The EU AI Act for teachers

01Why teach the AI Act at all

Your students already live with recommender systems, chatbots and automated screening. The AI Act gives a classroom three rare things: a concrete vocabulary for discussing those systems, a European civic frame, and real controversies that reward evidence-based argument rather than opinion.

It fits civics and digital citizenship, computer science and society modules, law and technology courses, media literacy, and professional development alike.

02Three entry points that work

• Start from a story — a social-scoring scenario (who gets housing priority? who decides?) raises the "prohibited practices" question naturally. Then name the regulation that answers it.
• Start from the students' tools — a chatbot they use daily leads to transparency obligations; a school's adaptive platform leads to high-risk systems.
• Start from the game — one 15-minute case of NO AI ACT gives the whole class a shared experience to dissect.

03Before, during and after

If you use the game, the practical session logistics (formats, teacher mode, pairing) are covered in the game's educator guide. In short: play the case yourself first; facilitate without revealing a "right answer"; debrief by comparing two different reports on the same evidence.

Without the game, the same structure works with the paper-based activities: scenario first, evidence second, classification debate third.

04Learning objectives you can actually assess

• Explain the AI Act's risk-based approach in their own words.
• Sort unfamiliar AI scenarios into risk categories and justify the sorting with specific facts.
• Distinguish effective human oversight from a rubber stamp, using an example.
• Name who carries obligations (provider vs deployer) in a concrete case.
• Formulate one question a citizen could ask about an automated decision that affects them.

05Discussion prompts

• Should a city be allowed to score its citizens' "reliability"? Does your answer change if the score only affects queue priority?
• When must you be told you are talking to a machine — and when does it not matter?
• Is a warning label ("AI-generated") enough? What would meaningful transparency look like?
• Your school buys an adaptive learning platform. Which questions should be answered before it is used on you?
• Who should be responsible when a general-purpose model gives harmful advice inside a public service?

06Facilitation notes

• Protect the ambiguity. The best cases are the grey ones; resist resolving them too early. "It depends — on what?" is the productive question.
• Demand evidence, not vibes. Push every claim back to a fact of the scenario: which data, which effect, which safeguard.
• Rotate roles. Let students argue as provider, deployer, affected citizen and inspector; the obligations question becomes tangible.
• Anchor to the real text at the end — one look at EUR-Lex shows both the distance and the connection between simulation and law.

07Limitations to keep in mind

• These materials and the game are educational simplifications — not legal training, not compliance guidance, not an official EU resource.
• The game covers core concepts through 13 cases; it does not cover the entire regulation.
• Educational effectiveness has not been empirically validated; treat the materials as a starting structure, not a proven method.

08Materials

• Lesson plan — introduction to the AI Act (60–90')
• Lesson plan — the risk-based approach (60')
• Lesson plan — transparency and users (50–60')
• Five classroom activities · Glossary · Plain-language guide · Privacy-conscious learning
• Play NO AI ACT (free, browser, no accounts)

09What actually applies to a school

This page is mostly about how to teach the AI Act. But the question school leaders ask most often is a different one: what falls on us? Here is the simplified educational answer — formal compliance for an individual institution still needs whoever carries the legal responsibility for it.

Three clarifications that prevent needless alarm in a staff meeting. First: the vast majority of digital tools used in schools are not high risk — a spell-checker or a file-sharing platform stay at minimal risk. Second: the "high risk" label prohibits nothing, it requires governing. Third: the heaviest obligations sit with whoever provides the system; the school answers as a deployer — for how it is used, who supervises the outcomes, and what families are told.

The full calendar, including the dates most often confused with one another, is on the application timeline.

10The questions that actually come up

The fastest way to get a staff room reasoning about these distinctions is to have them walk through the distinctions rather than hear them explained: the game contains two cases set in schools — emotion detection in class, and an adaptive learning platform — built precisely around the boundary between prohibited, high risk and legitimate use.