AI Act · Edtech

AI Act for Edtech: Assessment, Proctoring and Children's Data

Annex III point 3 of the AI Act classifies AI for access, assessment, level and educational proctoring as high-risk. Art. 5(1)(f) prohibits emotion recognition in education. Learn the obligations.

Is your educational AI high-risk under the AI Act?

The AI Act (Regulation (EU) 2024/1689) classifies four AI use cases in education as high-risk (Annex III, point 3):

  • 3(a): AI to determine access or admission to educational or vocational training institutions
  • 3(b): AI to evaluate learning outcomes, including where outcomes steer the learning process (adaptive learning)
  • 3(c): AI to assess the appropriate level of education
  • 3(d): AI to monitor and detect prohibited behaviour of students during tests (proctoring)

Emotion recognition prohibition: Art. 5(1)(f) prohibits AI systems inferring emotions of natural persons in the workplace and educational institutions, except for medical or safety reasons. This prohibition has been in force since 2 February 2025.

Emotion recognition as high-risk: Where emotion recognition is not prohibited, it is classified as high-risk under Annex III point 1(c).

Remote biometric identification: Remote biometric identification (Annex III point 1(a)/(b)) may be high-risk or prohibited depending on the context and inferred attributes.

Children's data (GDPR + LOPDGDD): Art. 8 GDPR sets the conditions for children's consent in information society services. The LOPDGDD Art. 7 sets the minimum consent age at 14 in Spain. Children under 14 cannot give valid consent; parental consent is required.

  • Annex III point 3 classification — We determine whether your educational AI is high-risk
  • Art. 5(1)(f) audit — We verify your AI does not infer emotions in education (prohibited)
  • Children's data analysis — We identify the legal basis and parental consent (LOPDGDD Art. 7)
  • Educational DPIA — We carry out the Art. 35 GDPR Impact Assessment for children's and biometric data
  • Technical documentation — We prepare the Art. 11 documentation for high-risk educational AI

AI Act obligations for high-risk Edtech

If your educational AI is high-risk (Annex III point 3), the AI Act obligations are:

For providers (developers):

  • Risk management system (Art. 9)
  • Training data quality (Art. 10), critical to avoid bias in assessment
  • Technical documentation and record-keeping (Art. 11)
  • Transparency and information for deployers (Art. 13)
  • Human oversight (Art. 14)
  • Accuracy, robustness and cybersecurity (Art. 15)
  • Conformity assessment (Art. 6 + Annex III)
  • Registration in the EU database (Art. 49)

For deployers (educational institutions):

  • Designate staff with competence for human oversight (Art. 26(2))
  • Carry out FRIA (Art. 27) before deploying the system
  • Notify serious incidents (Art. 73)
  • Follow the provider's instructions

Emotion prohibition (Art. 5(1)(f)): Any system inferring students' emotions is prohibited, except for medical or safety reasons. This prohibition has been in force since 2 February 2025.

Timeline: High-risk systems under Annex III must comply before 2 December 2027. Prohibited practices under Art. 5 have been in force since 2 February 2025.

AI use cases in Edtech and their classification

Use caseAI Act classificationApplicable rule
AI for admission to educational institutionsHigh riskAnnex III, point 3(a)
AI for learning assessmentHigh riskAnnex III, point 3(b)
AI for educational level assessmentHigh riskAnnex III, point 3(c)
AI for exam proctoringHigh riskAnnex III, point 3(d)
AI for emotion recognitionProhibitedArt. 5(1)(f)
AI for remote biometric identificationHigh-risk or prohibitedAnnex III point 1 + Art. 5
Tutoring chatbotLimited riskArt. 50(1) transparency
AI for content recommendationMinimal (generally)GDPR

Real enforcement cases in edtech

  • AEPD — School fine, €10,000 (2025) — Creation of a child’s electronic profile without parental consent (Arts. 6, 13 and 32 GDPR). Source: PSN Sercon blog
  • AEPD — Report on facial recognition in online exams: AEPD
  • AEPD — Resolution on biometric proctoring at a university: AEPD
  • AEPD — Guide for educational institutions: AEPD
  • AEPD — Principles for digital educational platforms: AEPD

