Overview
Schools, universities, continuing education providers and research institutions process data on particularly vulnerable people – children, students, study participants and patients. At the same time, they increasingly use AI: for admission, adaptive learning platforms, automated grading and exam proctoring. These are precisely the applications the AI Act classifies as high-risk or even prohibits.
Research brings its own topics: cybersecurity (research organisations are a NIS2 sector, and universities are popular targets), research data protection and, in future, access to health data via the EHDS. Many obligations depend on whether an institution is public or private and whether it conducts research, teaches or commercialises results.
The Most Important Regulations
| Regulation | What it means for education & research | Status / date |
|---|---|---|
| AI Act | High-risk for admission, assessment, placement and proctoring; ban on emotion recognition | bans since 02.02.2025, high-risk from 02.12.2027 |
| GDPR | Pupil, student and research data, learning platforms, research privilege | since 25.05.2018 |
| NIS2 | Commercially oriented research organisations as important entities | DE since 06.12.2025, AT since 01.10.2026 |
| EHDS | Access to health data for research via health data access bodies | applies from 26.03.2027, secondary use from 26.03.2029 |
| eIDAS 2.0 | Digital certificates and diplomas as attestations of attributes in the EUDI Wallet | wallets by end of 2026 |
| Whistleblower Directive | Internal reporting channel from 50 employees | since 17.12.2023 |
AI in Teaching and Examinations
The AI Act explicitly names education and vocational training as a high-risk area (Annex III). It covers AI systems that
- decide on access or admission to educational institutions,
- evaluate learning outcomes, including when used to steer the learning process,
- determine the appropriate level of education for a person,
- monitor and detect prohibited behaviour during tests (proctoring).
Providers of such systems – often EdTech companies – need risk management, technical documentation, conformity assessment and CE marking from 2 December 2027. Schools and universities as deployers must use the systems as intended, ensure human oversight and keep logs for at least six months; public institutions carry out a fundamental rights impact assessment beforehand.
Emotion recognition in educational institutions has already been prohibited since 2 February 2025, except for medical or safety reasons. Since 2 August 2026, transparency obligations apply, for example to chatbots in student advisory services. All institutions using AI must take measures to ensure the AI literacy of their staff.
AI systems developed and used solely for scientific research and development are excluded from the AI Act. The exclusion ends as soon as the system is used outside research or made available to others.
Data Protection for Learners and Research Data
The GDPR shapes everyday school life: learning platforms, video conferencing, grade management and cloud services need a legal basis, usually the statutory educational mandate, and data processing agreements. Where a service is offered directly to children, they can consent themselves from 16 in Germany and from 14 in Austria. Public schools and universities must appoint a data protection officer.
In research, the GDPR allows derogations where appropriate safeguards such as pseudonymisation are in place (Art. 89 GDPR); nationally, section 27 BDSG and the Austrian Research Organisation Act (FOG) supplement this. Health and genetic data are special categories and require strict safeguards. Via the EHDS, researchers will be able to use health data with a data permit in secure processing environments from 2029; research institutions may at the same time be obliged to make data available as data holders.
Cybersecurity at Universities and Institutes
Research is an other critical sector under NIS2 (Annex II). It covers research organisations whose primary goal is applied research or experimental development with a view to commercial exploitation – from 50 employees or more than EUR 10 million turnover and balance sheet total. Educational institutions are not research organisations; member states may, however, extend NIS2 to them. Universities that are not covered should still use the NIS2 minimum measures as a benchmark: open networks, many user accounts and valuable research data make them a frequent ransomware target.
Research involving sensitive technologies is also subject to export controls for dual-use items – including technology transfer to foreign partners. How to bundle partner and sanctions checks is explained in Business Partner Due Diligence.
Digital Credentials and Whistleblower Protection
With the EUDI Wallet under eIDAS 2.0, diplomas and certificates can be issued as qualified electronic attestations of attributes and presented across borders. Universities should consider how they will issue and accept digital credentials.
Institutions with 50 or more employees need an internal reporting channel (HinSchG, HSchG). Typical reports concern research misconduct, third-party funding, procurement and data protection.
Deadlines 2026/2027
| Date | What applies |
|---|---|
| since 02.08.2026 | AI Act: transparency obligations, e.g. for chatbots |
| since 01.10.2026 | Austria: NIS2 obligations under the NISG 2026 |
| 02.12.2026 | AI Act: machine-readable marking also for systems placed on the market before 02.08.2026 |
| 31.12.2026 | Austria: NIS2 registration; EUDI Wallet in all member states |
| 02.01.2027 | Announced launch of the German EUDI Wallet |
| 26.03.2027 | EHDS applies in principle |
| 02.12.2027 | AI Act: high-risk obligations for AI in education and examinations |
DACH Specifics
- Germany: Education is a matter for the federal states; schools and state universities are subject to state data protection and school laws. The BSIG covers research organisations but excludes educational institutions.
- Austria: Age of consent 14; research data under the FOG; NISG 2026 since 1 October 2026.
- Switzerland: The AI Act and NIS2 do not apply. Cantonal universities, the federal institutes of technology, universities of applied sciences and universities of teacher education have, however, had to report cyberattacks to the Federal Office for Cybersecurity (BACS) within 24 hours since 1 April 2025. EU research collaborations and EdTech offerings for the EU are subject to EU law.
Checklist
- Create an AI inventory: flag admission, assessment, placement and proctoring as high-risk
- Rule out emotion recognition and comparable analyses
- Agree with EdTech providers on conformity evidence for high-risk systems by 2 December 2027
- Train teaching and administrative staff in AI literacy
- Review learning platforms and cloud services for legal basis and data processing agreements
- Pseudonymise research data and maintain data management plans
- Check NIS2 applicability for research organisations; test contingency plans against ransomware
- Review export controls for research collaborations
- Set up an internal reporting channel from 50 employees