Starting today, every chatbot you talk to and every suspicious video in your feed is supposed to come with a label if AI made it. That's not a company being nice. That's EU law now, EU-wide, all at once. This is one of those moments where Brussels bureaucracy quietly rewrites how the internet is allowed to look — and almost nobody's talking about it yet.

What actually changes today

Sunday marks the start of the AI Act's transparency rules, which require companies to disclose when chatbots, images or video have been generated or manipulated by artificial intelligence, per euronews. Translation: if you're chatting with a bot, it has to tell you it's a bot. If a video or photo has been synthetically generated or doctored by AI, that has to be flagged too. This isn't a national law one country adopted early — it's a provision of the EU's AI Act, meaning it lands in all member states on the same day, no separate national vote required to make it bite.

Why now, specifically? Because the AI Act was built with a staggered rollout — different obligations switch on at different dates depending on how risky regulators judged them. Transparency duties for deepfakes and AI chat interfaces were always going to be an early tier, and that tier's clock just ran out.

The bigger arc: rules racing content

This is the latest chapter in a pattern that's been running for years now — regulation trying to catch up to generative AI faster than generative AI improves. Deepfakes went from a novelty to a genuine disinformation and fraud vector well before any government had a legal answer ready. The AI Act's labelling duty is the EU's attempt to close that gap with a blunt instrument: don't ban the technology, force it to identify itself.

Why experts aren't popping champagne

Here's the catch, and it's a big one. According to euronews, experts warn technical gaps and a lack of common standards could undermine enforcement of the new labelling requirement. Basically: the law says "label it," but there's no single agreed-upon technical way to do that across every platform, file type and piece of software producing synthetic media. Without a shared standard, labels can be inconsistent, strippable, or simply skipped by smaller or less compliant operators.

Open questions nobody's answered yet

The source reporting doesn't spell out exactly which authority polices this day to day, what the penalties look like in practice, or how a rule with EU-wide legal force gets checked against millions of individual pieces of content. That gap is not a minor detail — it's arguably the whole ballgame. A transparency duty only matters if someone is actually verifying compliance and someone pays a real cost for ignoring it. Hypothesis: enforcement will be uneven across the bloc at first, with larger platforms complying visibly while smaller AI tools and non-EU services lag or route around the requirement entirely. Supporting this: the euronews reporting itself flags a lack of common technical standards as a core weakness. Against this: EU-wide regulations before this one — from GDPR onward — did eventually converge toward more consistent enforcement once precedent and guidance built up, so day-one gaps aren't necessarily permanent.

Should you care?

Yes, but keep your expectations calibrated. The legal obligation to label AI content is real and now enforceable across all 27 member states. Whether that changes what actually shows up unlabeled in your feed next week is a separate, unresolved question — and per the experts cited above, the honest answer right now is probably not much, not yet. Watch for the first enforcement actions and whether Brussels moves to standardize labelling technically, because that's what will tell you if this law has teeth or just good intentions.