High-risk AI under the EU AI Act

The EU AI Act classifies AI systems as "high-risk" via two independent routes: the Annex I product route (AI as a safety component of CE-marked products) and the Annex III use-case route (8 listed application areas). Use this reference to check whether your AI system falls into either category - and whether the Article 6(3) filter brings it back out. Article 6

The short version

Route 1 - Annex I (product safety): AI that is a safety component of a product regulated under EU harmonisation law (machinery, medical devices, toys, vehicles, aviation, etc.) and required to undergo third-party conformity assessment.
Route 2 - Annex III (8 use-case areas): Stand-alone AI systems in biometrics, critical infrastructure, education, employment, essential services, law enforcement, migration/asylum, or administration of justice and democratic processes.
Article 6(3) filter: An Annex III system is not high-risk if it does not pose a significant risk of harm to health, safety or fundamental rights - but this exception does not apply where the system profiles people.

Two routes to high-risk

How an AI system becomes high-risk

Route 1 - Annex I

Safety component of a regulated product

AI that is a safety component of, or is itself, a product regulated under the EU harmonisation laws listed in Annex I (e.g. machinery, medical devices, toys, motor vehicles, aviation equipment) and that product is required to undergo third-party conformity assessment. Annex I

Obligations under Route 1 apply from 2 August 2027 under current law ([OMNIBUS - PROPOSED] deferred to 2 August 2028 if the Digital Omnibus is adopted - not yet law as of 9 June 2026).

Route 2 - Annex III

Listed high-risk use-case area

Stand-alone AI systems that fall in one of the 8 use-case areas in Annex III - unless the Article 6(3) filter applies. The filter requires a documented assessment that the system does not pose significant risk of harm. It cannot be used if the system profiles natural persons. Annex III

Obligations under Route 2 apply from 2 August 2026 under current law ([OMNIBUS - PROPOSED] deferred to 2 December 2027 if the Digital Omnibus is adopted - not yet law as of 9 June 2026).

Article 6(3) filter - not a free pass

The Article 6(3) filter allows an Annex III system to avoid the high-risk classification if it does not pose a significant risk of harm. Indicators of non-significant risk: the system performs a narrow procedural task; it improves a prior human activity; it detects decision-making patterns without replacing human judgement; it performs preparatory tasks only. The filter does not apply where the system profiles natural persons. You must document your filter assessment and be prepared to defend it. Article 6(3)

High-risk AI reference: 14 areas

Showing 8 Annex III use-case areas and 6 Annex I product categories. Use the search and filter to find relevant areas for your AI system.

