For educators · Trainers · Universities
NO AI ACT for educators
A ready-to-run classroom activity for the EU AI Act: your students investigate realistic AI systems, argue over the evidence and defend a risk classification. No accounts, no student data, no technical setup.
01Who it is for
- Upper-secondary schools — digital citizenship, rights, informed civic education.
- Universities and PhD programmes — law and technology, AI governance, AI ethics, legal informatics, media studies.
- Professional training — awareness of risk, transparency and accountability for people working with AI systems in companies or public administration.
02Learning objectives
After a session, learners should be able to:
- distinguish the AI Act's risk categories (prohibited practices, high-risk, transparency obligations, low risk) on concrete cases;
- recognise that risk depends on the context of use and the effects on people, not on the technology as such;
- connect a decision to specific evidence and tell effective human oversight apart from a merely formal one;
- reason about who carries the obligations: provider, deployer, authority.
NO AI ACT supports reflection and discussion: it is not a complete AI Act course or a certification, and it does not replace studying the legal text.
03Three session formats
One case, one discussion
5' introduction, 15' on a clear-cut case (e.g. "The city of scores"), 10' comparing outcomes: who decided what, citing which evidence? Ideal as the opener of an AI Act lesson.
Clear case + ambiguous case
Two cases in pairs or small groups: one clear and one deliberately grey (e.g. "Civic credit"). 20' guided debrief on the differences: why does the same logic lead to different outcomes?
Full path + teacher mode
A 4+ case path with teacher mode on: at the end, export the local (anonymous) debrief and discuss the city "ending" produced by the group's decisions.
04Before, during and after the lesson
Preparation (10 minutes of your time)
Play the case you plan to use: one run is enough to anticipate the discussions. Decide whether to enable teacher mode and whether to play in pairs (recommended: negotiating the classification is itself the exercise). Nothing to install — a browser is all you need.
In a classroom, turn the music off and keep the sound effects: the cues that say "you cited this exhibit" or "you can proceed" help students orient themselves, while a continuous music bed on projector speakers is tiring. They are two separate switches in SETTINGS, next to the CRT filter — worth turning off on a projector — and the text speed, useful if you read aloud. The whole game can also be played from the keyboard.
Facilitation
Launch the case without revealing a "right answer". Walk between groups listening to how they justify citing evidence. Stop the class once, halfway: "who changed their mind after an exhibit? which one?". Let the inspection reports arrive with diverging opinions.
Debrief and connection
Compare two different reports on the same case: same evidence, different outcomes — why? Use the game's decision debrief as a scaffold, then connect to the regulation's real text to show the distance between a simplified simulation and the norm.
05Classroom use cases
- Civic/digital education (upper secondary) — a clear-cut case (social scoring) as the entry point to digital rights and automated decisions: who decides? who answers for it? who can appeal?
- University law or AI governance seminar — the ambiguous cases (public chatbot, EdTech, GPAI) as legal-qualification exercises: the class defends opposite classifications with the same evidence, then compares with the actual articles.
- Professional development / teacher training — the EdTech case ("The profiled classroom") speaks directly to people working in schools: when does an adaptive platform stop being a helper and become a system to govern?
- Independent work / flipped classroom — assign 1–2 cases to play at home and spend class time only on comparing reports.
Ready-made materials from the education hub: the introductory lesson plan (60–90'), the five structured activities, the AI Act for teachers guide for framing it in the curriculum, and the glossary to hand out before playing.
06Suggested debrief questions
- Which piece of evidence changed your mind, and why?
- Was the system the problem, or the way it was used?
- Was the case's "human oversight" effective or merely declared? How could you tell?
- Who did you hold responsible? Who else could have answered for it?
- If the same system had been used in a different context, would the classification change?
- What would have made the harm preventable or governable?
07AI literacy, not only the AI Act
Although the setting is the European regulation, the habits the game trains are transversal to AI literacy and digital citizenship: asking which data a system uses, who it affects, who answers for it and how an automated decision can be challenged.
In that sense NO AI ACT can support AI literacy and responsible-AI programmes even where the AI Act is not the lesson's main goal: the regulation provides the structure, but the skill being exercised is reading an automated system critically. The full rationale is on the AI Act serious game page.
08Privacy: what does NOT happen in class
The game requires no accounts and collects no student data. Progress stays in each player's browser; the teacher-mode debrief export is local and anonymous. No results are ever sent to a server.
Full details: Privacy by design.
09Limitations to know
- It is an educational simplification of Regulation (EU) 2024/1689: a third-party legal review is recommended before formal institutional use.
- It works best on desktop or landscape tablets; portrait phones show a rotate-your-device guard.
- Educational effectiveness has not yet been empirically validated: the project is in its classroom-testing phase. This version embeds no external feedback forms.
10Educator FAQ
Do we need to install anything or create class accounts?
No. The game runs in the browser with no registration and no accounts: just share the link. Each student's progress stays in their own browser.
Are students tracked while they play?
No. Decisions, scores and reports are never sent to any server; the teacher-mode debrief export is local and anonymous. Details on the privacy by design page.
How much time does a meaningful activity take?
A single case takes 10–15 minutes of play: with an introduction and discussion, 30 minutes is enough for a complete activity. The 60 and 90-minute formats add a second, ambiguous case and the teacher-mode debrief.
Does it cover the whole AI Act?
No, and that is not the goal: it covers the load-bearing concepts (prohibited practices, high-risk systems, transparency, human oversight, deployer responsibility) through 13 cases. It is a reasoned introduction, not a complete course or a compliance tool.
Is it an official EU tool?
No. It is an independent, open-source educational project, not affiliated with the European institutions. It simplifies the regulation for teaching purposes and is not legal advice.
Can I adapt the materials or propose a collaboration?
Yes: the code is GPL-3.0-licensed and the educational content is CC BY-SA 4.0. For teaching or research collaborations, open an issue on GitHub.