Skip to main content
Praxikon
Free check
Nederlands

Annex III AI Act: High-risk AI systems referred to in Article 6(2)

Annex III of the AI Act lists eight areas in which an AI system is high-risk: biometrics, critical infrastructure, education, employment, essential services such as credit and benefits, law enforcement, migration and border control, and justice and elections. The high-risk duties apply to these systems from 2 December 2027 and to AI in Annex I products from 2 August 2028, both under Regulation (EU) 2026/1744. Article 6(3) takes some systems out, but never when they profile people; systems already in use before 2 December 2027 fall under Article 111(2), with public-authority systems due by 2 August 2030 at the latest.

, AI and privacy legal adviserobligations checked against the source on
Check your own case for free

Annex III at a glance

  • Annex III lists eight areas. An AI system whose intended purpose falls under one of the listed points is high-risk through Article 6(2). The intended purpose counts, not the technology.
  • The high-risk duties apply to Annex III systems from 2 December 2027 and to AI in Annex I products (Article 6(1)) from 2 August 2028. Both dates follow from Regulation (EU) 2026/1744, in force since 27 July 2026.
  • Article 6(3) takes a listed system out of the category if it poses no significant risk, for example when it performs a narrow procedural or preparatory task. That never applies when the system profiles natural persons. A provider relying on the exception documents the assessment and registers the system (Article 6(4) and Article 49(2)).
  • Systems already placed on the market or put into service before 2 December 2027 only come under the duties after a significant change in their design. Systems intended to be used by public authorities must comply by 2 August 2030 at the latest (Article 111(2)).
  • The list is exhaustive. The Commission can add or modify use cases by delegated act, but only within the eight areas and where the risk is equivalent or greater (Article 7(1)).

Go straight to the official text of Annex III

Which AI systems are high-risk under Annex III?

Two or three examples per area and the boundary that most often goes wrong. Whether your system is high-risk depends on its intended purpose and on the classification rules of Article 6. An exploration by sector and use case is available at high-risk use cases by sector.

  1. Point 1: Biometrics

    • Facial recognition that searches for a person in a database at a distance (one-to-many), also when done after the fact.
    • Biometric categorisation based on sensitive or protected attributes.
    • Emotion recognition from face or voice, for example in security screening or customer calls.

    Verification that only confirms a person is who they claim to be, such as face unlock, is excluded (point 1(a)). Emotion recognition at work and in education is prohibited, except for medical or safety reasons (Article 5(1)(f)).

    More on biometrics under Annex III

  2. Point 2: Critical infrastructure

    • AI as a safety component in the control of traffic lights or road traffic.
    • AI in the monitoring of drinking water treatment or in balancing an electricity grid.

    Only AI acting as a safety component is covered (Article 3(14)). Customer service, marketing or consumption forecasts of an energy company are not. No FRIA applies for this point (Article 27(1)).

    More on critical infrastructure under Annex III

  3. Point 3: Education and vocational training

    • AI that decides on admission or placement of pupils and students.
    • AI that marks tests or evaluates learning outcomes, also when this steers the learning process.
    • Online proctoring that detects prohibited behaviour during a test.

    Spelling feedback without a grade or assessment can fall outside; once the output affects a grade, placement or level advice, the system is covered (point 3(a) to (d)).

    More on education and vocational training under Annex III

  4. Point 4: Employment, workers management and self-employment

    • CV screening, ranking or scoring candidates, and targeted job advertisements.
    • Analysis of video interviews and online assessments.
    • Task allocation based on behaviour or personal traits, and monitoring and evaluating performance.

    Interview scheduling or acknowledgements without assessment can fall outside. A recruiter making the final click does not take a scoring system off the list. Employers inform workers and their representatives before use (Article 26(7)).

    More on employment, workers management and self-employment under Annex III

  5. Point 5: Essential private and public services

    • Assessment by or on behalf of public authorities of eligibility for benefits and services, including healthcare (point 5(a)).
    • Creditworthiness or credit scores of natural persons, and risk and pricing in life and health insurance (point 5(b) and (c)).
    • Triage of emergency calls and dispatch of emergency services (point 5(d)).

    AI used to detect financial fraud is expressly excluded from point 5(b). Non-life insurance is not in point 5(c). Deployers under point 5(b) and (c) carry out a FRIA (Article 27(1)).

    More on essential private and public services under Annex III

  6. Point 6: Law enforcement

    • Assessing the risk of a person becoming the victim of a criminal offence.
    • Assessing the reliability of evidence, and polygraphs or similar tools.
    • Re-offending risk tools and profiling during detection and prosecution.

