AI-generated models in fashion: what are the ethics and disclosure rules?

KEY TAKEAWAYS Summary by the editors

1. AI-generated models are either digital twins of real people created with their consent or fully synthetic people who do not exist.
2. H&M announced in March 2025 that it would create AI twins of 30 models, who retain control over their use and are paid in line with traditional shoots.
3. New York's Fashion Workers Act, effective 19 June 2025, requires clear written consent for creating or using a model's digital replica, specifying scope, purpose, pay and duration.
4. From 2 August 2026, Article 50 of the EU AI Act requires deployers to disclose deepfake content, and providers to mark synthetic outputs in machine-readable form.
5. A Guess advertisement featuring AI-generated models in the August 2025 issue of US Vogue drew strong public criticism despite a disclosure, showing that legal compliance does not guarantee audience acceptance.

AI-generated models are synthetic images or videos of people used to present clothing, either as consented digital twins of real models or as invented people. They can reduce production time and cost, but they raise questions of consent, payment, disclosure and trust, and new laws in New York and the EU now set concrete requirements.

What are AI-generated models?

Two categories should be kept apart, because the ethical and legal questions differ:

• Digital twins: AI replicas of real, identifiable models, trained on their images with permission. The person's likeness is the asset, so consent, scope and payment are central.
• Fully synthetic models: invented people generated by AI. No individual's likeness is used directly, but questions remain about training data, representation, labour displacement and transparency towards consumers.

Both are distinct from routine retouching or AI tools that only change a background, although the boundary is not always sharp.

Which fashion brands are using AI models?

The most structured publicly documented approach is H&M's. In March 2025 the retailer announced that it would create AI twins of 30 models who had consented. According to FashionUnited's report, the models retain control over how their replicas are used and receive compensation equivalent to traditional photo shoots, and H&M said initial images would carry watermarks identifying them as AI-generated. The same report notes earlier experiments by Levi Strauss in 2023 and Mango in July 2024.

Audience reaction can be sharp. In the August 2025 issue of US Vogue, an advertisement for Guess used AI-generated models created by the agency Seraphinne Vallora. The ad carried a note that the images were AI-generated, and Vogue said it came from its advertising department rather than its editorial team. The ad still drew significant backlash from readers, creators and image professionals, centred on job losses for models, photographers and production crews.

Read also

How are fashion brands using AI in marketing, from content to measurement?

Why are brands interested in AI models?

The commercial logic is straightforward. A conventional shoot requires models, photographers, stylists, locations and post-production, and every new colourway or market adaptation adds cost and time. AI models promise to show more products on more body types, in more settings, at lower marginal cost and with faster turnaround. For large assortments with frequent drops, that is attractive.

The savings are not free, however. Brands need high-quality reference imagery, specialist tools or agencies, legal review of consent and disclosure, and careful checking that the generated garment matches the real one. Synthetic imagery that misrepresents fit or fabric can raise return rates and complaints, eroding the cost advantage. And because audiences increasingly notice AI imagery, brands must weigh potential savings against the reputational risk of appearing to cut corners on people.

What does the law require for AI models?

Under the New York law, a digital replica is defined as a significant computer-generated or AI-enhanced representation of a model's likeness that substantially replicates or replaces their appearance or performance. Brands commissioning work involving New York models should check that agency contracts reflect these requirements.

In the EU, a realistic image of an identifiable real person in a situation that did not happen can fall within the AI Act's notion of a deepfake. Whether a given fully synthetic model does is a legal question that depends on the image and context, so brands should take advice and, as a minimum, follow their tool providers' marking.

What ethical questions should brands consider?

1. Consent and control: can the model withdraw consent, approve uses and see where their twin appears?
2. Fair payment: is compensation tied to actual use, including reuse across markets and seasons?
3. Representation: synthetic models can be used to signal diversity without employing diverse people, which critics see as tokenism.
4. Product truth: an AI model must not show a garment fitting or draping differently from reality.
5. Crew and supply chain: photographers, stylists, hair and make-up artists are also affected, as the Guess reaction showed.

How should a brand set an AI model policy?

A workable policy defines where AI models may be used (for example, social formats and colour variants but not hero campaigns), requires written, specific consent for any digital twin, sets disclosure standards that meet the strictest market the brand sells into, and keeps an audit trail of which images are synthetic. It should also state that product depiction must be checked against the physical garment. Involving model agencies early reduces both legal risk and reputational surprises.

Read also

What is GEO, and how can fashion brands be found in AI search?

What happens next?

The direction of travel is towards more consent requirements and more labelling, not less. With the EU transparency rules applying since August 2026 and New York's replica provisions in force, brands that build consent and disclosure into their content workflows now will avoid retrofitting later. The commercial case for AI models is real, but it rests on trust that can be lost quickly.

Frequently asked questions

Are AI-generated models legal in fashion advertising?

Generally yes, but conditions apply. In New York, using a model's digital replica requires specific written consent covering scope, purpose, pay and duration. In the EU, AI Act Article 50 requires disclosure of deepfakes from 2 August 2026, and ordinary advertising rules on misleading content apply everywhere.

Do brands have to label AI models?

In the EU, deployers must disclose deepfake content under Article 50 of the AI Act, and AI providers must mark synthetic outputs. In the UK there is no AI-specific disclosure rule, but ads must not mislead. Many brands, including H&M for its AI twins, have chosen to watermark voluntarily.

What is a model's digital twin?

A digital twin is an AI replica of a real model, created from their images with permission, that can be placed in new images or videos. H&M announced twins of 30 consenting models in 2025, with the models retaining control over use.

Why did the Guess AI model ad cause backlash?

The August 2025 Vogue ad used AI-generated models created by an agency. Although it was labelled, readers and industry professionals criticised it for displacing models, photographers and crews and for appearing in a prestige fashion title.

GuideThe complete guide to AI in fashion e-commerce, marketing and retailRead the complete guide