EU AI Act High-Risk AI Systems List: The Complete Reference for Compliance Teams
The EU AI Act creates a tiered risk framework, and high-risk is where the real compliance burden lives. This page breaks down every category in Annex III, maps the obligations that apply, and gives your team a clear starting point for assessing your AI portfolio.
If your organization develops, deploys, or procures AI systems that touch any of the eight high-risk domains, you face mandatory conformity assessments, risk management documentation, human oversight requirements, and ongoing monitoring obligations. Most teams underestimate the scope. This guide ensures you don't.
8
Annex III high-risk categories
EU AI Act, Annex III (2024/1689)
12+
Mandatory obligations per system
Articles 9 to 17, EU AI Act
<30%
Of organizations have completed an AI inventory
IAPP AI Governance Survey, 2024
From AI System Inventory to Ongoing Monitoring, Without the Spreadsheet Chaos
The EU AI Act's high-risk obligations span risk management, documentation, human oversight, and post-market monitoring. Priverion turns that regulatory burden into a structured, manageable workflow across every entity in your group.
AI Register for EU AI Act Compliance
Catalogue every AI system across all subsidiaries in a single, structured register. Map each system to Annex III categories, assign risk levels, and maintain the living inventory that Article 49 requires, without chasing business units for spreadsheet updates.
AXA achieved 100% ROPA recertification rate with automated workflows across their entire group.
AXA customer result, first 6 months on Priverion
AI-Assisted Impact Assessments
Generate draft Fundamental Rights Impact Assessments and DPIAs with AI-assisted analysis that pre-populates risk factors, maps relevant regulatory articles, and suggests mitigation measures. Every output is reviewed by your team before it becomes a compliance record. AI assists, humans decide.
Medtec saved 200+ hours preparing for ISO 27001 using Priverion's documentation workflows.
Medtec customer result, ISO 27001 preparation phase
Vendor AI Risk Management
The EU AI Act holds deployers accountable for the AI systems they procure, not just the ones they build. Priverion's vendor risk assessment workflows let you evaluate third-party AI providers against Annex III requirements, track conformity documentation, and maintain audit-ready evidence for every supplier relationship.
Zurzach Care achieved 100% vendor risk assessment coverage across their provider network.
Zurzach Care customer result, vendor assessment program
Group-Wide Compliance Dashboards
See the compliance posture of every subsidiary from one dashboard. Track which entities have completed their AI inventories, which high-risk assessments are pending, and where gaps exist, so you can report to the board with confidence instead of stitching together data from 47 spreadsheets.
Aircraft manufacturer reduced compliance admin time by 60% in their first 6 months.
Aircraft manufacturer customer result, first 6 months on Priverion
Incident Management and Serious Event Reporting
Article 62 requires providers of high-risk AI systems to report serious incidents to market surveillance authorities. Priverion's incident management workflow captures events, documents root causes, triggers notification timelines, and generates the evidence package regulators expect, before the deadline pressure hits.
Swiss Data Sovereignty: By Design, Not by Checkbox
Every AI-assisted analysis, every compliance record, and every piece of documentation your team creates lives within Swiss infrastructure. No customer data is used for AI model training. In a regulatory environment where data residency and cross-border transfer compliance are under constant scrutiny, this isn't a feature, it's a foundation.
All data processing within Swiss infrastructure: European data residency guaranteed.
Priverion infrastructure commitment, verified Swiss hosting
Enterprise-grade compliance without enterprise complexity
Mid-market organizations need powerful privacy management, not a platform built for Fortune 100 budgets and 18-month implementations. Why compliance teams are making the switch:
The OneTrust experience
Per-user, per-module pricing
Costs escalate unpredictably as you add users, subsidiaries, or modules. CFOs dread renewal conversations when the invoice doubles year over year.
US-hosted infrastructure
Since Schrems II, US-hosted compliance data creates the very cross-border transfer risks you're trying to manage. Additional SCCs required just to use your privacy tool.
Built for Fortune 100 complexity
Feature-rich to the point of overwhelm. Mid-market teams report using less than 30% of available functionality while paying for 100% of it.
Months-long implementation
Enterprise implementations routinely stretch 6 to 12 months. By the time you're live, regulatory requirements may have already shifted.
200+ shallow integrations
A marketplace of connectors that look impressive on paper but create maintenance overhead. Most teams need deep integration with a handful of core systems. HR, procurement, IT assets, not a catalogue.
The Priverion experience
Predictable, all-in-one pricing
Based on number of companies and organizational size, not per-user or per-module. Every capability included from day one. No expansion traps, no surprise invoices.
