Neither is arguing that AI should go away.
Both landed on the same question: did the person agree?
And it is not a fringe position held by people who dislike technology. SAG-AFTRA pushed the New York bill through. CAA and the Music Artists Coalition put their names on the registry. Cate Blanchett took the case to the European Parliament. None of them are asking anyone to stop building. They are asking for the same thing an audience asks for when it finds out after the fact that the face in an ad was never a person. Tell us. And make the work differently.
Which turns this from a debate into a to-do list.
What to actually do about it
None of this requires a strategy offsite. Four things are worth doing this quarter.
- Audit what is already live. If you have AI-generated people in market and you sell into New York, the disclosure question is live now, not next quarter. E-commerce lifestyle imagery is the most commonly missed category, because it rarely goes through the same review as a campaign.
- Decide your disclosure language before someone asks for it. The statute gives you no template and no safe harbor. Pick your wording, apply it consistently across formats, and keep a record of how you decided. Consistency is most of what conspicuous means in practice.
- Ask your vendors where the faces came from. Not which model. Which people. If your agency or your tool cannot answer that question, you have your answer.
- Assume the bar moves from labeling to consent. NO FAKES is pending. States keep expanding right-of-publicity law. SAG-AFTRA's commercials contract already spells out consent, notice, and compensation for digital replicas. None of that is converging on labeling.
You can add a label on Friday.
You cannot retrofit consent.
The future of AI is human
Complying with New York costs you a line of type. That is worth doing, and it is not the interesting part.
The interesting part is the direction all of this is pointing. New York's law and RSL Media's registry are two versions of the same finding, arrived at independently by a legislature and a group of actors: consent is the defining question of this technology, not a footnote to it. Moving fast has stopped being a reason to leave real people out of the deal.
That is not only a moral argument. AI can generate anything you ask it for. What it cannot generate on its own is a reason for anyone to believe it. That has to come from outside the model, and so far the only place it comes from is a person who agreed to be there and got paid for it.
The platforms that define the next decade of media will not be the ones with the biggest appetite. They will be the ones that built permission, payment, and respect into the floor rather than the marketing.
Trueth is an AI image generation platform built on a roster of real people who chose to be there, signed a commercial agreement, and earn a royalty every time their likeness is used.
If any of this is useful, that is enough. If you want to see how it works, we are at trueth.io.