EU AI Act Article 50: What fashion brands must disclose from August 2026
Article 50 of the EU AI Act requires brands to label AI-generated or manipulated images of people from 2 August 2026. Here's what must be disclosed and how.
KEY TAKEAWAYS Summary by the editors
- Article 50 of the EU AI Act mandates disclosure of AI-generated or AI-manipulated content depicting people, enforceable from 2 August 2026.
- Fashion brands must label synthetic imagery of people in marketing, e-commerce and social media with clear, machine-readable declarations.
- The disclosure applies when content creates realistic representations of people, including faces, bodies and clothing fit on virtual models.
- Non-compliance can result in fines up to 1.5 per cent of global annual turnover or 7.5 million euros, whichever is higher.
Article 50 of the EU Artificial Intelligence Act, which entered into force on 1 August 2024, imposes transparency obligations on deployers of AI systems that generate or manipulate image, audio or video content. From 2 August 2026, fashion brands must disclose when they use AI to generate or significantly alter depictions of people in their imagery. This requirement covers product photography, campaign visuals, social media posts and e-commerce sites where synthetic or manipulated human likenesses appear.
What content must be labelled under Article 50?
The transparency requirement applies to content that resembles existing persons, objects, places or events and would falsely appear authentic or truthful to a person. For fashion brands, this means any AI-generated or AI-manipulated imagery depicting people, whether real individuals or synthetic models, must carry a disclosure if the content could be mistaken for a genuine photograph.
The obligation covers deep fakes and other AI-generated or manipulated content depicting realistic images, audio or video of persons or events. In practice, this includes virtual models wearing garments, AI-enhanced product shots showing clothing on digitally created bodies, and images where a model's face, body shape or skin tone has been substantially altered using generative AI tools. Minor edits such as brightness adjustments or colour corrections are exempt if they do not materially alter the content.
Where must the disclosure appear?
Deployers must disclose that the content has been artificially generated or manipulated by marking it in a machine-readable format and ensuring it is detectable as such. For fashion brands publishing imagery on owned channels, this means embedding technical metadata alongside visible labels that inform viewers the content is synthetic or AI-altered.
The disclosure must be clear and conspicuous. Where technically feasible, the disclosure must be embedded in the content itself, for example through watermarking or metadata. On e-commerce product pages, this could mean a label near the image stating the model is AI-generated. On social media, brands should include a text disclaimer in the caption or overlaid on the visual itself. Burying the notice in terms of service or privacy policies is insufficient.
Are there exceptions for fashion content?
Certain uses are exempt, including content that is necessary for the exercise of the right to freedom of expression and the right to freedom of the arts and sciences, and content that forms part of an evidently artistic, creative, satirical, fictional or analogous work or programme. This carve-out may apply to avant-garde fashion campaigns explicitly presented as artistic projects or digital fashion collections marketed as virtual-only creations.
However, commercial product imagery and standard marketing materials do not qualify for the artistic exception. A brand selling physical garments using AI-generated models on its website or in paid advertising must label that content, even if the creative execution is highly stylised. The exception is narrow and requires the artistic or satirical intent to be evident to the audience.
What are the penalties for non-compliance?
For infringements of transparency obligations, companies can face fines of up to 1.5 per cent of total worldwide annual turnover in the preceding financial year or 7.5 million euros, whichever is higher. National authorities in each EU member state are responsible for enforcement, and penalties will be determined on a case-by-case basis depending on the severity and duration of the violation.
Beyond regulatory fines, brands risk reputational damage if consumers or advocacy groups highlight unlabelled synthetic imagery. In an industry where trust and authenticity are increasingly scrutinised, failure to comply with Article 50 can erode brand credibility and invite public criticism.
How should fashion brands prepare?
Compliance requires coordination across creative, e-commerce, legal and IT teams. Brands should begin by auditing existing and planned content to identify where AI-generated or AI-manipulated depictions of people are used. This includes reviewing product photography workflows, social media calendars, campaign assets and any third-party content such as influencer partnerships where the brand has editorial control.
Next, establish clear labelling protocols. Define the disclosure language, placement and format for each channel. For instance, e-commerce sites might add a badge reading 'AI-generated model' beneath product images, while Instagram posts could include '#AIgenerated' in the caption and a text overlay on the image. Ensure metadata standards are implemented so that disclosures are machine-readable and persist if content is downloaded or shared.
Train internal teams and external agencies on the requirement. Photographers, retouchers, social media managers and content creators must understand what triggers the disclosure obligation and how to apply labels correctly. Update creative briefs and vendor contracts to specify compliance responsibilities.
Finally, monitor enforcement guidance from national authorities and industry bodies as the deadline approaches. Interpretation of Article 50 may evolve through regulatory clarifications and case law, and brands should stay informed to adjust their practices accordingly.
Compliance checklist
- Audit all content channels to identify AI-generated or AI-manipulated depictions of people.
- Establish standardised disclosure language and placement for each platform and format.
- Embed machine-readable metadata in images and videos to indicate synthetic or altered content.
- Add visible labels on or adjacent to AI imagery, such as badges, captions or text overlays.
- Update creative workflows and vendor contracts to assign responsibility for disclosure compliance.
- Train internal teams and external partners on what content requires labelling and how to apply it.
- Review claims for artistic or creative exemptions with legal counsel to ensure they are defensible.
- Document compliance processes and retain records of labelling decisions for regulatory audits.
- Monitor guidance from national regulators and adjust practices as enforcement precedents emerge.
Frequently asked questions
Does Article 50 apply to AI-edited product photos without people?
No. Article 50 specifically targets content that depicts people, objects, places or events in a way that could falsely appear authentic. If a product image shows only garments on a hanger or flat lay with no human likeness, and no other realistic elements are synthetically generated, the disclosure requirement does not apply. However, if an AI tool creates a realistic background scene or setting that could be mistaken for a real place, labelling may be necessary.
Do brands need to label every image retouched with AI tools?
Not necessarily. Minor edits such as brightness, contrast or colour correction are exempt if they do not materially alter the content. The disclosure obligation arises when AI substantially changes the depiction of a person, such as generating a synthetic model, altering body shape or face, or creating a deep fake. Routine retouching that does not fundamentally change the realism or identity of the person may not require labelling, but brands should document their reasoning.
Can a brand avoid Article 50 by using AI-generated models only outside the EU?
No. The EU AI Act applies to deployers placing AI systems on the market or putting them into service within the EU, regardless of where the deployer is established. If a brand markets AI-generated imagery to EU consumers via its website, social media or other channels accessible in the EU, it must comply with Article 50. Geographic segmentation of content may reduce exposure, but enforcement depends on where the content is made available, not where the company is headquartered.