Guaranteed Swiss data sovereignty
Swiss-built and Swiss-hosted. All data processing within Swiss infrastructure. European data residency isn't a marketing checkbox. It's our architecture. Your compliance tool should never be a compliance risk.
Purpose-built for multi-entity management
Every feature designed around the reality of managing privacy across subsidiaries and jurisdictions. Aircraft manufacturer cut compliance admin time by 60% in their first six months, because the platform fits how group-level DPOs actually work.
Aircraft manufacturer case study, first 6 months post-implementation
Operational in weeks, not months
Clean UX designed for privacy professionals, not IT departments. Time-to-value measured in weeks. Your team is productive before the first quarterly review, not after the second.
Deep integrations where it matters
Focused integrations with HR, procurement, and IT asset management systems, the workflows that actually drive privacy compliance. Deep connectivity with fewer maintenance headaches.
An honest note: We don't cover ESG, ethics hotlines, or cookie consent. We're not built for single-entity companies. Our strength is group-wide privacy program management, and we do it exceptionally well.
Book a 30-min walkthroughThe Eight High-Risk AI Categories Your Team Needs to Assess
Annex III of the EU AI Act defines eight domains where AI systems are classified as high-risk. Each carries mandatory obligations under Articles 9 to 17. The sections below cover what compliance teams need to know about each category.
1. Biometric Identification and Categorisation
Remote biometric identification systems used in publicly accessible spaces. Includes real-time and post-identification systems, emotion recognition in workplaces and education, and biometric categorisation based on sensitive attributes.
Key obligation: Fundamental Rights Impact Assessment required before deployment (Article 27).
2. Critical Infrastructure Management
AI systems used as safety components in the management and operation of critical digital infrastructure, road traffic, and supply of water, gas, heating, and electricity.
Key obligation: Continuous post-market monitoring and serious incident reporting (Articles 61 to 62).
3. Education and Vocational Training
Systems that determine access to or assignment within educational institutions, evaluate learning outcomes, monitor prohibited behaviour during tests, or assess appropriate education levels for individuals.
Key obligation: Transparency to affected persons and human oversight mechanisms (Articles 13 to 14).
4. Employment and Worker Management
AI used in recruitment, CV screening, interview evaluation, promotion decisions, task allocation, performance monitoring, and termination of employment relationships.
Key obligation: Data governance requirements and bias testing documentation (Articles 10 to 11).
5. Essential Services Access
Systems evaluating eligibility for public assistance benefits, credit scoring, risk assessment for life and health insurance, and emergency service dispatch prioritisation.
Key obligation: Risk management system covering entire AI lifecycle (Article 9).
6. Law Enforcement
AI used for individual risk assessments, polygraphs, evidence reliability evaluation, crime prediction based on profiling, criminal offence analysis, and facial recognition database searches.
Key obligation: Logging and traceability for every decision output (Article 12).
7. Migration, Asylum, and Border Control
Systems used for polygraph assessments, application risk evaluation, document authenticity verification, and asylum/visa/residence permit application processing.
Key obligation: Human oversight with authority to override automated decisions (Article 14).
8. Administration of Justice and Democratic Processes
AI assisting judicial authorities in researching, interpreting facts and law, applying law to facts, and systems used to influence the outcome of elections or voting behaviour.
Key obligation: Quality management system and technical documentation (Articles 11, 17).
What Compliance Teams Ask About High-Risk AI Systems
What makes an AI system "high-risk" under the EU AI Act?
An AI system is classified as high-risk if it falls into one of the eight categories listed in Annex III of the EU AI Act (Regulation 2024/1689), or if it is used as a safety component of a product already covered by EU harmonised legislation listed in Annex I. High-risk classification triggers mandatory obligations including risk management, data governance, technical documentation, human oversight, accuracy requirements, and post-market monitoring under Articles 9 to 17.
When do the high-risk AI obligations take effect?
The EU AI Act entered into force on 1 August 2024. Prohibited AI practices apply from 2 February 2025. Most high-risk AI obligations, including conformity assessments and the requirements in Articles 9 to 17, apply from 2 August 2026. However, high-risk AI systems already on the market before that date will need to comply when significantly modified. Organizations should begin their AI inventory and gap assessment now to avoid a compliance scramble.
Do deployers (not just providers) have obligations for high-risk AI?
Yes. The EU AI Act places explicit obligations on deployers, the organizations using high-risk AI systems, not just those developing them. Deployers must ensure human oversight, monitor AI system operation, keep logs, conduct Fundamental Rights Impact Assessments for certain use cases (Article 27), and report serious incidents. If your organization procures AI from third-party vendors, you are a deployer and carry regulatory responsibility.