    Point 6 covers systems used by or on behalf of law enforcement authorities, where their use is permitted by law. Check the Article 5(1)(d) prohibition first: predicting criminal behaviour based solely on profiling or personality traits is prohibited.

    More on law enforcement under Annex III

  7. Point 7: Migration, asylum and border control

    • Risk assessment of persons who intend to enter or have entered the EU.
    • Support in examining asylum, visa and residence permit applications.
    • Detecting, recognising or identifying persons in the migration context.

    Verification of travel documents is excluded from point 7(d). Purely administrative sorting of files can fall under the Article 6(3) exception; substantive assessment of an application cannot.

    More on migration, asylum and border control under Annex III

  8. Point 8: Administration of justice and democratic processes

    • AI that helps a judge or dispute resolution body research and interpret facts and law in a concrete case.
    • AI intended to influence the outcome of elections or referenda or voting behaviour.

    Purely ancillary tasks such as anonymising judgments or administration fall outside point 8(a) (recital 61). Logistical campaign tools whose output does not directly reach voters are excluded from point 8(b).

    More on administration of justice and democratic processes under Annex III

What should you do if your system falls under Annex III?

  1. Map your AI systems and their intended purpose. Include purchased software in which a vendor has added AI. Record for each system what it is intended for and who uses the output.
  2. Check the prohibitions first, then Article 6. A prohibited practice (Article 5) does not become high-risk instead. If no prohibition applies, check whether the intended purpose falls under a point of Annex III, or under an Annex I product (Article 6(1)).
  3. Assess the exception and record it. Check whether one of the four conditions of Article 6(3) applies and whether the system profiles people. Record the outcome, also when the system falls outside the category, and register it in that case (Article 6(4) and Article 49(2)).
  4. Determine your role and the duties that come with it. A provider meets the requirements of Articles 8 to 15, completes the conformity assessment and registers the system (Articles 16, 43 and 49(1)). A deployer follows the instructions for use, arranges human oversight, keeps the logs and informs workers (Article 26). Public authorities also register their use (Article 49(3)).
  5. Carry out a FRIA where Article 27 requires it. The fundamental rights impact assessment applies to bodies governed by public law, private entities providing public services and deployers under point 5(b) and (c), not to point 2. For an existing public-authority system the Article 111(2) path applies: by 2 August 2030 at the latest, or earlier after a significant change.

In doubt about one system? Put it to the case check. More on the Article 6(3) filter is in the explanation of the Commission guidelines on high-risk AI; the duties per role are in Article 26 and Article 27 (FRIA).

Prefer a reasoned classification with a signature, or a legal adviser who carries it out in your team? See signed assessment or hire an AI and privacy legal adviser.

Sources for this explanation

Official text

High-risk AI systems referred to in Article 6(2) High-risk AI systems pursuant to Article 6(2) are the AI systems listed in any of the following areas:

1. Biometrics, in so far as their use is permitted under relevant Union or national law:

(a) remote biometric identification systems. This shall not include AI systems intended to be used for biometric verification the sole purpose of which is to confirm that a specific natural person is the person he or she claims to be;

(b) AI systems intended to be used for biometric categorisation, according to sensitive or protected attributes or characteristics based on the inference of those attributes or characteristics;

(c) AI systems intended to be used for emotion recognition.

2. Critical infrastructure: AI systems intended to be used as safety components in the management and operation of critical digital infrastructure, road traffic, or in the supply of water, gas, heating or electricity.

3. Education and vocational training:

(a) AI systems intended to be used to determine access or admission or to assign natural persons to educational and vocational training institutions at all levels;

(b) AI systems intended to be used to evaluate learning outcomes, including when those outcomes are used to steer the learning process of natural persons in educational and vocational training institutions at all levels;

(c) AI systems intended to be used for the purpose of assessing the appropriate level of education that an individual will receive or will be able to access, in the context of or within educational and vocational training institutions at all levels;

(d) AI systems intended to be used for monitoring and detecting prohibited behaviour of students during tests in the context of or within educational and vocational training institutions at all levels.

4. Employment, workers’ management and access to self-employment:

(a) AI systems intended to be used for the recruitment or selection of natural persons, in particular to place targeted job advertisements, to analyse and filter job applications, and to evaluate candidates;

(b) AI systems intended to be used to make decisions affecting terms of work-related relationships, the promotion or termination of work-related contractual relationships, to allocate tasks based on individual behaviour or personal traits or characteristics or to monitor and evaluate the performance and behaviour of persons in such relationships.

