The Annex III high-risk areas, explained

An AI system is high-risk under Art. 6(2) when it is used in one of the eight areas listed in Annex III — unless the narrow Art. 6(3) exception applies (see The Article 6(3) exception). The Annex III obligations apply from 2 December 2027, deferred from the original date by Regulation (EU) 2026/1744.

The eight areas

1. Biometrics — remote biometric identification (not mere verification of who someone claims to be), categorisation by sensitive attributes, emotion recognition. Example: face-matching people against a watchlist.
2. Critical infrastructure — safety components in the management of critical digital infrastructure, road traffic, or the supply of water, gas, heating and electricity. Example: AI managing grid load balancing.
3. Education and vocational training — access and admission, evaluating learning outcomes, assessing the education level a person can receive, monitoring prohibited behaviour during tests. Example: automated proctoring that flags exam cheating.
4. Employment, workers management and access to self-employment — recruitment and selection, filtering applications, evaluating candidates, decisions on terms, promotion and termination, task allocation from behaviour or personal traits, monitoring and evaluating performance. Example: a CV-ranking tool that shortlists applicants.
5. Access to essential private and public services and benefits — four sub-points: 5(a) eligibility for essential public assistance benefits and services, including healthcare; 5(b) creditworthiness and credit scoring, excluding fraud detection; 5(c) risk assessment and pricing in life and health insurance; 5(d) emergency-call classification and dispatch, including emergency healthcare triage. Example: a model that sets a loan decision or a life-insurance premium.
6. Law enforcement — risk assessment of individuals, polygraphs and similar tools, evidence-reliability evaluation, profiling in investigations, and assessing the risk of offending or reoffending where that is not based solely on profiling. Example: predicting reoffending risk for bail decisions.
7. Migration, asylum and border control — risk assessment of people entering the EU, examining applications for asylum, visas and residence permits, detecting and identifying people at borders. Example: AI triaging visa applications.
8. Administration of justice and democratic processes — assisting judicial authorities in researching and interpreting facts and law, alternative dispute resolution, influencing the outcome of an election or referendum. Example: a tool that drafts judicial rulings.

Why the area matters

The area is not just a label. Veritome stores it on your system's classification, and the obligation engine uses it — together with your role and risk tier — to decide which obligations apply. Two examples that turn on the area:

• The FRIA (Art. 27) attaches to deployers that are public bodies or private providers of public services in the biometrics, education, employment, law-enforcement, migration and justice areas — and to every deployer of a creditworthiness (5(b)) or life and health insurance (5(c)) system, public or private. See FRIA: the fundamental-rights impact assessment, step by step.
• The Art. 6(3) exception can never be claimed for a system that profiles natural persons, whichever area it sits in.

The area also determines which competent authority is relevant for incidents and registration, and it is one of the Annex VIII fields the EU-database registration asks for.

In the registration wizard you pick one Annex III area — the most relevant match — and, for essential services, the sub-point. Choosing "None of the Annex III domains" routes you to the transparency (Art. 50) and minimal-risk checks instead.