Key Takeaways
- Eurocommerce wants AI-generated advertising exempted from EU deepfake transparency rules taking effect August 2
- Zalando reports 90% of its platform marketing content is now AI-generated, with turnaround times targeting under 24 hours
- The EU's broad deepfake definition now encompasses everything from fraud videos to AI-rendered furniture images
Major European retailers including Amazon, H&M, Ikea, and Inditex are pushing the EU to carve out an exception for AI-generated advertising from upcoming deepfake transparency rules. The problem: the EU AI Act, effective August 2, defines 'deepfake' so broadly that an AI-rendered image of a sofa technically falls under the same labeling requirements as synthetic videos designed to spread disinformation.
Eurocommerce, the trade association behind the lobbying effort, sent a letter to EU tech commissioner Henna Virkkunen arguing that product images created by AI shouldn't require the same disclosure as content meant to deceive. The association represents retailers generating €4.8 trillion in annual turnover across 5.4 million companies.
Why retailers are worried about EU AI Act deepfake labeling
The scale of AI adoption in retail marketing explains the urgency. Zalando says 90% of the content on its platform is now AI-generated. Matthias Haase, VP of Content Solutions at Zalando, told Reuters that generative AI cut their production cycle from weeks to days, with a target this year of under 24 hours from trend identification to live content.
H&M and Zara already use AI-generated clones of human models. If every one of these images requires a deepfake label, the disclosure would appear on the vast majority of fashion advertising in Europe. Eurocommerce's Director General Christel Delberghe argues this would water down the transparency rule's value for consumers. When everything carries a warning, nothing stands out.
The definition problem: what counts as a deepfake?
The term 'deepfake' originated in non-consensual pornography communities and became associated with fraud, election interference, and identity theft. The EU's application of this loaded term to commercial product photography reveals a drafting gap in the regulation.
Under the current framework, the EU distinguishes between fully AI-generated content (requiring an 'AI Generated' label) and partially AI-modified content (requiring a different disclosure). But both categories sweep in use cases that have nothing to do with deception. An AI-furnished apartment photo for a real estate listing, an AI-enhanced product shot with better lighting, and a synthetic video of a politician saying things they never said all fall under the same regulatory umbrella.
The disclosure requirements include a base icon indicating AI involvement, plus specific labels like 'Fully AI-generated' for content created entirely by AI without human creative input beyond prompting. For retailers running automated content pipelines, this creates compliance overhead at scale.
What the EU Commission hasn't said
The Commission has not responded to Eurocommerce's request. With the August 2 deadline approaching, retailers face a binary choice: label everything and accept the visual noise, or risk non-compliance. The lack of guidance on commercial AI content suggests the regulation's drafters may not have fully anticipated how quickly generative AI would become standard in advertising production.
Digital rights advocates push back on the exemption idea. Their argument: carving out advertising creates a loophole that could be exploited. What stops a bad actor from claiming their synthetic content is 'just advertising'? The line between commercial persuasion and manipulation has never been clean.
Implications for AI product teams building retail tools
For teams building AI content generation tools targeting European retail clients, the regulatory ambiguity creates product design questions. Should tools automatically watermark outputs? Embed metadata flags for downstream compliance? The safest approach may be building disclosure capabilities into the generation pipeline now, even if exemptions eventually materialize.
The 24-hour content turnaround Zalando is targeting only works if compliance doesn't add manual review steps. Automation platforms like Zapier or Make that connect content generation to publishing workflows will need to account for labeling requirements. The alternative is building compliance logic directly into generation tools.
Disclosure
Some links in this post are affiliate links — Logicity earns a commission if you sign up, at no extra cost to you. We only link products we have used or actively recommend.
Frequently Asked Questions
When do EU AI Act deepfake labeling requirements take effect?
The transparency requirements for AI-generated and AI-altered content take effect on August 2, 2025. Content qualifying as a 'deepfake' under the regulation must carry clear disclosure labels from that date.
Does AI-generated product photography require disclosure under EU law?
Under the current EU AI Act definition, yes. AI-generated or AI-modified images technically fall under deepfake transparency requirements. Eurocommerce is lobbying for an exemption for non-deceptive commercial content, but no exception exists yet.
What retailers are pushing for AI advertising exemptions?
Eurocommerce members including Amazon, H&M, Inditex (Zara's parent company), and Ikea are behind the lobbying effort. The association represents 5.4 million retail companies across the EU.
How much retail content is AI-generated?
Zalando reports that 90% of the marketing content on its platform is now AI-generated. Industry surveys suggest around 77% of major fashion retailers are experimenting with AI imagery.
What labels does the EU require for AI-generated content?
The EU requires a base icon indicating AI involvement, plus specific labels. 'AI Generated' applies to content created entirely by AI without human creative input. Partially AI-modified content requires different disclosure indicating human-created material was altered.
Logicity's Take
The EU wrote rules for one problem (political deepfakes, fraud) and accidentally regulated another (commercial content production). For AI product teams, this is a reminder that regulatory definitions often lag behind deployment speed. If you're building generative tools for e-commerce, bake disclosure metadata into your outputs now. Don't wait for clarity. Companies like Adobe already embed Content Credentials; open-source alternatives like C2PA offer similar provenance tracking. The cost of retrofitting compliance later will exceed the cost of building it in from the start.
Relevant for teams automating content workflows at the scale retailers now require
The deeper question for European regulators: can a single framework handle both election integrity and furniture photography? The current approach suggests they think so. Retailers betting on AI-first content strategies are now learning the limits of that assumption.
Need Help Implementing This?
Building AI content tools for European markets? We help product teams design compliance-ready generative pipelines. Contact us to discuss your architecture.
Source: The Decoder / Matthias Bastian
Huma Shazia
Senior AI & Tech Writer
Produced with AI assistance and reviewed by the Logicity editorial team. Learn more in our Editorial Policy.
