The EU AI Act’s Labelling Rule Now Applies: What It Changes for Schools, and What It Doesn’t

Since 2 August 2026, EU law has required AI-generated content to be marked and deepfakes to be disclosed. Almost everything written about that rule is for companies. Very little of it says what a school should do with it, or what the rule leaves untouched.

What Article 50 says

The rule is Article 50 of the EU’s AI Act, and it puts duties on two groups.

Providers — the companies that build and sell AI systems — must make sure people know when they’re talking to one. They must also mark synthetic audio, image, video and text “in a machine-readable format” so that it’s “detectable as artificially generated or manipulated”.

Deployers — the Act’s word for anyone using an AI system in a professional capacity — must disclose deepfakes. They must also label AI-generated text that’s published to inform the public on matters of public interest, unless a person has reviewed it and someone takes editorial responsibility.

The Commission has published a voluntary code of practice on how marking and labelling should work, and three optional icons for labelling. One date matters if you’re checking a tool: generative systems already on the market before 2 August 2026 have a grace period for the marking duty until December 2026.

The definition worth teaching

The Act defines a deepfake in one sentence: “AI-generated or manipulated image, audio or video content that resembles existing persons, objects, places, entities or events and would falsely appear to a person to be authentic or truthful.”

Three things in that sentence are worth a lesson. It covers objects, places and events, not only people. It covers audio. And the test is whether the content would falsely appear authentic — not whether anyone meant harm.

What the rule doesn’t do

It doesn’t reach a student at home. A deployer is someone using an AI system “except where the AI system is used in the course of a personal non-professional activity”. A student who makes a fake clip of a classmate on their own phone isn’t a deployer under this regulation. Other law may well apply to what they’ve done, but not this one.

It doesn’t make unlabelled content real. The Commission says deepfakes made before 2 August 2026 don’t have to be labelled retroactively. Older tools have until December. And a machine-readable mark is there for software to read, not for a student scrolling past. A label tells you something. The absence of one tells you nothing.

It isn’t the law on harassment. Sexualised deepfakes of real people are dealt with in a separate directive from 2024.

Where a school is on the other side of the rule

A school isn’t only an audience for this rule. A school that uses an AI tool for its own work is using it professionally, which is what the Act means by a deployer.

Most of what a school generates sits well outside the two deployer duties. A diagram, a worksheet illustration or a draft quiz isn’t a deepfake and isn’t news. Some of it comes closer: a generated “photograph” of a historical event for a history lesson, a cloned voice of a real person, an item in the school newsletter that AI wrote and nobody edited.

Whether a particular use creates a legal duty is a question for your authority’s legal adviser, not for a blog post. Whether you label it is a decision you can take today.

What a school can do this term

None of this needs a budget:

• Agree one line. Decide the wording the school puts on any image, audio or video it generates or alters with AI — in lessons, on the website, on social media — and use the same line everywhere.
• Teach the absence problem. Students tend to read “no label” as “real”. Show them why that doesn’t follow.
• Put the definition on the board. Give students the Act’s one sentence and a handful of examples, and ask which ones meet it. The disagreements are the lesson.
• Ask your tools. For each generative tool the school uses, find out whether it marks its output yet.

We think checking what you see has become a basic school skill, in the way reading a source used to be, and that’s why this rule matters to us more than its length suggests. We work through it with teachers in three courses: Media Literacy, which reads Article 50 for what it requires and what it leaves out; the AI Literacy Course for Teachers, on how these systems work and where they go wrong; and Digital Storytelling, where the rights session covers what a class should label before it shows its own work outside the classroom. You can start a registration once you know which one fits.