EU dispute: should AI-made ads carry a label?

What this is about

Europe’s retail sector is pushing for an exception before the next AI Act phase: AI-generated advertising should not automatically fall under transparency duties for synthetic content. Reuters reported on June 19, 2026 that EuroCommerce sent EU tech chief Henna Virkkunen a letter asking for that clarification.

This sounds like a legal detail, but it touches everyday imagery: a sofa in an artificially generated living room, a digitally altered dress photo, or an ad video with a synthetic background. From August 2, 2026, the EU AI Act transparency rules move closer to this kind of content.

What Article 50 actually does

Article 50 of the AI Act sets transparency duties for certain AI systems. For synthetic or manipulated image, audio, and video content, the point is that people should be able to recognize when material was artificially generated or changed. Providers also face machine-readable marking duties.

The conflict sits in the definition of deepfakes and in the question of when advertising actually deceives. According to Reuters, EuroCommerce argues that ordinary product visualization should not be treated like misleading deepfakes. An AI-generated living room used to show a sofa is different from a manipulated video of a real person.

Why it matters

Advertising is one of the first areas where generative image tools are being used at scale. Reuters cites Zalando as an example: the company said AI reduced content production costs by about 90 percent. H&M and Zara have also used AI-generated models or digital replicas, according to the report.

For consumers, the question is still not trivial. As synthetic product imagery becomes more realistic, an image can show properties that were never photographed: light, fit, room atmosphere, body shape, or material feel. A label is not a moral warning. It is a provenance signal.

In plain language

It is like a cake photo on packaging. Nobody expects the cake to come out of every kitchen looking exactly like the picture. But if the picture is not a photo of any real cake and was fully created on a computer, many people would want to know. Not because the cake is automatically bad, but because the image was made differently.

A practical example

A furniture retailer creates 5,000 online ads per week. Previously, 500 were photographed and 4,500 were assembled from existing product images. In the future, 4,000 ads might show AI-generated rooms: the same table in 20 apartments, at morning light, evening light, and with different wall colors.

Without clear rules, the company must decide whether every one of those visuals needs a label. With a blanket advertising carve-out, shoppers may no longer see which images are real photography and which are synthetic visualization. That is the political conflict.

Scope and limits

First, it is not clear whether the Commission will support the requested exception. TNW writes that there is so far no sign of a broad category carve-out.

Second, not every AI-assisted ad graphic is risky. A generated background for a lamp is different from a synthetic person endorsing a product.

Third, too much labeling can also numb people. If every banner carries an AI label, consumers may pay less attention to warnings that genuinely matter.

SEO & GEO keywords

EU AI Act, Article 50, EuroCommerce, AI-generated ads, AI advertising, deepfake transparency, retail marketing, Zalando, H&M, Zara, synthetic media, consumer rights