The EU AI Act, explained for brand marketers

Most brands hear "EU AI Act" and assume it means slapping a disclaimer on every image and hoping that's enough. It's not that simple, and it's not that hard either.

Here's what actually matters if you're commissioning AI-generated or AI-assisted imagery for a campaign running in the EU - and this isn't a future deadline to plan around. The labelling obligation is already active.

Transparency is about the output, not the tool.

The regulation isn't concerned with which software made an image. It's concerned with whether a viewer can tell the image was artificially generated or manipulated when it matters for them to know. That distinction shapes almost everything else.

Not every AI-touched image needs a label.

A product shot that's had AI-assisted retouching, colour correction, or background cleanup isn't the same as a fully synthetic scene with no photographic origin. Minor technical use generally sits outside disclosure requirements. A wholly generated image used in a context where realism could mislead a viewer is a different case entirely.

What counts as a "deepfake," and what doesn't.

Under the Act, a deepfake is AI-generated or manipulated content that resembles a real person, object, place, entity, or event closely enough that it could be mistaken for authentic. That's a narrower, stricter category than AI imagery in general, and it carries its own explicit disclosure duty. A generated product-in-context scene, an imagined location, or a stylised campaign world isn't a deepfake, since nothing in it is presented as a real, identifiable person or event. A generated video of a real spokesperson saying something they never said, or a fabricated "real" customer testimonial, is. The test isn't how realistic the image looks, it's whether it depicts something specific and real that didn't actually happen that way.

Context changes the obligation.

The same image can sit on either side of the line depending on where and how it's published. A generated lifestyle scene used as an obvious stylised campaign visual is read differently than one presented as if it were a real customer photo or testimonial.

There's now an official way to label it.

The European Commission has published a standard set of icons for marking AI-generated content, designed to help viewers recognise it clearly and consistently. Using the icons themselves is optional, but the underlying disclosure obligation is not, and has applied since August 2026. The icon alone doesn't create compliance on its own. You can see the official icon set here: EU icons for labelling AI-generated content.

"Audit trail" isn't optional caution, it's practical protection.

Regardless of what a specific campaign requires, keeping a clear record of what was shot, what was generated, and what was blended protects you if a client, platform, or regulator ever asks. It's the difference between answering a question in a minute and reconstructing a project's history under pressure.

Enforcement is still catching up to the rule.

The obligation is live, but guidance continues to develop, and platform-level policies (Meta, Google, TikTok) often move faster and further than the regulation itself. What's compliant today may need a stricter label next year, not because the law changed, but because a platform's own policy did.

What we do with this in practice

Every project we deliver comes with a clear internal record of method: shot, generated, or hybrid, down to the individual asset. Where disclosure applies, whether that's general AI-content labelling or the stricter deepfake disclosure, we handle it as part of delivery, including the official EU icons where appropriate, not as an afterthought. Where it doesn't, you still have the record if you ever need it.

This is general information, not legal advice. If AI disclosure is a compliance-critical question for your brand, your legal team should review the specifics of your campaign and jurisdiction.