Does a fictional AI model need a model release?

A fictional AI person who replicates nobody has no one to sign a release, and New York's Fashion Workers Act consent rule addresses a real model's likeness. A release is needed if the AI model is built from a real person's photo or likeness.

Do I have to label AI models on my clothing website?

It depends on the market. New York General Business Law section 396-b requires conspicuous disclosure in advertisements, although no decision has settled whether a product page counts. EU AI Act Article 50(4) can require a visible label for deep fakes, and whether a photorealistic AI model counts is unsettled. The UK's ASA said in May 2025 there is no blanket legal duty.

Is a metadata tag enough to meet New York's disclosure law?

That is not settled. General Business Law section 396-b requires a person to “conspicuously disclose in such advertisement”, without defining conspicuous, and a viewer cannot see a metadata tag. A tag that satisfies a platform may not settle the statutory duty.

Does the EU AI Act ban AI fashion models?

No. Article 50 of the EU AI Act is a transparency rule, not a ban. A disclosure does not make an otherwise misleading ad lawful, as the Commission's Guidelines of 20 July 2026 note.

Can I legally use AI models for my clothing brand? 2026

Are AI fashion models legal to use in advertising?

AI fashion models are legal to use in advertising in the jurisdictions reviewed here, provided the brand meets the disclosure, consent and truthfulness rules below.

The table gives the position as of 31 October 2026. Other US states and other countries have their own rules, which were not reviewed. This is general information, not legal advice.

Jurisdiction	Rule	What it asks of a clothing brand	Date and status
New York	General Business Law section 396-b(3)	Conspicuous ad disclosure of a synthetic performer. Civil penalty $1,000, then $5,000.	Signed 11 December 2025; in effect 9 June 2026.
New York	Labor Law sections 1037(7) and 1038, Fashion Workers Act	Prior written consent for a real model's digital replica. Civil penalty $3,000, then $5,000.	In effect 19 June 2025.
New York	Civil Rights Law sections 50 and 51	Written consent to use a living person's likeness in advertising or trade.	Text read 31 October 2026.
United States	FTC Act section 5; FTC Deception Statement	No representation likely to mislead consumers on a material point.	Statement dated 14 October 1983.
European Union	AI Act Articles 50(2), 50(4) and 50(5); Article 111(4), added by Regulation (EU) 2026/1744	Providers mark output; deployers disclose deep fakes visibly.	Applies from 2 August 2026. Systems on the market before that date have until 2 December 2026 for marking only.
United Kingdom	CAP Code rule 3.1	Marketing “must not materially mislead or be likely to do so”.	No blanket UK duty to disclose AI (ASA, 29 May 2025).

What does New York's synthetic performer law require?

New York's synthetic performer law requires anyone dealing in property or services who produces or creates an advertisement to “conspicuously disclose in such advertisement that a synthetic performer is in such advertisement, where such person has actual knowledge”. The bill, S.8420-A, was signed on 11 December 2025 as chapter 617 and took effect on the 180th day, 9 June 2026, as General Business Law section 396-b(3).

A synthetic performer is a digitally created asset, made or modified using generative AI or a software algorithm, “intended to create 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”.

The section exempts audio advertisements, AI used solely to translate a human performer's language, and ads for expressive works such as films and games.

Several points are unsettled as of 31 October 2026. The section defines neither “advertisement” nor “conspicuously” and prescribes no wording. Subdivision 3 has no “in this state” limit, so its reach to sellers outside New York is untested. No decision was found on whether a catalogue image on a product page counts; treating it as covered is the cautious reading. Whether a metadata keyword alone is conspicuous is also open, since a viewer cannot see one. Platform tagging rules are in the 2026 rules guide.

Check whether a file carries an AI label

Do I need consent to make an AI version of a real model?

Yes. In New York, a client must “obtain clear and conspicuous prior written consent from the model for any creation or use of a model's digital replica, detailing the scope, purpose, rate of pay, and duration of such use” (Labor Law section 1037(7)). The New York Department of Labor says client duties apply from 19 June 2025.

A digital replica is “a significant, computer-generated or artificial intelligence-enhanced representation of a model's likeness”, excluding routine edits such as colour correction and minor retouching. A client is a person or entity that contracts for and manages modelling services.

A release that omits those four items does not meet that wording. The section addresses a real model's likeness, so on its wording a wholly fictional AI person who replicates nobody falls outside the definition; no decision testing that line was found.

Create a synthetic studio model

Can I be sued if my AI model looks like a celebrity or influencer?

Yes, a brand can be sued if an AI model is recognisable as a real person. New York Civil Rights Law section 51 lets a person whose “name, portrait, picture, likeness or voice” is used in the state “for advertising purposes or for the purposes of trade” without written consent seek an injunction and damages, with exemplary damages for knowing use. Section 50 makes that use a misdemeanor.

Section 396-b does not help here: a recognisable person is outside the synthetic performer definition, and the section says it does not reduce rights under sections 50, 50-f or 51. A disclosure label is not a substitute for consent.

US courts have reached beyond exact copies: in White v. Samsung (9th Circuit, 1992, California law), an ad showing a robot in a wig and gown beside a game-show board revived common-law publicity and Lanham Act claims, though not the California statutory claim.