5. Access to and enjoyment of essential private services and essential public services and benefits:

(a) AI systems intended to be used by public authorities or on behalf of public authorities to evaluate the eligibility of natural persons for essential public assistance benefits and services, including healthcare services, as well as to grant, reduce, revoke, or reclaim such benefits and services;

(b) AI systems intended to be used to evaluate the creditworthiness of natural persons or establish their credit score, with the exception of AI systems used for the purpose of detecting financial fraud;

(c) AI systems intended to be used for risk assessment and pricing in relation to natural persons in the case of life and health insurance;

(d) AI systems intended to evaluate and classify emergency calls by natural persons or to be used to dispatch, or to establish priority in the dispatching of, emergency first response services, including by police, firefighters and medical aid, as well as of emergency healthcare patient triage systems.

6. Law enforcement, in so far as their use is permitted under relevant Union or national law:

(a) AI systems intended to be used by or on behalf of law enforcement authorities, or by Union institutions, bodies, offices or agencies in support of law enforcement authorities or on their behalf to assess the risk of a natural person becoming the victim of criminal offences;

(b) AI systems intended to be used by or on behalf of law enforcement authorities or by Union institutions, bodies, offices or agencies in support of law enforcement authorities as polygraphs or similar tools;

(c) AI systems intended to be used by or on behalf of law enforcement authorities, or by Union institutions, bodies, offices or agencies, in support of law enforcement authorities to evaluate the reliability of evidence in the course of the investigation or prosecution of criminal offences;

(d) AI systems intended to be used by law enforcement authorities or on their behalf or by Union institutions, bodies, offices or agencies in support of law enforcement authorities for assessing the risk of a natural person offending or re-offending not solely on the basis of the profiling of natural persons as referred to in Article 3(4) of Directive (EU) 2016/680, or to assess personality traits and characteristics or past criminal behaviour of natural persons or groups;

(e) AI systems intended to be used by or on behalf of law enforcement authorities or by Union institutions, bodies, offices or agencies in support of law enforcement authorities for the profiling of natural persons as referred to in Article 3(4) of Directive (EU) 2016/680 in the course of the detection, investigation or prosecution of criminal offences.

7. Migration, asylum and border control management, in so far as their use is permitted under relevant Union or national law:

(a) AI systems intended to be used by or on behalf of competent public authorities or by Union institutions, bodies, offices or agencies as polygraphs or similar tools;

(b) AI systems intended to be used by or on behalf of competent public authorities or by Union institutions, bodies, offices or agencies to assess a risk, including a security risk, a risk of irregular migration, or a health risk, posed by a natural person who intends to enter or who has entered into the territory of a Member State;

(c) AI systems intended to be used by or on behalf of competent public authorities or by Union institutions, bodies, offices or agencies to assist competent public authorities for the examination of applications for asylum, visa or residence permits and for associated complaints with regard to the eligibility of the natural persons applying for a status, including related assessments of the reliability of evidence;

(d) AI systems intended to be used by or on behalf of competent public authorities, or by Union institutions, bodies, offices or agencies, in the context of migration, asylum or border control management, for the purpose of detecting, recognising or identifying natural persons, with the exception of the verification of travel documents.

8. Administration of justice and democratic processes:

(a) AI systems intended to be used by a judicial authority or on their behalf to assist a judicial authority in researching and interpreting facts and the law and in applying the law to a concrete set of facts, or to be used in a similar way in alternative dispute resolution;

(b) AI systems intended to be used for influencing the outcome of an election or referendum or the voting behaviour of natural persons in the exercise of their vote in elections or referenda. This does not include AI systems to the output of which natural persons are not directly exposed, such as tools used to organise, optimise or structure political campaigns from an administrative or logistical point of view.

Source: EUR-Lex, Regulation (EU) 2024/1689. Text reproduced verbatim.

Does your AI system fall under these rules?

Determine your risk class in minutes and see where your organisation stands.

Run the risk checkOr take the AI Readiness Score

Frequently asked questions

Which AI systems are listed in Annex III of the AI Act?

Annex III contains eight categories of high-risk AI systems: (1) biometrics, (2) critical infrastructure, (3) education and vocational training, (4) employment and workers management, (5) essential services, (6) law enforcement, (7) migration and border control, and (8) administration of justice and democratic processes.

Is my AI system high-risk according to Annex III?

Check whether your AI system falls under one of the eight categories in Annex III. The system must be 'intended to be used' for the stated purpose. Classification depends on the intended use, not the underlying technology.

Which AI systems in HR and recruitment are high-risk?

Under Annex III, point 4: AI for (a) recruitment and selection, including targeted job ads, filtering applications and evaluating candidates, and (b) decisions on working conditions, promotion, termination, task allocation and monitoring performance.

Do AI systems in the banking sector fall under Annex III?

Yes, if the AI system is used for creditworthiness assessments or establishing credit scores of natural persons (point 5b). Exception: AI systems for detecting financial fraud are excluded.

