How NO AI ACT works

01The investigative loop

Every case follows the same cycle. There are no one-shot quiz questions: each step asks you to reason about evidence, context of use and effects on people.

1. Scenario — you open the case file: an automated city, an AI system at work, something that doesn't add up. The scenario gives you the context of use: who runs the system, who is affected.
2. Evidence — you examine the exhibits: award decisions, vendor manuals, citizens' complaints, data tables, internal notes, public statements. Each exhibit has a different source and weight; you choose which ones to cite in your report.
3. Context and rules — during the investigation you can re-read the case, consult the archive of rules you have acquired and the case's relevant rule, all read-only: consulting never changes the outcome.
4. A 4-step decision — you classify the risk (prohibited practice, high-risk, transparency obligation, low risk, out of scope), choose the measure, attribute the obligations to the right subject and ground the report on a rationale.
5. Inspection report — the document you built gets stamped with the outcome: compliant, partial, open to challenge or non-compliant, with the main finding and a decision analysis.
6. Decision debrief — a panel explains the reasoning: what to observe in the file, which rule connects the evidence to the decision, how to reason in the next case. It never changes the score.
7. Consequences — the city reacts: the indicators (efficiency, social control, fundamental rights, public trust) move with your decision, and an investigative note tells you the concrete effect.
8. Norm acquired — closing the case unlocks the provision in a simplified educational version: it enters your archive and stays available in later cases.

02Why it is not a quiz

A quiz asks for the right answer. NO AI ACT asks for a defensible decision: the same classification can stand or collapse depending on the evidence you cite and the subject you hold responsible.

• The cited evidence grounds the classification: a report built on the wrong exhibits is open to challenge even when the classification is right.
• The context of use matters more than the technology: the same system can be lawful in one context and prohibited in another.
• Mistakes have visible consequences for the city, not just a score.

0313 cases, from clear-cut to deliberately ambiguous

You start with prohibited social scoring and end with the grey areas: public chatbots, AI procurement, adaptive learning platforms, generative models used downstream. Each file tests a different part of the EU AI Act.

The full case list is on the main page. A single case takes 10–15 minutes; the full paths 60–90 minutes.

04What you receive, and what it is designed to teach

Every closed case gives you three concrete things: the inspection report stamped with the outcome and the main finding, the decision debrief that explains the reasoning, and the acquired norm that enters your consultable archive. At the end of a path, the state of the city tells the cumulative effect of your decisions.

The game is designed to train four habits: classifying risk from the context of use, grounding decisions on cited evidence, telling effective human oversight apart from formal oversight, and attributing obligations to the right subject. It does not cover the entire regulation and does not replace studying the text: the full framing is on the AI Act serious game page.

05And in the classroom?

The loop is built for group discussion too: the inspection report and the debrief are natural starting points for comparing different decisions on the same evidence. Teacher mode adds a local, exportable debrief — no accounts, no student data.

06Learn more

• Why a serious game about the AI Act — the educational rationale of the project.
• For educators — session formats, learning objectives, debrief questions.
• Privacy by design — what the game does not collect, and why.
• Education hub — the complete hub: guides, lesson plans, activities and glossary.
• The AI Act, explained simply — the game's subject matter, in plain language.
• Classroom activities — five ready exercises to pair with the cases.