That case involved deliberate evocation of a famous person. No decision was found on an AI model that resembles a private person or influencer by accident, so that question is unsettled. Not naming real people in a brief, not uploading their photos and checking results for an obvious match all lower the risk.

Elsewhere, the UK government said on 18 March 2026 that existing protections “do not cover all situations” and that it will consider a new digital replica or personality right.

Can an AI model image mislead buyers about the garment?

Yes, and an AI model image that misrepresents a garment is judged like any other advertising claim. In the US, section 5 of the FTC Act declares unlawful “unfair or deceptive acts or practices in or affecting commerce”. The FTC's Deception Statement of 14 October 1983 finds deception where a representation, omission or practice is “likely to mislead the consumer acting reasonably in the circumstances, to the consumer's detriment”, on a point likely to affect the buying decision.

Images count: the FTC's advertising guide says it looks at “words, phrases, and pictures”. In FTC v. Colgate-Palmolive (1965), the Supreme Court agreed that an undisclosed mock-up shown as proof of a claim was a material deceptive practice, while distinguishing a prop “not being used for additional proof of the product claim”.

No FTC rule or decision specific to AI fashion models was found as of 31 October 2026, so how these principles apply is unsettled. The practical risk is colour, drape, length and fit. Herofab's own measurement of 303 judged outputs found 4.0% showed a different product and 9.6% added invented detail. The check covered images made between 24 September and 12 October 2026, before Herofab's late-October image engine change.

FTC rule 16 CFR 465.2 separately bars a business from creating a consumer testimonial that materially misrepresents “That the reviewer or testimonialist exists”.

A label does not cure a misleading image. The UK's ASA said on 29 May 2025 that “disclosure alone is very unlikely to mitigate the harm caused by a fundamentally misleading message”.

Dress a model in your real garment photo

Does the EU AI Act apply to AI models in product images?

The EU AI Act can apply to AI model images through two duties in Article 50, which the European Commission says applies from 2 August 2026. Article 50(2) requires providers of generative systems to mark output in a machine-readable format; under Article 111(4), added by Regulation (EU) 2026/1744 of 8 July 2026, providers of systems placed on the market before 2 August 2026 have until 2 December 2026. That extension covers Article 50(2) only. Article 50(4) requires deployers to disclose deep fakes. A brand using an AI system professionally can be a deployer; the Commission's Guidelines say a company that merely commissions an agency without controlling its AI use is not.

A deep fake is image, audio or video content that “resembles existing persons, objects, places, entities or events and would falsely appear to a person to be authentic or truthful” (Article 3(60)). The Commission's Guidelines, C(2026) 5054 final of 20 July 2026, are non-binding, and their cover note says formal adoption follows once all language versions exist. They count “realistic AI-generated human avatars” as persons and say a subject that “can plausibly exist” is enough.

The Guidelines give two product examples. “An AI-generated image of a product in advertisement or packaging” is a deep fake where it can mislead on actual appearance, including “making the product appear not identical to the real product, more appealing or with improved quality than in real life”. A real product on an AI background is not, “as long as the ad is not likely to mislead the audience”. The test in both is whether the image misleads.

They give no example of a catalogue photo of a fictional AI model. The nearest examples are videos: a realistic synthetic influencer testing a sponsored product, and simulated consumers in a teleshopping-style ad, both listed as deep fakes. They also say “photorealism alone is not determinative”. On that reasoning, a photorealistic AI model shown as a real person in an EU-facing image may qualify, depending on context. That is Herofab's reading, not a ruling, and it is unsettled until a regulator or court addresses it.

Where a deep fake exists, the Guidelines say deployers “cannot rely on the machine-readable marking” and must disclose visibly, at the latest at first exposure; a disclosure fails if it “can be easily overlooked or missed”. The Commission says content generated before 2 August 2026 need not be labelled retroactively. Under the Guidelines, brands outside the EU are covered where they themselves foresee use in the EU, including by posting on the globally accessible internet. Article 99 allows fines of up to EUR 15 million or 3% of worldwide turnover, with the lower figure for SMEs and small mid-caps. Zalando's image guidelines, read on 31 October 2026, do not allow visible on-image labels and say not to provide content that requires visible labelling under applicable laws, naming deepfakes. A seller on Zalando therefore cannot resolve the point with a label on the image.

When should a clothing brand not use an AI model?

A clothing brand should not use an AI model where the image is meant to prove how a garment fits or performs, because the garment was never worn in that image. Real try-on photography is the better method for fit evidence.

The same applies to an AI model that copies a real model without consent, resembles a known person or is presented as a real customer. Images of children raise separate rules not reviewed here, so take advice first.

Checklist

Add a conspicuous disclosure to New York-facing ads that show a synthetic performer.
Get written consent covering scope, purpose, rate of pay and duration for any real model's replica.
Do not name, upload or imitate a real person when briefing an AI model.
Compare colour, length, drape and details with the real garment before publishing.
Never present an AI person as a real customer.
Follow the channel's marking rules in the 2026 rules guide, and keep the file's metadata.

Sources

Checked 31 October 2026. Rules change; the official pages are the final reference.

Can I legally use AI models for my clothing brand?