Model Releases and Likeness Rights When Using AI Models

What Is a Model Release and Why AI Complicates It

A model release is a legal agreement in which a person grants a brand or photographer the right to use their image for commercial purposes. In traditional fashion photography, this is straightforward: a model signs before the shoot, and the release defines where and how the images may appear.

AI-generated content disrupts this framework at two points. First, many AI image and video systems were trained on datasets containing real people’s photographs, sometimes without those individuals’ knowledge or consent. Second, some AI tools allow users to upload reference images of real people to guide the visual output. In both scenarios, an argument exists that the original person’s likeness is being commercially exploited, which triggers release obligations in many jurisdictions.

The safer alternative, and the one most professional AI fashion tools now take, is to generate entirely synthetic models built from no single identifiable person. When the AI output is provably not derived from a specific individual, likeness claims become substantially harder to sustain. That said, brands still need to understand where the training data came from, because that question sits upstream of the output.

How Likeness Rights Work Across Jurisdictions

Likeness rights, sometimes called the right of publicity, vary significantly by country and, in the United States, by state. Key points include:

• In the United States, states including California, New York and Tennessee have strong right-of-publicity statutes. California’s law now explicitly addresses AI-generated replicas of performers, requiring consent before a digital likeness is used commercially.
• In the European Union, likeness rights fall partly under the General Data Protection Regulation where biometric data is involved, and partly under national personality rights laws. Using a recognisable person’s face in commercial AI output without consent can constitute both a data protection breach and a tort.
• In the United Kingdom, there is no standalone right of publicity, but passing-off law and data protection rules under the UK GDPR offer meaningful protection to individuals whose appearance is used without consent.
• Australia, Canada and Japan each have their own personality rights or privacy frameworks that a fashion brand publishing globally must consider.

The practical implication for fashion e-commerce teams is this: publishing AI model content globally means complying with the most restrictive applicable law, not the most permissive one.

AI Model Legal Obligations When Using Third-Party Tools

When a fashion brand uses an third-party AI video or image platform, legal responsibility does not transfer entirely to the vendor. Contracts and terms of service matter enormously. Before signing up to any AI model tool, brands should confirm the following:

1. Training data provenance: Was the model trained on licensed, ethically sourced imagery? Reputable platforms provide documentation or public statements on this point.
2. Output ownership: Who owns the generated video or image? Some platforms retain a licence to reuse your outputs; others assign full ownership to the customer.
3. Indemnification clauses: Does the platform indemnify you if a third party claims their likeness was used without consent in the training data? Many do not, so your legal exposure may be greater than you expect.
4. Platform-specific rules: TikTok, Instagram and other distribution channels have their own policies on AI-generated content disclosure. Violations can result in content removal or account suspension. If you are creating content for TikTok Shop product videos, review those disclosure requirements before publishing.

Outfit Video generates fashion videos using synthetic model representations derived from your garment photography rather than from identifiable real people, which removes the primary likeness risk for brands focused on product-first content.

When You Use Your Own Photos or Real-Model Inputs

Some AI video workflows accept photographs of real models as input, animating or adapting them to create short-form content. If your brand takes this approach, the release you obtained for the original photoshoot may not automatically cover AI-generated derivative content. Standard photography releases typically specify the media types and uses permitted. AI-generated video is a new use category, and many legacy release templates do not address it.

The fix is straightforward. Update your model release template to include explicit language covering AI-generated derivatives, animated outputs and synthetic adaptations of the original imagery. Obtain fresh releases from existing models if you intend to use their photographs as AI input. If you work with a model agency, check the agency contract too, because usage rights are often layered between the agency agreement, the model’s personal release and any union rules that apply.

For brands that use ghost mannequin or flat-lay photography as the primary input rather than images of real people, this concern largely disappears. The ghost mannequin approach is one reason many fashion sellers choose it as their standard AI video input method.

Resellers and Small Sellers: The Same Rules Apply

Individual resellers on platforms such as Depop and Vinted increasingly use AI-generated imagery to make listings look more polished. The legal framework does not scale down for small sellers. A sole trader publishing AI model content without understanding the underlying likeness questions carries the same theoretical liability as a large brand, though enforcement risk is practically lower.

More concretely, platform terms of service apply to all sellers regardless of size. Both Depop and Vinted prohibit misleading imagery, and AI-generated models that misrepresent how a garment fits or looks could trigger listing removal. If you are using AI video to improve your resale listings, the guidance in our post on Depop and Vinted listing videos that sell faster covers best practices for accurate, compelling presentation without crossing into misleading territory.

Building a Compliant AI Model Workflow

Compliance does not have to slow down production. A structured approach lets you move fast and stay within legal boundaries.

• Choose tools with transparent training data policies. If a vendor cannot tell you where its training images came from, treat that as a red flag.
• Use garment-only or ghost-mannequin inputs wherever possible. This removes likeness questions from the equation entirely.
• Update all model release templates to include AI-generated derivative content as an explicit permitted use.
• Disclose AI-generated content on platforms that require it, and adopt a consistent disclosure practice even on platforms that currently do not mandate it. Regulatory trends point toward more disclosure requirements, not fewer.
• Review vendor contracts annually. AI tool terms of service change frequently, and a clause that was acceptable last year may not be today.
• Consult an intellectual property or entertainment lawyer if your brand operates across multiple jurisdictions or uses real people’s images as AI inputs at scale.

The cost of building a compliant workflow is modest relative to the cost of a likeness claim or platform suspension. Fashion brands that treat legal hygiene as part of their content operations, rather than an afterthought, are better positioned to scale AI video production without interruption.

Ready to turn your outfit photos into scroll-stopping videos? Try Outfit Video free and create your first AI fashion video in minutes.

Ready to turn your outfit photos into scroll-stopping videos? Try Outfit Video free and create your first AI fashion video in minutes.