No federal likeness right, fifty state regimes
The US has no federal right of publicity, so likeness exposure follows the subject and the audience. How to set a governing-law posture that actually holds.
There is no American likeness law. There are roughly fifty of them, they disagree, and which one decides your case is not determined by where your company is.
This is the single most common structural mistake in AI content planning. Teams internalise copyright — federal, uniform, one statute for the whole country — and then reason about likeness the same way. Likeness does not work like that. The United States has no federal right of publicity. What exists is a patchwork of state statutes and common-law doctrines that vary in what they protect, how long protection lasts, whether it survives death, and how hard the First Amendment pushes back on expressive uses.
This is general information, not legal advice. The point of this piece is to give you a posture, not an answer.
What "no federal right" actually means in practice
It means three things that change how you plan.
There is no single standard to comply with. You cannot be "compliant with US likeness law" because there is no such thing to be compliant with. The best you can do is identify which regimes plausibly reach your content and satisfy the strictest of them.
The differences are not cosmetic. Tennessee's ELVIS Act, effective 1 July 2024, made voice a protected personal right and reaches tool providers, not only publishers. California's AB 2602 and AB 1836, both effective 1 January 2025, attack the contract side and the deceased-personality side respectively. Those are structurally different instruments, not local variants of one rule. A process built only around one of them will miss what the other covers.
The gaps get filled unevenly and fast. New state laws land every session, and they do not converge on a template.
A federal right has been proposed. The NO FAKES Act advanced out of the Senate Judiciary Committee on 18 June 2026 as S.4591. It is a bill. Plan on the patchwork, and treat federal preemption as a thing that might happen later. Our piece on digital replica rights turning into property covers what would change if it passes.
Federal law does touch synthetic media — just not as a publicity right
Worth separating, because "no federal right of publicity" gets misheard as "nothing federal applies."
- The TAKE IT DOWN Act (Pub. L. 119-12) criminalises non-consensual intimate imagery including AI-generated material, and requires platforms to operate notice-and-removal within 48 hours. FTC enforcement of the platform duties began 19 May 2026.
- FTC Section 5 deception and unfairness authority applies to advertising with no AI-specific rule needed. The Rule on the Use of Consumer Reviews and Testimonials, effective 21 October 2024, prohibits AI-generated fake reviews and fabricated testimonials outright.
- The Government and Business Impersonation Rule, effective April 2024, covers impersonating firms and agencies. A supplemental rulemaking to extend it to individuals appears not to have been finalised — do not rely on it either way without checking current status.
None of those give a person a property-style right in their own face. They constrain specific harms.
The three location questions
Replace "what does US law say" with these, in this order.
1. Where does the subject sit? The person being depicted anchors the publicity analysis. Their domicile — or, for a deceased personality, where the estate sits and where they were domiciled at death — is the first regime to check. This is the question most teams skip entirely.
2. Where does the audience sit? Distribution creates its own hooks, and increasingly these are disclosure duties rather than likeness duties. New York's synthetic performer disclosure took effect 9 June 2026 and turns on ads reaching New York consumers, at $1,000 for a first violation and $5,000 thereafter, with an exemption for promotion of expressive works. EU AI Act Article 50 deployer obligations apply from 2 August 2026 to content put in front of EU viewers. Neither cares where you are incorporated.
3. Where do you sit? Last, not first. Your own jurisdiction matters for enforcement practicality and for your own contracts, but it does not shrink the set of regimes that can reach content you distributed elsewhere.
The anchors worth checking first
Not exhaustive. This is the shortlist that covers most real productions.
| Anchor | What it turns on | Live date |
|---|---|---|
| Tennessee ELVIS Act | Voice as a protected personal right; tool providers actionable | 1 Jul 2024 |
| California AB 2602 | Validity of digital-replica clauses in performer contracts | 1 Jan 2025 |
| California AB 1836 | Estate consent for replicas of deceased personalities | 1 Jan 2025 |
| New York synthetic performer disclosure | Ads reaching NY consumers featuring a fully synthetic human performer | 9 Jun 2026 |
| EU AI Act Art. 50 | Deployer labelling of deepfakes on first exposure | 2 Aug 2026 |
| California AI Transparency Act (SB 942, as amended by AB 853) | Latent provenance marking by covered generative AI providers | 2 Aug 2026 |
| Colorado SB 189 | Narrowed disclosure and transparency regime, replacing the repealed Colorado AI Act | 1 Jan 2027 |
Two things to note about that table. Colorado's original AI Act was repealed and replaced by SB 189, signed 14 May 2026 — a lot of secondhand writing still cites the repealed version, so check what you are reading against the current text. And outside the US, Denmark's likeness amendment is drafted to cover every person rather than only public figures, though we could not resolve its in-force date against a primary source; the Denmark write-up sets out that caveat honestly.
The posture that survives
You cannot run a fifty-state analysis per asset. What works is a default plus an escalation trigger.
Default: build to the strictest regime you plausibly touch. Written consent for any real person's face or voice, scoped and time-limited, regardless of what the most permissive applicable law would allow. Voice cleared separately from likeness. Disclosure applied by default rather than by exception. That default is not expensive, and it makes the jurisdictional question mostly moot.
Escalate when any of these is true:
- The subject is a named public figure, living or dead.
- The subject did not participate in creating the material.
- The content will run as paid media in a jurisdiction you have not checked.
- Voice is involved and the depiction is recognisable.
- The output is a replica of a specific person rather than a synthetic individual.
The last one is the sharpest filter. A fully synthetic person triggers disclosure duties but not publicity claims, because there is no rights holder. A replica of a real person triggers both. Keeping those two categories distinct in your brief, your file naming and your review step removes most of the ambiguity before it reaches counsel.
Choice-of-law clauses do less than people assume
A governing-law clause allocates risk between the parties who signed it. It does not bind a third party whose likeness was used and who never signed anything, and validity rules of the kind California enacted are frequently applied regardless of what a contract nominates.
So a contract is the right tool for the counterparty relationship — the performer, the agency, the client — and the wrong tool for regulatory exposure. Both need handling and they are not the same handling. Legal and licensing for AI content in business covers the contract half, and the cross-platform labelling checklist covers the disclosure half.
Making it operational
The record is what turns a posture into a defence. For every asset depicting a real person, keep the consent document, its scope and term, the model used, the date, and the jurisdictions the piece ran in, stored with the file rather than in someone's inbox. Usage rights is the vocabulary; agencies producing at volume will find the process version in our guide for agencies, and the ad-side disclosure obligations in AI ad disclosure compliance.
FAQ
If my company is in a state with weak publicity law, am I safer?
No. Your own state's law is the least relevant of the three location questions. Exposure follows the subject and, for disclosure duties, the audience. Incorporating somewhere permissive does not shrink either set.
Does the First Amendment protect parody and commentary?
There is genuine and unresolved tension between publicity claims and expressive uses, and courts apply different tests to it. That tension is a real defence in real cases — it is not a category you can self-certify into before publishing. Commercial advertising sits at the weakest end of it.
Would the NO FAKES Act replace all of this?
It would create a federal right, but whether and how far it preempts state regimes is a question about the enacted text, which does not exist yet. Until it passes, the patchwork is the whole picture. Our voice and likeness law overview tracks where the pieces currently sit.
How do I handle content that runs everywhere at once?
Assume the strictest regime in your distribution set and build to it. For most teams that means written consent plus disclosure as the default, which is cheaper than segmenting creative by territory and far cheaper than getting it wrong in one market.