EU AI Act resource · Last verified 2026-08-02

Annex III categories — the 8 high-risk areas

Annex III is where the EU AI Act's high-risk regime bites for everyday AI. Here are the eight categories, what triggers each one, and the systems that typically fall inside them.

1

§1 Biometrics

Biometric identification, categorisation, and emotion recognition. Includes facial recognition for access control, biometric identity verification, and emotion inference — but not the banned Art. 5 uses, which are worse.

2

§2 Critical infrastructure

AI that is a safety component of critical infrastructure: traffic management, water and gas supply, and electricity grids. If failure could threaten life or critical services, expect the full obligations.

3

§3 Education and vocational training

AI used to determine access to education, evaluate learning outcomes, or monitor students during tests — admissions screening, automated grading, and exam proctoring are the classic examples.

4

§4 Employment and worker management

AI for recruitment, candidate evaluation, promotion, and monitoring of workers. Hiring assistants, CV screeners, and productivity trackers all land here. The most common high-risk trigger for private companies.

5

§5 Essential private and public services

AI that decides access to essential services: creditworthiness scoring, insurance risk pricing, and emergency-call dispatch. Denying someone credit or coverage through AI is exactly what the Act is watching.

6

§6 Law enforcement

AI used for crime prediction, risk assessment, profiling, and evidence evaluation. Heavy safeguards and, for some uses, separate rules on biometric identification apply.

7

§7 Migration and border control

AI for travel documents, visa processing, and border surveillance. High stakes for fundamental rights — asylum and migration decisions are specifically protected.

8

§8 Administration of justice

AI that assists judicial authorities in researching and applying the law, or that supports alternative dispute resolution. Human decision-making by judges is the non-negotiable centrepiece.

9

The significant-risk test

An Annex III system is high-risk only if it also poses a significant risk to health, safety, or fundamental rights (Art. 6(3)). Providers can document a carve-out — but 'we don't think it matters' is not documentation.

Go deeper

Related guides & tools

See where your AI actually lands

Describe your system in the free Risk Scanner for a preliminary classification and the obligations that likely apply.

Open the Risk Scanner

Preliminary EU AI Act clarity summary. Not legal advice.