EU Retailers Want AI Ads Exempt from Transparency Rules — While Brands Flood Social Media with Fake Influencers
Three days ago, The Guardian published an investigation revealing something quietly unsettling: brands are deploying AI-generated influencers on social media who purport to be real customers, with no indication that the people in the videos don’t exist. A “crying bride” praising a wedding photo app. An AI woman saying she could “kiss the interior designer.” A fashion brand posting images of models with extra fingers.
On the same day in Brussels, Eurocommerce — the retail trade association representing Amazon, H&M, Inditex (Zara’s parent company), and Ikea — formally asked the EU’s top technology official to carve advertising out of the bloc’s incoming AI transparency rules.
The timing is too neat to ignore. The rules kick in on 2 August, giving retailers barely two months to adjust.
Article 50 — the rules retailers don’t want
Article 50 of the EU AI Act requires deployers of AI systems to clearly disclose when images, audio, or video have been artificially generated or manipulated. Providers of generative AI tools must also embed machine-readable watermarks in their output — a digital fingerprint that flags AI-generated content even if the human-facing label is stripped away.
There is no minimum spend threshold. There is no blanket exemption for advertising. As drafted, every AI-touched advert — from a local bakery’s Instagram post to a global fashion campaign — falls under the same transparency requirement.
Eurocommerce’s argument, as reported by Reuters on 19 June, is one of proportionality. Labelling every piece of AI-touched promotional material, they argue, imposes a compliance burden disproportionate to the actual risk to consumers, particularly for routine marketing that nobody mistakes for journalism.
The existing law already has narrow exemptions. If a human editor takes responsibility for AI-assisted copy, that’s one carve-out. If the AI involvement is “obvious to a reasonably observant person,” that’s another. But both are conditional and case-by-case — not the clean exclusion the retailers are looking for. What they want is a category-wide exemption for advertising.
The double-edged sword of dual liability
Here’s the detail that makes this interesting from a technical perspective: Article 50 creates dual responsibility for a single piece of content. Both the provider of the AI tool (the company that made the image generator) AND the deployer (the advertiser who used it) are on the hook.
For a sector that has spent the last 18 months rushing to adopt generative tools for everything from product photography to social media campaigns, that dual liability is the real sting. It means an advertiser can’t just blame their AI vendor — they’re equally responsible for making sure the disclosure is there.
The stakes are calibrated accordingly. Non-compliance draws fines of up to €15 million or 3% of global annual turnover, whichever is higher. That kind of exposure turns what might seem like a labelling debate into a board-level discussion.
The Guardian investigation — what consumers actually see
While Eurocommerce is lobbying in Brussels, the rest of us are seeing what unlabelled AI advertising looks like in practice.
The Guardian’s investigation, published on 21 June, found concrete examples across Instagram of brands using AI-generated influencers to create content that passes as genuine customer testimony. A photo app called Once posted videos of a woman crying about how much she loved using the app at her wedding. Reality Defenders, a deepfake detection company, identified the woman as AI-generated. Once didn’t respond to requests for comment.
A Dubai-based fashion brand called Ashle posted photographs appearing to show a woman wearing its clothes at a restaurant — except the woman had an extra finger. When The Guardian contacted them, Ashle deleted the images and claimed the removal was because “those designs are no longer part of the collection, not because they were AI-generated.”
NDAs and the 40-60% estimate
The most striking number came from Lisa Mansbridge, a freelance creator who makes AI influencer content for brands under her Mia Metaverse brand. She estimates that 40% to 60% of promotional content from some of the biggest brands is actually being generated through AI.
The catch? “A lot of the creators are under NDA,” she told The Guardian. Brands are paying people to create hyperrealistic digital humans that appear as authentic user-generated content — and then asking those creators to sign non-disclosure agreements so they can’t talk about it.
The consumer group Which? added context: their own investigation found that 70% of people are unable to correctly identify real and fake videos, meaning consumers are frequently being misled without realising it.
The UK’s regulatory vacuum
In the UK, the Advertising Standards Authority’s position is notably different. An ASA spokesperson told The Guardian there is “nothing in our rules that explicitly prohibits brands from posting AI-generated promotional content without disclosing it.” The regulator would only intervene if the content was found to be misleading — but determining whether AI-generated content is misleading requires someone to flag it first, investigate it, and make a judgment.
In the EU, the transparency requirement is proactive — you label it before anyone has to complain. In the UK, the system is reactive — complain first, investigate later. Whether that difference matters depends on how many people actually notice the difference between a real customer and an AI one.
The irony is almost too perfect
The central irony of this story is almost too perfect: the industry association representing the world’s largest advertisers is asking regulators to remove transparency requirements for AI-generated advertising — at the exact moment an investigation reveals how widely that technology is being used to create content that consumers cannot distinguish from reality.
Eurocommerce’s argument is technically coherent. Not every AI-touched advert is deceptive — a bakery using Midjourney to stylise a photo of their sourdough isn’t trying to fool anyone. But a blanket exemption for all advertising would also cover the crying brides, the extra-fingered models, and the NDAs. It would remove the requirement for labelling precisely in the cases where labelling matters most.
The rules start on 2 August. The lobbying is happening now. Whether Eurocommerce gets its exemption would tell us a lot about what the EU thinks transparency actually means.
Sources: The Next Web (19 June), Reuters (19 June), The Guardian (21 June)
