Education
Preparing lessons, giving feedback, lowering barriers: what is possible in schools and universities and what the AI Act expressly prohibits.
What this is about
| Task | Classification |
|---|---|
| Preparing and differentiating teaching material | unproblematic |
| Rendering texts in plain language | unproblematic, with review |
| Generating practice exercises | unproblematic, with review |
| Giving feedback on drafts | unproblematic, where not grading |
| Marking assessments | high risk, Annex III |
| Deciding on admission | high risk, Annex III |
| Recognising emotions | prohibited |
The productive area
- 01
Differentiating material
The same content at three levels of difficulty, with examples from different contexts. That is work which otherwise does not get done.
- 02
Lowering barriers
Plain language, translation, read-aloud, subtitles. The benefit for learners with additional needs is immediate.
- 03
Formative feedback
Comments on a draft, without a grade. That relieves pressure without touching assessment.
- 04
Only then consider assessment
And then with full awareness of the duties attached.
Handling AI in submitted work
- A written rule about what is allowed: research yes, drafting no, or the other way round. Without a rule, any sanction is contestable.
- Task formats that assess process rather than product: interim stages, oral explanation, reference to class work.
- No detection tools as evidence. Their error rate for non-native speakers is documented as high, and a false accusation does considerable harm.
- Disclosure duty rather than prohibition: learners state where and how they used AI.
Annex III in education
High risk includes systems for:
- Access, admission or assignment to educational institutions.
- Evaluating learning outcomes, including steering the learning path from them.
- Assessing the appropriate level of education for a person.
- Monitoring and detecting prohibited behaviour during examinations.
The last point hits camera-based exam invigilation directly. Such systems are not prohibited but attract the full set of high-risk duties, including a fundamental rights impact assessment for public bodies.
Deployer duties
| Duty | Concretely |
|---|---|
| Human oversight | A named person able and permitted to review the output |
| Information | Those affected learn that a high-risk system is in use |
| Logging | Retention of the automatically generated logs |
| Input data | Ensuring it is suitable and representative for the purpose |
| AI literacy | Adequate training of the staff deploying it |
See Duties by role and AI literacy.
Data protection for minors
- Consent by children under 16 requires authorisation by a guardian; some member states lower the age to 13.
- In schools, consent is rarely the right basis, because voluntariness within a compulsory relationship is doubtful. National education law takes precedence.
- Attainment data is particularly sensitive and must not feed profiles beyond the educational purpose.
- Tools transmitting learner usage data to providers outside the EU are not deployable without a checked basis.
Entry sheet: education
FREE ACCOUNT
Entry sheet for education
This sector's tasks by risk class, the points to settle beforehand, and the metrics the benefit shows up in.
Checklist3 items