How does Priverion help with EU AI Act compliance specifically?
Priverion's AI Register allows you to catalogue every AI system across all subsidiaries, map each to Annex III categories, and assign risk classifications in one structured platform. AI-assisted impact assessments pre-populate risk factors and suggest mitigation measures, with every output reviewed by your team before it becomes a compliance record. Vendor risk workflows ensure you're assessing third-party AI providers against the same standards. And because everything runs on Swiss infrastructure with no customer data used for model training, you reduce the compliance risk that your compliance tool itself introduces.
Can Priverion handle EU AI Act compliance alongside GDPR?
Yes, that's the core design principle. Privacy program management and AI governance share significant overlap: data mapping, impact assessments, vendor management, incident response, and cross-border transfer documentation. Priverion manages both in a single platform, so your team isn't duplicating effort across separate tools. ROPA management, DPIAs, and the AI Register all live in one place with shared workflows.
What if we're not sure which of our AI systems are high-risk?
That's exactly where most organizations start. Fewer than 30% of organizations have completed an AI inventory (IAPP AI Governance Survey, 2024). Priverion's AI Register helps you begin with a structured inventory, cataloguing AI systems across your group, mapping them to Annex III categories, and identifying which require conformity assessments. The platform guides classification decisions rather than assuming your team already has the answers.
Stop managing privacy in spreadsheets
Your group-wide privacy program deserves 30 minutes of clarity
See how organizations like Aircraft manufacturer cut compliance admin time by 60% in their first six months, with automated ROPA recertification, AI-assisted DPIAs, and cross-entity visibility, all hosted on Swiss infrastructure.
No sales pitch, a live walkthrough tailored to your group structure and compliance needs.
▸ About this page: references, definitions, and FAQs
Key Takeaways: EU AI Act High-Risk AI Systems
The EU AI Act (Regulation 2024/1689) establishes the world's first comprehensive legal framework for artificial intelligence. High-risk AI systems, defined in Annex III across eight domains, carry the heaviest compliance burden: mandatory conformity assessments, risk management systems, technical documentation, human oversight, and post-market monitoring. With the high-risk obligations becoming applicable on 2 August 2026, compliance teams must inventory their AI systems, classify risk levels, and implement governance structures now. According to the IAPP-EY 2024 Privacy Governance Report, fewer than 30% of organizations have completed an AI system inventory, leaving the majority unprepared for the regulatory deadline.
Definitions
What is a high-risk AI system?
High-risk AI system refers to an AI system listed in Annex III of the EU AI Act or used as a safety component of a product covered by EU harmonisation legislation listed in Annex I. These systems are subject to mandatory requirements under Articles 9 to 17 before they can be placed on the EU market. Source: EU AI Act, Regulation 2024/1689
What is Annex III of the EU AI Act?
Annex III enumerates eight areas of high-risk AI use cases: (1) biometric identification, (2) critical infrastructure, (3) education and vocational training, (4) employment and worker management, (5) essential private and public services, (6) law enforcement, (7) migration and border control, and (8) administration of justice and democratic processes. Source: EU AI Act, Annex III
What is a conformity assessment?
A conformity assessment is the process by which a provider demonstrates that a high-risk AI system meets the requirements of the EU AI Act before placing it on the market. For most Annex III systems, providers may use internal control procedures (Annex VI), while biometric identification systems used by law enforcement require third-party assessment by a notified body (Annex VII). Source: EU AI Act, Articles 43 to 46
What is a Fundamental Rights Impact Assessment (FRIA)?
A Fundamental Rights Impact Assessment is required under Article 27 of the EU AI Act for deployers that are public bodies or private entities providing essential services. It evaluates the potential impact of a high-risk AI system on fundamental rights before deployment. Source: EU AI Act, Article 27
EU AI Act High-Risk Compliance Timeline
| Date | Milestone | Source |
|---|---|---|
| 1 August 2024 | EU AI Act enters into force | Regulation 2024/1689, Art. 113 |
| 2 February 2025 | Prohibited AI practices take effect | Art. 5 obligations apply |
| 2 August 2025 | General-purpose AI model obligations apply | Art. 51 to 56 apply |
| 2 August 2026 | High-risk AI system obligations (Annex III) become applicable | Art. 6, Annex III apply |
| 2 August 2027 | High-risk AI in Annex I products (existing legislation) fully applicable | Art. 113(3)(a) |
Key Statistics on AI Governance Readiness
According to the IAPP-EY 2024 Privacy Governance Report, fewer than 30% of organizations have completed a comprehensive AI system inventory. A 2024 McKinsey Global Survey on AI found that 72% of organizations now use AI in at least one business function, up from 55% in 2023, yet only 42% report having formal AI governance policies in place. The European Commission's impact assessment estimated that compliance costs for high-risk AI providers range from €6,000 to €7,000 per system for conformity assessments. According to ENISA's report on AI cybersecurity challenges, adversarial attacks on AI systems increased by over 50% between 2022 and 2024, underscoring the importance of the robustness requirements in Article 15.
