Synthetic Faces, Digital Twins, and Fashion’s New Compliance Divide
11/9/20265 min read
Fashion imagery made with generative tools can fall into two categories. A fully synthetic model is an appearance created without an intention to depict a particular, identifiable person. Its face, body, and styling may be assembled or generated from prompts, reference material, or model outputs, but the creative aim is not to reproduce an individual’s identity. By contrast, a digital replica is designed from, or materially derived from, a real person’s likeness. It may draw on photographs, scans, recordings, or personal data, and can reproduce recognizable features even when the image is rendered, retouched, or placed in a setting.
The boundary is not always clear. A face generated without a named subject might coincidentally resemble someone, raising questions about recognition and context rather than original intent alone. Conversely, a digital twin may be modified with invented traits, combined with visual elements, or placed in a scene that never existed. Those changes do not automatically erase the source person’s identity or rights. Relevant facts can include the inputs used, degree of resemblance, whether audiences could identify the person, and whether consent or licensing covers the particular transformation and use. No single label resolves every issue.
For brands, classification should be a starting point, not a shortcut. Teams should examine how an image was made, whose identity it may evoke, the campaign’s purpose and distribution, and the relevant jurisdictions. Applicable rules vary in definitions, scope, effective dates, and treatment of advertising, employment, privacy, publicity rights, and disclosure. Some requirements may target certain synthetic depictions or contexts; others may depend on recognizable likeness, consent, or data practices. A synthetic fashion image does not, by itself, trigger identical disclosure or contract duties everywhere. Assessment should be specific to the asset and its planned use for each campaign.
For fashion brands, transparency around synthetic models and digitally altered campaign imagery is not a single-label exercise. Article 50 of the EU AI Act establishes different duties according to the content, the actor involved and the context of use. Providers of systems that generate synthetic audio, image, video or text must generally ensure outputs are marked in a machine-readable, detectable way, subject to technical feasibility and specified exceptions. This is distinct from the deployer’s obligation to disclose when an image, audio or video constitutes a deepfake: a realistic depiction of a person, object, place or event that has been generated or manipulated and could falsely appear authentic.
Article 50 also contains context-specific provisions, including rules for certain AI-generated text published to inform the public on matters of public interest. Exceptions apply in defined circumstances, such as where content is evidently artistic, creative, satirical or fictional; disclosure must still be provided in an appropriate manner that does not hamper the work’s display or enjoyment. The Act’s general application date is 2 August 2026, although brands should verify the relevant timetable and any applicable transitional provisions. New York has adopted separate disclosure requirements concerning synthetic performers in advertisements, with an effective date and scope that should be checked against current law and the campaign’s circumstances.
In practice, a fashion business should assess whether it is acting as a provider, deployer, advertiser or another relevant party, and determine where the campaign is directed and which territorial rules apply. A clear notice placed close to the affected image or product presentation can help audiences understand what they are seeing, but a caption, icon or “AI-generated” label does not automatically satisfy every legal duty. Some obligations concern machine-readable marking rather than consumer-facing wording; others require a disclosure that is perceptible to viewers. Brands should document their assessment, review exceptions carefully and coordinate notices across websites, social platforms and advertisements. Missing an applicable disclosure can bring regulatory and reputational exposure, while a presentation that misleads consumers may also raise separate consumer-protection concerns.
A campaign that uses a model’s digital replica raises questions distinct from ordinary photography rights. Permission to pose for a shoot does not automatically authorize a brand or agency to generate a synthetic likeness, train systems on it, or deploy it in new contexts. Those activities may implicate publicity and privacy rights, intellectual property, data protection rules, employment obligations, and contractual promises. Applicable rules vary by jurisdiction and circumstances, so parties should distinguish capture of original images from creation and later use of a replica. Consent should be informed and affirmative, with enough detail for a model to understand the proposed uses.
New York’s Fashion Workers Act adds protections for covered fashion work and relationships, including provisions relevant to digital replicas. Its application depends on statutory definitions and the parties’ roles; brands and agencies should verify whether a particular engagement is covered and confirm current requirements with qualified counsel. Where applicable, the Act reinforces the need for written, appropriately specific authorization rather than reliance on a generic, open-ended buyout. A broad release may not clearly establish permission for materially different synthetic uses. The statute also addresses working conditions and protections, including work-hour requirements in covered circumstances; requirements must be checked in context.
A written agreement should identify the replica and obtain informed consent, then specify the intended purposes, media, geographic territory, and duration. It should state what edits, transformations, or combinations with other material are permitted, and whether exclusivity applies. Compensation for replica creation and use should be addressed separately from modeling fees, with clear payment terms. The contract should also limit reuse, transfer, and sublicensing, naming any authorized recipients and requiring fresh approval for uses beyond the grant. Practical safeguards may include access controls, retention limits, deletion procedures, and a process for reporting unauthorized use. Clear drafting clarifies the rights granted.
Fashion brands should begin with an asset register that distinguishes provenance from permission to use. For each face or digital model, record whether it is fully synthetic, derived from licensed source material, or connected to an identifiable person. This classification is a starting point, not legal clearance: lawful creation does not automatically authorize every commercial use, and a license may not cover likeness rights or particular channels.
Next, audit the systems and materials used to generate the asset. Document permissions for training and reference datasets, including applicable licenses, model releases, and any restrictions on commercial use. Assess whether collecting or processing images raises right-of-publicity or biometric-privacy concerns in relevant jurisdictions. Retain an auditable record of vendors, prompts, source files, model versions, licenses, and internal approvals so the brand can trace how an image was made and the basis for its use.
Before commercial publication, conduct a review for unintended resemblance to real people. If reviewers, talent representatives, or other stakeholders identify a plausible resemblance, pause release and escalate it to legal and appropriate business leads. Do not treat a synthetic label as a substitute for investigation. Separately, use a jurisdiction-specific disclosure checklist to determine whether audiences must be told that an image or model is generated or digitally altered. When a campaign uses a person’s digital replica, obtain a separate written agreement specifying informed consent, permitted uses and media, territory, duration, approval rights, and compensation.
Misclassification can expose a brand to private claims, statutory remedies where available, and regulatory scrutiny, while also delaying campaigns and undermining consumer and talent trust. A clear workflow helps teams identify problems before publication rather than relying on informal assurances from a vendor. Assign ownership for each approval, preserve records for the life of the campaign and any applicable retention period, and schedule periodic legal review. Laws, disclosure rules, and enforcement guidance continue to evolve, so processes should be updated as requirements change.
Luxury
Elevate your brand with our exclusive AI models.
Contact us
© 2026. All rights reserved.
(609) 901-8073
