Yesterday, Pro TV asked me about the European Union's new system for labelling AI-generated content — a set of marks that becomes mandatory to apply this August.
The EU has published a set of labels companies can attach to photos and videos made with artificial intelligence. They're free to use, and they come in three levels: a basic mark for content where only a part has been altered — a voice swapped, an element edited; an "AI Generated" mark for fully synthetic material, the deepfake clips of politicians or events that never happened; and a middle mark for clips that are partly real, partly AI. The label designs themselves are voluntary, but from August, flagging AI-generated or manipulated content becomes an obligation across online platforms and media.
My take on camera was that the direction is right but the execution is everything.
"The idea of labels that flag when something is an AI product is a good one. What worries me is the implementation — without a real mechanism to enforce it, the labels end up useless. Whoever wants to deceive simply won't apply them."
A label only helps with honest content. The people most likely to mislead — the ones producing political deepfakes or scams — are exactly the ones who will skip the mark, and a voluntary badge gives them nothing to fear. So the rule is only as strong as what backs it: detection that can spot unlabelled AI content, and consequences when the obligation is ignored. Without that, we get a false sense of safety — people learn to trust the absence of a label, which is precisely what a bad actor leaves off.
That's why I think the labels have to come with public education, not replace it. The durable defence isn't a sticker on a video — it's people who know that any image, voice, or clip can now be synthetic, and who pause before they believe and share. The label is a helpful prompt; the habit of doubting is the real protection.