Europe’s AI labels are now law

If you chatted with a customer-service bot this week and it told you upfront that it was artificial intelligence, that small disclosure was not a courtesy. Since 2 August, it is European law. Article 50 of the EU AI Act requires AI systems and synthetic content to identify themselves as such (EU AI Act, Commission summary). The rule does not ban anything. It makes the synthetic layer of the internet slightly more legible.

What actually has to happen

Three situations, three obligations. A chatbot must say it is AI at the start of the conversation. A realistic synthetic image, video, or audio clip must carry a visible label. And AI-generated text on matters of public interest — election explainers, public-health notices, policy summaries, not product blurbs — must be disclosed as machine-made, unless a human editor reviewed it and takes editorial responsibility (Article 50 guidance).

The system works on two tracks. The visible layer is what users notice: a tag on a deepfake video, a disclosure beside a chatbot window. The hidden layer is metadata embedded in the file itself, a record of where the content came from, readable by detection tools even when a human eye sees nothing unusual (Code of Practice). Think of it like nutritional labels on food: one tells the shopper, the other lets regulators test what is actually inside.

Obligations split depending on where you sit in the chain. OpenAI or another chatbot vendor has one set of duties: design the system so users know they are talking to a machine, and bake origin-tracking metadata into every synthetic output. The bank, newsroom, or political campaign using that tool has a different duty: tell audiences when they encounter deepfakes or AI-written public-interest material (Stibbe). A notice buried in terms and conditions does not count. The Commission says disclosure must happen at the start of the first interaction, clearly and accessibly (AI Act FAQ).

If you post a personal AI meme, you are not the target. The rules apply to professional and commercial use, not purely personal activity (EU AI Act).

Complaints now have somewhere to go

Brussels wrote the rules, but national authorities enforce them. Germany offers the clearest picture so far: the Bundesnetzagentur, the federal network agency, became the default complaint office on 2 August, and citizens can file reports free of charge through an online form (Bundesnetzagentur, German government). Ireland has set up a national AI Office for coordination (Department of Enterprise Ireland). Other member states are building equivalent structures.

Fines for transparency violations can reach up to €15 million or 3% of global annual turnover. That is lower than the €35 million / 7% ceiling some coverage has cited; the higher tier covers prohibited AI practices, not labelling failures (EU AI Act, Commission enforcement note).

The label is a starting point, not a solution

Bad actors will ignore the rules. Content generated before the deadline carries no retroactive obligation. And a missing label does not prove something is real, because the technical standards for origin-tracking metadata are still maturing (Two Birds). Courts will settle the boundary questions that remain: what counts as "obviously AI"? How prominent must a label be inside a specific app?

This is also not the entire AI Act arriving at once. Generative AI systems already on the market get until 2 December 2026 to add machine-readable markings. Broader high-risk rules covering employment, education, and credit scoring take effect in December 2027 for standalone systems and August 2028 for those embedded in regulated products (AI Omnibus, RegDossier).

The system is incomplete and tamperable, but it gives users and regulators something to point to for the first time. Whether labels change behaviour, build trust, or become background noise that people learn to scroll past is the experiment that starts now.