Eigma
EN 中 DE
The EU AI Act in Everyday Life Open Eigma
Contents Admission and assessment

Work and education

Admission and assessment

AI determining admission, educational access, placement or educational level, and AI evaluating learning outcomes, including those used to steer learning, are listed as high-risk. The main provider and deployer safeguards are scheduled for 2 December 2027 after the AI Omnibus postponed them.12

Use Duty-holder and duty Enforcer; start
Training and testing data Articles 10 and 16: provider; use of relevant, sufficiently representative data.1 Relevant national market-surveillance authority; 2 December 2027
Human oversight Articles 14 and 16: provider; effective oversight. Article 26: educational deployer; assignment of competent, trained, authorised people.1 Relevant national market-surveillance authority; 2 December 2027
AI-assisted personal decision Article 26: educational deployer; notice to affected person of high-risk AI use.3 Relevant national market-surveillance authority; 2 December 2027
Public service use Article 27: public-law body or private entity providing public services; fundamental-rights risk assessment before first use; authority notification.1 Relevant national market-surveillance authority; 2 December 2027
Public authority registration Article 49: public authority or body deploying the system; registration of itself and its use; no use of an unregistered system.1 Relevant national market-surveillance authority; 2 December 2027

These are duties on providers and institutions, not rights to technical documentation or replacements for institutional appeals.

Since 2 August 2026, Article 85 has allowed applicants or students with grounds to suspect an infringement to complain under national procedures. As of 12 September 2026, those procedures were country-specific, and in many countries no relevant market-surveillance authority or contact point had yet been designated. Article 86 has formally applied since then. On request, an institution acting as deployer must explain the role and main elements of covered AI in an adverse admission, grading, placement or access decision with legal or similarly significant effects. The market-surveillance authority enforces that duty.1

The Act sets no Article 86 request form, response deadline or appeal route. Its operation before high-risk classification begins is unresolved; national education appeals are separate.

For personal data, since 25 May 2018, Article 15 has required the education provider acting as controller to give the person's data and meaningful information about automated logic and consequences. Under Article 22, since the same date, the education provider acting as controller must not subject a person to a solely automated decision with legal or similarly significant effects unless an exception and safeguards apply. University admission can meet that threshold. National data-protection authorities and courts enforce both rules.4

Where this stops

Not every educational AI tool is high-risk. Article 6 can exclude a listed use only when it poses no significant risk and does not materially influence outcomes. Not every private institution provides public services. Older private systems may remain outside the deferred safeguards unless significantly redesigned.1

References

  1. [1]Consolidated EU AI Act
  2. [2]Regulation (EU) 2026/1744, the AI Omnibus
  3. [3]European Commission AI Act FAQ
  4. [4]European Data Protection Board guidance on individual rights
Quizzes

  1. An applicant suspects AI affected an admission decision in September 2026. Which response correctly separates available options from deferred safeguards?

    • GDPR protections may apply, and a person with grounds to suspect an AI Act infringement may complain; the main safeguards remain deferred
    • Provider data-quality and institutional human-oversight duties already apply, but GDPR protections begin only when those AI Act duties begin
    • GDPR may provide access to personal data, but an AI Act complaint becomes possible only after the deferred safeguards begin

    GDPR protections may already apply, and AI Act complaints have been possible since August 2026. The main data, oversight and notice safeguards remain deferred to December 2027.

  2. Every AI tool that recommends study material is automatically treated as a high-risk educational system.

    • True
    • False

    The classification concerns listed uses that influence admission, access, educational level or learning outcomes. A non-influential tool can fall outside high-risk status.

Comments

No comments yet. Start the conversation.