Skip to main content
    Guide
    Annex III

    High-Risk AI Systems (Annex III)

    The EU AI Act defines 8 categories of high-risk AI in Annex III. If your AI falls into these categories, you have specific obligations as a deployer. Here's what you need to know.

    Understanding High-Risk Classification

    High-risk AI systems are those that pose significant risks to health, safety, or fundamental rights. The EU AI Act identifies these through two mechanisms:

    Annex III Categories

    Eight specific use-case areas defined in the Act: biometrics, critical infrastructure, education, employment, essential services, law enforcement, migration, and justice/democracy.

    Product Safety Components

    AI systems that are safety components of products already covered by EU harmonization legislation (medical devices, machinery, vehicles, etc.).

    The 8 Annex III Categories

    Biometrics

    Remote biometric identification and biometric categorisation of natural persons

    Examples:
    • Facial recognition for identification
    • Biometric categorisation systems
    • Emotion recognition in certain contexts

    Critical Infrastructure

    AI as safety components in management/operation of critical infrastructure

    Examples:
    • Energy grid management
    • Water supply systems
    • Transport infrastructure
    • Digital infrastructure

    Education & Training

    AI used in education and vocational training contexts

    Examples:
    • Student assessment and scoring
    • Admissions decisions
    • Exam proctoring
    • Learning progress evaluation

    Employment & Workers

    AI in employment, worker management, and self-employment access

    Examples:
    • CV screening and recruitment
    • Interview analysis
    • Performance evaluation
    • Task allocation
    • Workforce monitoring

    Essential Services

    AI affecting access to essential private and public services

    Examples:
    • Credit scoring
    • Insurance pricing
    • Emergency services dispatch
    • Healthcare access
    • Social benefits eligibility

    Law Enforcement

    AI used by law enforcement authorities

    Examples:
    • Evidence assessment
    • Risk of offending/reoffending
    • Lie detection (polygraph)
    • Criminal profiling

    Migration & Border

    AI in migration, asylum, and border control

    Examples:
    • Visa application assessment
    • Border patrol systems
    • Asylum claim processing
    • Security risk assessment

    Justice & Democracy

    AI in justice administration and democratic processes

    Examples:
    • Case outcome research
    • Judicial decision support
    • Election-related systems
    • Democratic process influence detection

    Key Deployer Obligations

    If you deploy high-risk AI, you must:

    • Use the system according to provider instructions
    • Assign human oversight to competent persons with authority to intervene
    • Ensure input data is relevant and representative (if under your control)
    • Monitor the system's operation and inform provider of risks
    • Keep logs under your control for at least 6 months
    • Inform workers before using AI that affects them
    • Report serious incidents to providers and authorities
    • Complete a FRIA if you're a public body or provide public services

    Frequently Asked Questions

    What makes an AI system 'high-risk'?

    An AI system is high-risk if it falls into one of the Annex III categories (biometrics, employment, credit, etc.) or is a safety component of a product covered by EU harmonization legislation (e.g., medical devices, vehicles).

    When do high-risk obligations apply?

    Most high-risk AI obligations apply from 2 August 2026. However, for AI systems that are safety components of products already covered by EU law (Annex I), there's an extended transition until 2 August 2027.

    What are the main deployer obligations for high-risk AI?

    Deployers must: use systems according to instructions, assign competent human oversight, ensure input data relevance, monitor operation, report risks and incidents, keep logs for 6+ months, and inform workers (where applicable).

    Do I need a FRIA for high-risk AI?

    A Fundamental Rights Impact Assessment (FRIA) is required for public bodies and certain private entities using high-risk AI, particularly those affecting individuals' rights. Use our FRIA tool to check your requirements.

    How do I know if my AI is high-risk?

    Use our High-Risk Checker tool to assess your AI systems against Annex III categories. The tool asks targeted questions to determine if your use case triggers high-risk classification.

    Manage High-Risk AI Compliance

    Klarvo helps you classify, control, and evidence compliance for all your high-risk AI systems.

    No credit card
    14-day trial
    Cancel anytime