Are AI systems in healthcare high-risk under Annex III?

Yes, under point 5a: AI used by or on behalf of public authorities to evaluate eligibility for essential public services, including healthcare. AI for emergency triage (point 5d) is also high-risk.

Which biometric AI systems are high-risk?

Under Annex III, point 1: (a) remote biometric identification systems, (b) biometric categorisation based on sensitive characteristics, and (c) emotion recognition. Biometric verification (confirming someone is who they claim to be) is excluded.

Do government AI systems fall under Annex III?

Yes, multiple categories are relevant for government: AI for public services (point 5a), law enforcement (point 6), migration and border control (point 7), and administration of justice (point 8). AI in education (point 3) is also relevant for public institutions.

Is AI for emotion recognition high-risk?

Yes, AI systems intended for emotion recognition are listed in Annex III, point 1c. In the workplace and in education, emotion recognition is even prohibited (Article 5(1)(f)), unless the system is used for medical or safety reasons.

Which AI in education is high-risk?

Annex III, point 3 lists four categories: (a) admission to educational institutions, (b) evaluating learning outcomes, (c) assessing the appropriate education level, and (d) monitoring prohibited behaviour during tests (proctoring).

Is AI for emergency calls high-risk?

Yes, Annex III, point 5d classifies AI systems that evaluate and classify emergency calls, or are used for dispatching emergency services (police, fire, ambulance) and emergency healthcare triage as high-risk.

Can the European Commission expand Annex III?

Yes, but only within the eight existing areas. Article 7(1) empowers the Commission to add or modify Annex III use cases by delegated act, where they are intended to be used in one of the areas listed in the annex and pose a risk equivalent to or greater than the use cases already listed.

Do AI systems for election influence fall under Annex III?

Yes, Annex III, point 8b classifies AI systems intended to influence the outcome of elections or referenda, or voting behaviour, as high-risk. Logistical campaign tools not directly exposed to citizens are excluded.

Does AI for insurance fall under Annex III?

Yes, Annex III point 5c classifies AI systems used for risk assessment and pricing in relation to natural persons for life and health insurance as high-risk. AI for claims detection and fraud prevention in insurance does not automatically fall under this category.

Is AI-powered CV screening high-risk?

Yes, Annex III point 4a explicitly classifies AI systems for filtering and evaluating job applications as high-risk. This includes automated CV screening, AI-powered candidate matching and automated assessments in recruitment processes.

How do I know if my AI system is 'intended to be used' for an Annex III purpose?

Classification depends on the intended purpose as determined by the provider, not on technical capabilities. If an AI system is technically capable of an Annex III purpose but not intended for it, it is not automatically high-risk. However, repurposing for an Annex III purpose can lead to reclassification.

When do the Annex III rules apply, and is every listed system high-risk?

Since the Digital Omnibus (Regulation (EU) 2026/1744) the high-risk rules for Annex III apply from 2 December 2027. Not every listed system is high-risk: under Article 6(3) a system falls outside if it poses no significant risk, for example because it only performs a narrow procedural or preparatory task, but never if it profiles natural persons. For systems already in use before 2 December 2027, Article 111(2) applies: only once significantly changed, and for public-authority systems by 2 August 2030 at the latest.

Can an AI system fall under more than one Annex III area?

Yes. A system used for both recruitment (point 4) and creditworthiness assessment (point 5(b)) falls under both points. It is high-risk as soon as one point applies, and the obligations then apply to the whole system.

Does Annex III also apply to general-purpose AI models?

No, general-purpose AI models have their own regime in Chapter V of the AI Act. An AI system built on such a model and used for an Annex III purpose can still be high-risk. The high-risk obligations then apply to the provider of that system.

What if my AI system falls under none of the eight areas?

Then it is not high-risk through Article 6(2). Do check the other routes: high-risk as a product or safety component under Annex I (Article 6(1), from 2 August 2028), a prohibited practice (Article 5) and the transparency obligations of Article 50, which have applied since 2 August 2026.

Does a public authority with an existing Annex III system have to comply on 2 December 2027?

Not necessarily. Systems already placed on the market or put into service before 2 December 2027 only come under the high-risk duties after a significant change in their design. Systems intended to be used by public authorities must comply by 2 August 2030 at the latest, even without a change (Article 111(2)). This also covers the FRIA.

Who has to carry out a FRIA for an Annex III system?

Article 27 applies to deployers that are bodies governed by public law or private entities providing public services, and to deployers of systems for creditworthiness (point 5(b)) and for life and health insurance (point 5(c)). No FRIA applies to critical infrastructure (point 2). A private employer using a recruitment system is therefore not automatically covered.

Official guidance on this annex

13