EU AI Act for Education & EdTech: High-Risk Rules for Learning AI
AI now decides who gets admitted, how exams are marked and whether a candidate is flagged for cheating — decisions that shape people's futures. The EU AI Act (Regulation (EU) 2024/1689) reflects this by placing much of education and vocational-training AI in the high-risk tier under Annex III. EdTech vendors and institutions alike are in scope whenever the output is used within the EU.
Is it in scope?
Annex III lists education and vocational training as a high-risk area, covering AI used to determine access or admission to institutions, to evaluate learning outcomes, to assess the appropriate level of education a person will receive, and to monitor and detect prohibited behaviour during tests. Automated exam scoring and proctoring are therefore squarely in scope. Note also that emotion recognition in educational settings is prohibited under Art. 5, which constrains certain proctoring and engagement-analysis designs.
Typical AI use cases
- Automated admissions and enrolment decisions
- AI exam scoring and essay grading
- Remote proctoring and cheating detection
- Adaptive placement and streaming of learners
- Learner performance evaluation and prediction
- Automated eligibility assessment for programmes
Risk classification
Admissions, evaluation and proctoring systems are high-risk under the Annex III education listing. The EdTech company that builds the grading or admissions engine is the provider, responsible for Art. 9–15 — with data governance (Art. 10) and human oversight (Art. 14) especially critical given the impact on students. The school, university or exam board deploying it is a deployer under Art. 26 and must ensure meaningful human review and inform affected learners. Proctoring tools must also be checked against the Art. 5 prohibition on emotion recognition.
Obligations to prepare for
FAQ
Is AI essay grading high-risk?
Yes. Evaluating learning outcomes and scoring assessments falls within the Annex III education category, so automated grading systems are treated as high-risk under the AI Act.
Can we use AI proctoring that reads student emotions?
No. Emotion recognition in educational settings is prohibited under Art. 5. Proctoring can flag defined behaviours, but inferring emotional states of learners is off-limits.
We're a school buying EdTech, not building it — are we regulated?
Yes, as a deployer under Art. 26. You must use the system per the provider's instructions, ensure genuine human oversight of automated decisions, and inform the students affected.
Run the free 2-minute risk checker — no sign-up.