AI Act and Education: What German Schools Need to Know
Education AI is a tale of two categories. Learning tools that help students study? Generally low risk. AI that decides who gets admitted or how students are assessed? High-risk, with serious compliance obligations.
The distinction matters enormously for German educational institutions.
Admissions and Assessment Are High-Risk
The AI Act explicitly lists educational AI as high-risk when it’s used for admissions decisions or assessing learning outcomes that affect educational paths. University admissions algorithms, automated grading that determines advancement, AI that streams students into different educational tracks—all high-risk.
This means risk management, bias testing, transparency, human oversight, and documentation. For German institutions bound by state education laws, this layers on top of existing requirements.
Learning Tools Are Mostly Fine
Adaptive learning platforms, AI tutors, language learning apps, research assistants—these are generally lower risk. They assist learning without making gatekeeping decisions. Use them, document them, but don’t expect heavy compliance burdens.
The line is whether the AI determines access or outcomes. Helping a student learn? Low risk. Deciding whether they pass? Higher risk.
Proctoring and Monitoring
Remote exam proctoring with AI raises flags. If it’s monitoring student behavior, detecting “suspicious” activity, or making determinations about cheating, you’re in sensitive territory. Emotion recognition is prohibited. Any biometric processing needs careful justification.
What This Means Practically
Educational institutions should map their AI systems by function. Admissions and assessment AI needs the full high-risk treatment. Learning support tools need basic documentation. Proctoring systems need careful review for prohibited practices.
How Compound Law Helps
- AI system inventory and classification
- Admissions AI compliance frameworks
- Assessment system documentation
- Proctoring policy review
- State education law integration
Frequently Asked Questions
Is adaptive learning software high-risk? Usually no. If it helps students learn without determining grades or advancement, it’s lower risk.
What about automated grading? Depends on consequences. Grading that affects progression, graduation, or admissions is high-risk. Practice quiz scoring isn’t.
Do private schools have the same obligations? Yes. The AI Act applies based on what the AI does, not who operates it.