Annex III High-Risk Categories: Comparison Table
| Category | Domain | Example AI Systems | Conformity Route |
|---|---|---|---|
| 1 | Biometric identification & categorisation | Real-time facial recognition, emotion recognition in workplaces | Third-party (law enforcement) or internal |
| 2 | Critical infrastructure | AI managing electricity grids, water supply, road traffic | Internal control (Annex VI) |
| 3 | Education & vocational training | AI determining access to education, automated grading | Internal control (Annex VI) |
| 4 | Employment & worker management | CV screening, automated interview analysis, task allocation | Internal control (Annex VI) |
| 5 | Essential services | Credit scoring, insurance risk assessment, emergency dispatch | Internal control (Annex VI) |
| 6 | Law enforcement | Predictive policing, evidence analysis, lie detection | Third-party for biometric; internal for others |
| 7 | Migration & border control | Automated visa processing, border surveillance | Internal control (Annex VI) |
| 8 | Justice & democratic processes | AI assisting judicial decisions, election influence analysis | Internal control (Annex VI) |
Frequently Asked Questions
What are high-risk AI systems under the EU AI Act?
High-risk AI systems are AI applications listed in Annex III of the EU AI Act (Regulation 2024/1689) that pose significant risks to health, safety, or fundamental rights. They span eight domains, from biometric identification to administration of justice, and require mandatory conformity assessments, risk management, and human oversight under Articles 9 to 17.
What are the 8 Annex III high-risk AI categories?
The eight categories are: (1) biometric identification and categorisation, (2) critical infrastructure management, (3) education and vocational training, (4) employment and worker management, (5) essential private and public services, (6) law enforcement, (7) migration, asylum and border control, and (8) administration of justice and democratic processes. Each category lists specific use cases that trigger high-risk classification. Source: EU AI Act, Annex III
When do high-risk AI obligations take effect?
High-risk AI system obligations under Annex III become applicable on 2 August 2026, two years after the Act entered into force. Prohibited practices already apply since 2 February 2025, and general-purpose AI model rules apply from 2 August 2025. Source: EU AI Act, Article 113
Do deployers have obligations under the EU AI Act?
Yes. Article 26 requires deployers to use high-risk AI systems according to instructions, ensure human oversight, monitor operations, maintain logs, and report serious incidents. Public-sector deployers and certain private entities must also conduct Fundamental Rights Impact Assessments under Article 27. Source: EU AI Act, Articles 26 to 27
What penalties apply for non-compliance?
Under Article 99, non-compliance with high-risk obligations can result in fines of up to €15 million or 3% of worldwide annual turnover, whichever is higher. Violations of prohibited AI practices carry fines of up to €35 million or 7% of turnover. SMEs and startups benefit from proportionate caps. Source: EU AI Act, Article 99
How does the EU AI Act interact with GDPR?
The EU AI Act complements the GDPR. High-risk AI systems processing personal data must comply with both frameworks. Article 10 requires data governance aligned with GDPR principles, and Article 26(9) mandates GDPR-compliant use of AI outputs. Organizations can leverage existing DPIAs and ROPA as a foundation for AI Act compliance. Source: EU AI Act, Recital 10 and Article 10
What is the EU AI database for high-risk systems?
Article 71 of the EU AI Act establishes an EU-wide database for high-risk AI systems. Providers must register their systems before placing them on the market, and deployers that are public bodies must also register. The database is managed by the European Commission and is publicly accessible to enhance transparency. Source: EU AI Act, Article 71
How can organizations prepare for high-risk AI compliance?
Organizations should: (1) conduct a comprehensive AI system inventory across all business units, (2) classify each system against Annex III categories, (3) perform gap analyses against Articles 9 to 17 requirements, (4) implement risk management systems and technical documentation, (5) establish human oversight protocols, and (6) set up post-market monitoring and incident reporting workflows. According to the IAPP-EY 2024 report, organizations that start with an AI inventory are 2.5× more likely to achieve compliance readiness on schedule.
