EU AI Labeling Rules Now in Force

A new set of EU AI labeling rules took effect on 2 August 2026, giving everyone in Germany a clearer right to know when they are looking at machine-made content or talking to a machine. The rules come from Article 50 of the EU AI Act, the bloc’s regulation on artificial intelligence known in German as the KI-Verordnung. Under this AI labeling requirement, companies that build or professionally use AI systems must mark AI-generated or manipulated material and make chatbots plainly recognizable as software rather than people.

What the new AI labeling rules require

Article 50 sets transparency duties for two groups: the providers who develop AI systems and the professional users, called deployers, who put them to work. When you interact with a chatbot or a voice assistant, the system must make clear that you are dealing with a machine, unless that is already obvious from the situation. This closes a gap where a customer could believe they were chatting with a human agent.

The rules also cover generated media. AI-produced or AI-edited text, images, audio and video must be marked as such. Deepfakes, meaning realistic but fabricated images, clips or recordings of real people, have to carry a visible label so viewers are not misled. Generated media must additionally be marked in a machine-readable way, for example through digital watermarks or signed metadata that other software can detect even after the content is shared.

Who enforces AI labeling in Germany

In Germany the supervisor is the Bundesnetzagentur, the Federal Network Agency that already oversees telecoms and energy markets. It is responsible for checking that companies apply the transparency standards correctly and for handling complaints when they do not.

The penalties are significant. According to boerse-express and Handelsblatt, breaches of the Article 50 transparency duties can draw fines of up to 15 million euros or up to 3 percent of a company’s total worldwide annual turnover, whichever is higher. A separate and higher tier of up to 35 million euros or 7 percent applies to outright prohibited AI practices, not to the labeling duties, so most businesses face the 15 million or 3 percent ceiling.

What is exempt from the AI labeling duty

The rules are not meant to cover every private snapshot or joke. Purely private, personal use of AI falls outside the labeling duty. So do clearly artistic, creative, fictional or satirical works, where the disclosure must be arranged so it does not spoil the enjoyment or presentation of the piece. In practice a film using AI effects or an obvious parody can be marked discreetly rather than plastered with a warning.

There is also a grace period for systems that were already on the market. Generative AI tools launched before 2 August 2026 have until 2 December 2026 to meet the machine-readable marking standard, while brand-new systems must comply immediately. The European Commission published guidance and a voluntary code of practice in July 2026 to help companies interpret the details.

Your new right to know

For everyday internet users, the clearest benefit is transparency. When you contact a company’s support line and reach a chatbot, it should now tell you it is a machine. When you see a striking image or video of a public figure, a visible AI label, or its absence, becomes part of judging whether it is real. Combined with machine-readable watermarks, this is meant to make manipulated media easier to spot before it spreads.

These consumer protections sit alongside other recent EU moves to rein in large tech platforms, such as the penalties covered in our report on EU fines against Google under the Digital Markets Act. The AI labeling rules add a transparency layer on top of those market rules, and they strengthen the broader framework explained in our guide to consumer protection laws in Germany.

What this means for expats running a business

If you are self-employed or run a company in Germany and you use AI in customer-facing ways, the AI labeling duty applies to you as a professional user, not just to the big AI developers. A chatbot on your website must reveal that it is automated, marketing images generated by AI should be labeled, and any synthetic audio or video of a person needs a visible mark. Reviewing where AI already appears in your customer communication is the practical first step, and the December deadline gives some breathing room for tools you adopted before August.

Smaller businesses do not need to panic, but they should not ignore the rules either. The safest approach is to document where you use AI, add clear notices where customers interact with it, and keep an eye on the Bundesnetzagentur guidance as it develops. Getting the basics right now is far cheaper than facing a complaint once enforcement ramps up.