The “AI” Label Enters the Frame
From August 2, the EU’s AI Act brings new transparency obligations for the use of artificial intelligence in advertising and commercial communications. The changes concern almost everyone who touches a brief today: advertisers, agencies, creative teams, marketers. The framework does not ban AI or restrict innovation. It asks for something simpler and more demanding at the same time: to say when something was made by a machine.
Let’s take it calmly, because more panic than information has gathered around the date.
What Article 50 actually requires
The provision that concerns us is one: Article 50, “Transparency Obligations for Providers and Deployers”. In plain Greek, it asks for four things.
- The user must know when they are interacting with an AI system, unless it is obvious.
- Synthetic content, and especially deepfakes, must be labelled as artificially generated.
- When AI is used for emotion recognition or biometric categorisation, the user must be informed.
- When content is generated or altered with AI, this must be disclosed, with an exception (among others) for material that is obviously artistic or satirical.
In practice: the AI-generated face in a spot, the synthetic voice in a radio ad, the “retouch” that is no longer retouch but re-creation — all of it falls under labelling. The essence lies in the “how”.
The “how”: two letters
How do you “label” an advertisement? The EU’s Code of Practice (together with the overview prepared by EASA for the advertising market) gives the practical answer, more tangible than it sounds. The core is two letters: an icon or equivalent indication with the acronym “AI”. You decide the size and style, as long as the indication is clear, distinct and accessible. The common European icon exists, but its use remains optional.
Placement has rules with common sense: the indication appears no later than the first exposure to the content, embedded on top of it, visible for long enough, without getting lost among other overlays.
- In a deepfake video it appears at the beginning and again after every interruption (an ad break, for example).
- In audio-only it requires a short audio disclaimer at the beginning.
- For artistic or creative works there is flexibility: the indication can appear even next to the frame, not inside it, as long as it remains clearly visible.
The distinction many overlook
Care is needed here. The Code of Practice is optional (guidance, not law) and is still under assessment by the Commission and the AI Board. The transparency obligations of Article 50, by contrast, are a legal obligation with direct and automatic effect in Greece. Put simply: you may or may not follow the Code’s “how”; the “that” of Article 50 is not at your discretion.
And one last clarification, because it causes confusion. The Regulation splits roles into providers (those who build the models) and deployers (those who use them). EASA’s overview is clear: the first part of the Code concerns providers and does not directly touch the advertising market; the second speaks to deployers and focuses on deepfakes.
So, in the overwhelming majority of cases, an agency or an advertiser is a deployer and its primary obligation is labelling. The line shifts only if you start building and making available your own AI tools—then you also wear the provider hat.
Article 50 changes the terms: transparency about what is real and what is synthetic becomes a prerequisite, not a polite gesture.
For years, advertising sold authenticity without having to prove it. Two letters on a frame may seem insignificant, yet they are the point where the industry decides whether it will see transparency as a burden or as an advantage.