Search use cases

14 areas shown

High-risk AI use cases under the EU AI Act. Columns: area, category, example use cases, and a compliance note.
Area	Category	Example use cases	Note
Biometrics (area 1)	Annex III	Remote biometric identification systems (face, gait, fingerprint) Biometric categorisation systems inferring attributes from biometric data Emotion recognition systems used outside prohibited contexts	Excludes uses already banned by Article 5 (untargeted facial-image scraping, real-time biometric ID in public for law enforcement without authorisation, and biometric categorisation inferring sensitive attributes). Source: https://artificialintelligenceact.eu/annex/3/
Critical infrastructure (area 2)	Annex III	AI used as safety components in road-traffic management systems AI in management or operation of water, gas or heating distribution AI safety components in electricity-grid or digital-infrastructure management	Covers AI that is a safety component in, or manages/operates, critical infrastructure. The Article 6(3) filter may apply to purely administrative or monitoring tools that do not directly affect safety. Source: https://artificialintelligenceact.eu/annex/3/
Education & vocational training (area 3)	Annex III	AI determining access or assignment to educational or vocational institutions AI assessing or evaluating learning outcomes AI evaluating a person's learning level AI monitoring or proctoring students during exams	Covers AI that determines or materially influences educational access or assessment. AI that merely recommends learning resources or provides study aids may qualify for the Article 6(3) filter exception. Source: https://artificialintelligenceact.eu/annex/3/
Employment & workers management (area 4)	Annex III	AI for recruitment and selection of candidates (screening CVs, shortlisting) AI making or substantially influencing decisions on promotion or termination AI monitoring and evaluating employee performance or behaviour AI allocating tasks to workers (including gig workers)	One of the most compliance-relevant areas for enterprise deployers. HR software vendors offering AI-assisted screening, performance management or workflow automation should assess this area carefully. Source: https://artificialintelligenceact.eu/annex/3/
Access to essential private & public services (area 5)	Annex III	Credit-scoring and creditworthiness evaluation AI (excluding fraud detection) AI assessing eligibility for public benefits or social services AI for life insurance or health insurance risk assessment and pricing Emergency service dispatch and triage AI	Credit scoring and insurance risk assessment are the highest-profile commercial use cases here. These also trigger the Article 27 FRIA requirement for deployers. Fraud-detection AI is excluded from the credit-scoring entry. Source: https://artificialintelligenceact.eu/annex/3/
Law enforcement (area 6)	Annex III	AI assessing the risk of a person committing an offence or re-offending AI polygraph and similar lie-detection tools AI evaluating reliability of evidence in criminal investigations AI profiling individuals in the context of criminal investigations	Narrow sector - primarily relevant to law-enforcement authorities and vendors supplying them. Some uses may overlap with Article 5 prohibitions (predictive policing based solely on profiling). Source: https://artificialintelligenceact.eu/annex/3/
Migration, asylum & border control (area 7)	Annex III	AI polygraph or lie-detection tools used in migration/asylum contexts AI assessing risk of irregular migration or other risks AI examining or deciding on asylum, visa or residence permit applications AI detecting or recognising persons at borders	Primarily relevant to government/border-control authorities and vendors supplying them. High sensitivity given fundamental rights implications for applicants. Source: https://artificialintelligenceact.eu/annex/3/
Administration of justice & democratic processes (area 8)	Annex III	AI tools that assist judges or judicial authorities in researching, interpreting or applying the law AI used to influence election or referendum outcomes AI targeting voters based on political preferences or behaviour	Very narrow. Pure legal-research tools that help lawyers find relevant case law may qualify for the Article 6(3) filter if they perform preparatory tasks without replacing judicial judgement. Electoral AI uses raise the most severe fundamental-rights concerns. Source: https://artificialintelligenceact.eu/annex/3/
Medical devices & IVDs	Annex I product	AI clinical decision-support software used as a safety component in a medical device (MDR/IVDR) AI-enabled diagnostic imaging analysis classified as a medical device AI systems embedded in in-vitro diagnostic equipment	AI that is a safety component of, or is itself, a medical device under Regulation (EU) 2017/745 (MDR) or in-vitro diagnostic device under Regulation (EU) 2017/746 (IVDR), and required to undergo third-party conformity assessment. This is the most commercially significant Annex I category. Source: https://artificialintelligenceact.eu/article/6/
Machinery	Annex I product	AI safety functions in industrial robots under the Machinery Regulation (EU) 2023/1230 AI collision-avoidance or emergency-stop systems in automated machinery AI monitoring systems that constitute safety components in machinery	AI that is a safety component of a product regulated under the Machinery Regulation (EU) 2023/1230 (which replaces Directive 2006/42/EC). Third-party conformity assessment triggers the AI Act high-risk classification. Source: https://artificialintelligenceact.eu/article/6/
Motor vehicles & automotive	Annex I product	AI safety systems in vehicles subject to EU vehicle type-approval (Regulation (EU) 2019/2144) AI advanced driver-assistance systems (ADAS) in scope of the automotive type-approval regulation AI autonomous emergency braking, lane-keeping or occupant protection systems	AI in motor vehicles regulated under EU type-approval legislation (Regulation (EU) 2018/858 and Regulation (EU) 2019/2144). Where these systems must undergo third-party conformity assessment, they trigger the Annex I high-risk route. Source: https://artificialintelligenceact.eu/article/6/
Aviation	Annex I product	AI safety components in civil aircraft and equipment under EASA regulations AI used in flight-control or autopilot systems subject to EU Aviation Safety Agency oversight AI ground-based safety systems covered by EU aviation safety rules	AI that is a safety component of civil aviation products regulated under Regulation (EU) 2018/1139. Typically subject to stringent EASA certification requirements that also trigger the AI Act high-risk classification. Source: https://artificialintelligenceact.eu/article/6/
Toys	Annex I product	AI embedded in connected toys as a safety component under the Toy Safety Directive (2009/48/EC) AI content-filtering or interaction systems in toys that constitute a safety feature	AI that is a safety component of toys regulated under Directive 2009/48/EC, where the underlying product legislation requires third-party conformity assessment. Source: https://artificialintelligenceact.eu/article/6/
Lifts, pressure equipment & other products	Annex I product	AI safety systems in lifts regulated under Directive 2014/33/EU AI monitoring in pressure equipment under Directive 2014/68/EU AI components in radio equipment under Directive 2014/53/EU (RED)	Several further product categories in Annex I are less commercially prominent for AI but follow the same rule: AI safety component + third-party conformity assessment required = high-risk. Source: https://artificialintelligenceact.eu/article/6/

This reference covers the areas defined in Annex III and Annex I of Regulation (EU) 2024/1689. It is a summary - always confirm the precise scope against the full regulation text. This is guidance, not legal advice.

Need to classify your specific AI system? Try the risk classifier. For the full obligations once classified, see the obligations guide.

Sources

[1]Annex III of Regulation (EU) 2024/1689 - AI Act Explorerretrieved 9 Jun 2026
[2]Article 6 of Regulation (EU) 2024/1689 - classification as high-risk AI systemretrieved 9 Jun 2026
[3]Annex I of Regulation (EU) 2024/1689 - Union harmonisation legislationretrieved 9 Jun 2026
[4]Regulation (EU) 2024/1689 (EU AI Act) - EUR-Lexretrieved 9 Jun 2026
[5]AI Act high-level summary - AI Act Explorerretrieved 9 Jun 2026

The AI Act Brief

Stay up to date with the EU AI Act

We watch Brussels so you don't. Plain-English EU AI Act updates, free.

No spam. Unsubscribe anytime.