AI Act compliance checklist for Edtech

  • Identify all AI uses in the educational product
  • Classify under Annex III point 3 (a, b, c, d)
  • Verify no emotion inference (Art. 5(1)(f), prohibited)
  • Verify no prohibited remote biometric identification
  • Implement risk management system (Art. 9)
  • Document training data quality (Art. 10), avoiding bias
  • Prepare technical documentation (Art. 11)
  • Establish human oversight (Art. 14)
  • Carry out conformity assessment (Art. 6 + Annex III)
  • Register in EU database (Art. 49)
  • Carry out Art. 27 FRIA (educational institution deployers)
  • Carry out Art. 35 GDPR DPIA for children’s and biometric data
  • Verify parental consent (LOPDGDD Art. 7, under 14s)
  • Comply with Art. 8 GDPR (children’s consent conditions)

Official sources


This page is general information, not legal advice. Each AI system must be analysed individually. Fines and deadlines cited are based on Regulation (EU) 2024/1689 in force at the date of publication.

Featured Services

Frequently Asked Questions

What educational AI is high-risk under the AI Act?

Annex III point 3 of the AI Act classifies four cases as high-risk: (a) AI to determine access or admission to educational institutions; (b) AI to evaluate learning outcomes, including adaptive learning; (c) AI to assess the appropriate level of education; (d) AI to monitor and detect prohibited behaviour during tests (proctoring). These systems require conformity assessment, technical documentation and registration in the EU database.

Is emotion recognition prohibited in education?

Yes. Art. 5(1)(f) of the AI Act prohibits AI systems inferring emotions of natural persons in the workplace and educational institutions, except for medical or safety reasons. This prohibition has been in force since 2 February 2025. Any AI system analysing facial expressions, voice or biometrics to infer students' emotions is prohibited, except in medical or safety contexts.

What does the LOPDGDD say about children's data?

The LOPDGDD Art. 7 establishes that children under 14 cannot give valid consent for the processing of their data in information society services. Parental consent is required. For children aged 14 to 18, the child's consent is valid but must be complemented with adapted information. Art. 8 GDPR sets the conditions for children's consent at European level, allowing each Member State to set the minimum age between 13 and 16 (Spain: 14).

Is proctoring with biometrics high-risk?

Yes. Proctoring with biometrics (facial recognition, iris scanning) during tests is classified as high-risk under Annex III point 3(d). In addition, biometric data are special categories (Art. 9 GDPR) and require a specific legal basis. The AEPD has published a report on facial recognition in online exams, requiring that consent be free and with a genuine alternative.

What fines does the AI Act provide for edtech?

Maximum fines are: up to €35 million or 7% of global turnover for prohibited practices (including emotion recognition in education); up to €15 million or 3% for breach of high-risk obligations. For SMEs and startups, the lower absolute amount applies. The AEPD fined a school €10,000 (2025) for creating a child's electronic profile without parental consent (Arts. 6, 13 and 32 GDPR).

What enforcement cases exist in edtech?

The AEPD fined a school €10,000 (2025) for creating a child's electronic profile without parental consent (Arts. 6, 13 and 32 GDPR). It has also published a report on facial recognition for online exams, requiring that biometric consent be free and with a genuine alternative, and has ruled against the use of biometric proctoring at a Spanish university.

When is the DPIA mandatory in edtech?

The DPIA (Art. 35 GDPR) is mandatory for high-risk processing, including large-scale processing of special categories of data (biometrics) and systematic large-scale monitoring. In edtech, biometric proctoring, adaptive assessment with profiling and large-scale processing of children's data require a DPIA. It must document the risk analysis, mitigation measures and residual assessment.

Which authorities supervise the AI Act in edtech in Spain?

The AESIA supervises the AI Act. The AEPD supervises the processing of personal data, especially of children, under the GDPR and the LOPDGDD. In educational AI cases, both authorities may intervene. The AEPD has published specific guides for educational institutions and digital educational platforms.

Contact

If you want to talk about intellectual property protection, AI and data regulation, startup formation or international expansion, write us a few lines and schedule a conversation with the team.

A2 Estudio Legal
María de Molina, 41
28006 Madrid · Spain

Tel: +34 913 451 406
Email: info@a2estudiolegal.com

Tell us briefly about your project, development stage and target countries. We will respond with the next steps to start working together.

Let's talk
Contact us