AI Models in Fashion Ads: The Disclosure and Consent Rules Brands Face in 2026

AI-generated models in fashion ads are now regulated in three places at once. Since 9 June 2026, New York has required advertisers to “conspicuously disclose” a synthetic performer when they know one is in the ad. California will require the same from 1 January 2027 under SB 1050, which Governor Newsom signed on 16 September 2026. In the EU, the AI Act’s deepfake labelling rule has applied since 2 August 2026. Consent is regulated too. Under New York’s Fashion Workers Act, a model’s digital twin can only be created or used with the model’s separate written consent. Using an AI model is legal. Hiding that you used one, or cloning a real model without a signed agreement, increasingly is not.

This guide covers what each rule requires, the campaigns that brought the issue to a head, and a checklist for brands, agencies and models.

why this became a fashion law issue

The flashpoint was a single page. In the August 2025 issue of US Vogue, a Guess advertisement featured a model generated with AI by the agency Seraphinne Vallora, with the AI use noted in small print. The backlash was immediate. Readers, models and campaigners asked why a real person had not been booked, and why the disclosure was so easy to miss.

The industry had been moving this way for a while:

• Levi’s, 2023: announced a partnership with Lalaland.ai for AI-generated models and faced criticism. It said the models would “supplement but not replace” live shoots.
• Mango, 2024: ran its first fully AI-generated campaign, for its Teen line, across 95 markets.
• H&M, 2025: launched a digital twin campaign on 2 July 2025, and said the models keep ownership of their twins and share in the revenue, with the images labelled.

The H&M model is roughly where the law is heading: consent, pay and a label.

what does New York’s synthetic performer law require?

New York’s law was passed as S.8420-A / A.8887-B and signed on 11 December 2025. It adds section 396-b to the General Business Law. The bill text defines a “synthetic performer” as a digitally created asset, created or modified by computer using generative AI or a software algorithm. It must give the impression of a human performer who is “not recognizable as any identifiable natural performer”.

• The duty: an advertiser must “conspicuously disclose” a synthetic performer in an ad “where such person has actual knowledge”.
• Penalties: $1,000 for a first violation and $5,000 for each later one.
• Exemptions: ads for expressive works such as films, TV and games; audio-only ads; AI used only for language translation; and media outlets that publish the ad, unless they have been given written notice.
• Effective date: 180 days after signing, which was 9 June 2026, as Cooley notes.

Governor Hochul signed a companion law the same day. It requires consent from heirs or executors before a deceased person’s name, image or likeness is used commercially.

The first complaints are in. In August 2026, Straight Arrow News reported that two complaints had been made to the New York Attorney General. They named an athletic apparel brand and an eyewear brand over ads with apparently undisclosed AI models. No enforcement action had been announced.

what does California’s SB 1050 add?

California followed on 16 September 2026, when Governor Newsom signed SB 1050. According to Davis+Gilbert’s summary, it:

• takes effect on 1 January 2027
• sets required wording, such as “this performance features a synthetic performer” or “no human performer is depicted”
• covers audio ads, unlike New York
• applies where the synthetic performer appears prominently
• is enforced through California’s false advertising and unfair competition laws

California already had two digital replica laws. AB 2602, in force since 1 January 2025, makes a contract term allowing a digital replica unenforceable if three things are all true. The replica replaces work the person would have done in person. The intended uses are not reasonably specifically described. And the person had neither a lawyer nor a union representing them. AB 1836 covers digital replicas of deceased personalities, with damages of at least $10,000.

what does the Fashion Workers Act say about digital twins?

New York’s Fashion Workers Act took effect on 19 June 2025. It is the first law written specifically for models, and it covers digital replicas directly:

• Model management companies must obtain “clear written consent for the creation or use of a model’s digital replica, detailing the scope, purpose, rate of pay, and duration of such use”. That consent must be separate from the representation agreement.
• Clients, meaning brands and retailers, need “clear and conspicuous prior written consent”.
• Management companies may not create, alter or manipulate a replica without clear, separate written consent.
• A power of attorney cannot cover a model’s digital replica. The New York Department of Labor’s FAQ confirms that existing agreements that tried to do so became void on 19 June 2025.

A digital replica is defined as a significant, computer-generated or AI-enhanced representation of a model’s likeness. Routine photographic edits are excluded. So retouching a shoot is fine, but building a reusable 3D version of the model is not fine without a signed consent that names the scope, pay and duration.

For models, the practical point is to read every agreement for replica language before signing. Our piece on how long modelling careers last explains why the long-term value of a likeness matters so much.

what does the EU AI Act require?

