The EU AI Act, Regulation (EU) 2024/1689, entered into force on 1 August 2024. The transparency rules that matter most to fashion brands become enforceable on 2 August 2026. That sounds comfortable. It isn't. Building compliance documentation, auditing your existing campaigns, updating your model agreements, and establishing disclosure procedures takes months. Brands that start in the summer of 2026 will be rushing.
This is the guide for independent designers who want to understand what the Act actually requires. Not the summary that tells you to "watch this space," but the practical breakdown of what changes, what doesn't, and what you need to do about it.
Does the EU AI Act apply to your brand?
If you sell into European markets, through your own website, through EU-based retailers, at European trade shows, or via any distributor who reaches EU consumers, the Act can apply to you. The geographic location of your brand's incorporation does not provide an exemption.
This is the most commonly misunderstood aspect of the regulation. A New York-based designer with an EU stockist and a European DTC website is within scope. An LA-based brand that ships to Germany is within scope. The EU AI Act is extraterritorial in the same way that GDPR is: it follows the market, not the company.
The risk-based tier system, where fashion brands land
The Act categorises AI applications into risk tiers: unacceptable risk (banned entirely), high risk (strict controls, registration, auditing), limited risk (transparency obligations), and minimal risk (no specific obligations). The critical question for fashion brands is: which tier does your AI use fall into?
The good news: most fashion-related AI use, synthetic model imagery, AI design assistants, personalisation tools, generative pattern creation, falls under the transparency tier, not high risk. Your primary obligation is disclosure and documentation, not algorithmic auditing or regulatory registration.
The exception: AI used in hiring, credit scoring, or other high-stakes decisions falls into the high-risk tier. If you use AI to filter job applications, that's a different conversation.
What Article 50 actually requires
Article 50 creates transparency obligations for AI systems that generate or manipulate image, audio, video, or text content. It splits the duties between providers (the companies that build the AI tools) and deployers (the businesses that use them). A fashion brand generating campaign imagery with Midjourney or a similar tool is a deployer. These obligations apply from 2 August 2026.
For deployers, the core duty sits in Article 50(4): if you publish AI-generated or AI-manipulated image, audio, or video content that qualifies as a deep fake, content that resembles real people, objects, places, or events and would falsely appear authentic, you must disclose that the content is artificially generated or manipulated. For fashion campaigns, this maps directly onto photorealistic synthetic model imagery, AI-generated product visualisations, and substantially AI-altered photography.
Is a fully invented face a "deep fake" when it resembles no specific existing person? There is an argument that it isn't. I would not bet a campaign on that argument. The definition also covers realistic depictions of objects, places, and events, and the Commission's draft guidelines on Article 50, published in May 2026, read the obligation broadly.
Practically, Article 50 means you should:
Disclose when campaign imagery includes AI-generated synthetic models. If a person depicted in your campaign was generated by AI, whether fully synthetic or created by AI rendering over a real person's base image, plan to disclose it. Article 50(5) requires the disclosure to be clear, distinguishable, and delivered no later than the consumer's first exposure to the content. It cannot be buried in fine print.
Disclose when real model imagery has been substantially AI-altered. Standard post-production (colour grading, basic skin retouching) is not the target. AI that substantially alters body proportions, replaces physical characteristics, or composites a realistic depiction that differs materially from the original shoot is in scope. The line between "retouching" and "AI manipulation" is still being drawn; the Commission's draft guidelines and the final guidance to follow are where that line will come from.
Understand the marking layer. Article 50(2) requires providers of generative AI systems to mark outputs in a machine-readable format, watermarks and metadata. That burden falls mainly on your tools, not on you, but it means provenance data will increasingly travel with your imagery. One wrinkle: under the Digital Omnibus provisional agreement reached on 7 May 2026, generative systems already on the market before 2 August 2026 get until 2 December 2026 to meet the marking requirement. The deployer labelling duties were not delayed. I explain how provenance marking works in our provenance guide.
Maintain documentation of AI systems used in commercial content production. This is the most actionable near-term step. Keep records of which AI tools you use, which campaigns they appear in, and what oversight your human creative team exercises over the output.
The guidance finally exists
For two years the honest answer to "what exactly will disclosure look like?" was "wait for guidance." That wait is mostly over. On 10 June 2026 the Commission published the final Code of Practice on Transparency of AI-Generated Content, a voluntary framework that providers and deployers can sign to demonstrate compliance with Article 50. It is now undergoing an adequacy assessment. The EU has also released a set of standard icons that deployers may use to label AI-generated content. Commission guidelines on the scope of Article 50 are coming behind the draft published in May.
What disclosure looks like in practice
The Act does not mandate one specific disclosure format, which gives brands flexibility but also creates uncertainty. Based on the Code of Practice, the EU labelling icons, and parallel requirements in other jurisdictions (particularly New York's synthetic performer disclosure law and the FTC framework), effective disclosures include:
In-image or adjacent labelling: "Created with AI" or "AI-generated image" placed clearly on or near the campaign image in digital contexts. The EU icons are designed for exactly this.
Caption disclosure in social media: a clear statement in the post caption before any "more" truncation. Not a buried hashtag.
Ad disclosure in paid media: all paid advertising using AI-generated imagery should carry the disclosure in the ad unit itself, not just the landing page.
If you want a starting point, our disclosure generator produces channel-specific language you can adapt.
Penalties for non-compliance
Under Article 99(4), violations of the Article 50 transparency obligations can draw fines of up to €15 million or 3% of worldwide annual turnover, whichever is higher. Here is the detail that matters for independent designers: for SMEs and startups, the Act caps the fine at whichever of those figures is lower. For a brand generating €3M annually, that means maximum exposure of €90,000, not €15 million. Still real money. Not existential.
These are maxima. My expectation is that early enforcement will focus on deliberate deception rather than paperwork mistakes, but that is a prediction, not a promise. Enforcement is handled by national market surveillance authorities in each member state rather than a single central body, so enforcement patterns may vary by country and will develop over the first few years after August 2026. I track developments on the tracker.
The compliance checklist for independent designers
Audit your current campaigns now. Identify every piece of campaign content that involved AI generation or substantial AI alteration. Document what tools were used, what the human creative involvement was, and what's already published without disclosure.
Update your model agreements. Your standard model releases need language that distinguishes between traditional photography rights, AI alteration rights, and synthetic likeness rights. These are legally distinct and the consent for each needs to be explicit. California's AB 2602 is a useful benchmark for how specific that consent language should be.
Establish a disclosure template and workflow. Create standard disclosure language for each distribution channel, your website, social media, paid advertising, press materials, and add it to your pre-publish workflow as a non-optional step.
Document your AI tool stack. Keep a running record of which AI tools your brand uses, updated each time you adopt something new. This documentation is your first line of evidence if regulatory questions arise.
Brief your agencies and photographers. If you work with external creative partners who use AI in their process, make sure they understand your disclosure requirements and contractually agree to flag AI use to you.
This article is editorial analysis, not legal advice. For questions about your specific situation, consult a qualified attorney.