The AI Act defines a deepfake as AI-generated or manipulated image, audio or video content that resembles existing persons, objects, places, entities or events and would falsely appear to be authentic (Article 3(60)). Under Article 50(4), anyone deploying an AI system that generates deepfakes must disclose that the content has been artificially generated or manipulated. The disclosure must be “clear and distinguishable” at the latest at first exposure (Article 50(5)). For evidently artistic or creative works, the disclosure may be lighter, in a way that does not spoil the work.

The rule has applied since 2 August 2026. The EU’s Digital Omnibus on AI, in force from 27 July 2026, did not postpone it. It only gave providers of generative systems already on the market until 2 December 2026 to add machine-readable marking. Breaches can bring fines of up to €15 million or 3 per cent of worldwide turnover.

Does an invented AI model count as a deepfake? The definition covers content resembling “existing persons”, which clearly catches a digital twin of a real model. A wholly invented face is less certain, but a photorealistic “person” in a product ad would falsely appear authentic, and many brands are labelling both to be safe.

what about the UK?

The UK has no specific law. The Advertising Standards Authority said in May 2025 guidance that “there is no blanket legal requirement in the UK to disclose the use of AI in ads”, and that the existing CAP Code applies. It also warned that “disclosure alone is very unlikely to mitigate harm caused by a fundamentally misleading message”. An AI image that exaggerates how a product performs or fits is misleading whether or not it is labelled. Beauty and shapewear ads are the obvious risk.

is there a federal US law?

Not yet. The NO FAKES Act would create a federal right to control digital replicas, with a notice-and-takedown system. A revised version was reported unanimously by the Senate Judiciary Committee on 18 June 2026. It would preempt future state laws while keeping state laws in force as of 2 January 2025, such as Tennessee’s ELVIS Act. It has not passed the full Senate or the House.

can a brand own the copyright in an AI model image?

Possibly not, at least in the US. The US Copyright Office’s January 2025 report says copyright “does not extend to purely AI-generated material, or material where there is insufficient human control over the expressive elements”. Prompts alone are not enough. In Thaler v Perlmutter, the D.C. Circuit held in March 2025 that the Copyright Act requires human authorship, and the Supreme Court denied review in March 2026.

For brands, this means a campaign built entirely from AI output may be easy for competitors to copy. Human selection, arrangement and editing can be protected, so keep records of the creative work people did. Our guide to fashion law versus IP law covers the wider framework.

a checklist for brands, agencies and models

Brands and agencies

1. Keep a register of every asset that uses a synthetic performer or a digital replica.
2. Add a clear disclosure wherever the ad runs in New York, and in California from 2027. Use California’s set wording if you run national campaigns.
3. For EU audiences, label AI-generated people in ads, and check your production tools for machine-readable marking.
4. Get separate written consent for any digital twin, naming scope, purpose, pay and duration. Do not rely on a power of attorney.
5. Check that AI imagery does not exaggerate fit, finish or results, or the UK and US advertising regulators may still treat it as misleading.
6. Document human creative input so you can claim copyright where it exists.

Models

1. Look for digital replica clauses in every agency and client agreement.
2. Ask for scope, duration, territory and pay in writing, and the right to approve uses.
3. In California, get a lawyer or union to review any replica clause, since AB 2602 protects unrepresented performers.

Contracts with creators raise the same issues. See how influencer contracts should handle IP and usage rights.

frequently asked questions

Do brands have to disclose AI-generated models in ads?

In New York, yes, since 9 June 2026, where the advertiser knows a synthetic performer is used. California requires it from 1 January 2027. In the EU, AI-generated deepfakes must be labelled under the AI Act from 2 August 2026. The UK has no blanket rule, but ads must not mislead.

Do models have to consent to a digital twin?

In New York, yes. The Fashion Workers Act requires separate, clear written consent covering scope, purpose, pay and duration, and a power of attorney cannot cover it. California’s AB 2602 makes some replica clauses unenforceable where the performer was not represented.

What is the fine for not disclosing an AI model in New York?

$1,000 for a first violation and $5,000 for each later violation, under General Business Law section 396-b.

Does the EU AI Act require labelling AI images in advertising?

Article 50(4) requires deployers of AI systems that generate deepfakes to disclose that the content is AI-generated or manipulated. It has applied since 2 August 2026, and it was not delayed by the Digital Omnibus.

Can you copyright an AI-generated model?

In the US, purely AI-generated material is not protected by copyright, according to the Copyright Office and the D.C. Circuit in Thaler v Perlmutter. Human creative contributions to the image can be protected.

sources

• New York S.8420-A, synthetic performer disclosure
• Office of the Governor of California, SB 1050
• New York Fashion Workers Act, S9832
• EU AI Act, Article 50
• ASA, disclosure of AI in advertising
• US Copyright Office, Copyrightability report

Get fashion law news and the weekly jobs and internships list in FLJ Insider. This article is general information, not legal advice.