What this page is for
It helps users produce a concrete artifact, checklist, or decision note instead of only reading generic advice. The page is written for people comparing options and preparing a real workflow.
High-risk checklist
EU AI Act High-Risk Checklist
Use this checklist when an AI feature may affect safety, rights, eligibility, access, employment, education, public services, law enforcement, migration, courts, or democratic processes.
Disclaimer: This checklist is a practical starting point for internal review. It is not legal advice and does not determine your final regulatory classification.
Step 1: classify the use case
| Question | What to record |
|---|---|
| What is the intended purpose? | Write the product task, user group, decision context, and expected output. |
| Is the system a safety component? | Check whether it is part of a regulated product or safety-related workflow. |
| Does it match an Annex III area? | Screen employment, education, essential services, biometrics, public services, law enforcement, migration, courts, and democratic processes. |
| Does it make or influence consequential decisions? | Note whether people can lose access, ranking, opportunity, services, money, or rights. |
| Can a human meaningfully review the output? | Describe the reviewer role, override path, escalation rule, and timing. |
Step 2: prepare the evidence file
- Risk management process and known foreseeable misuse.
- Training, validation, and testing data description where relevant to your role.
- Accuracy, robustness, cybersecurity, and monitoring notes.
- Human oversight instructions and escalation thresholds.
- Logging or record-keeping approach for important decisions.
- User instructions, limitations, and transparency materials.
- Post-market monitoring or customer feedback process.
Step 3: decide the launch lane
| Result | Recommended product action |
|---|---|
| No high-risk signal | Keep a dated classification memo and still check transparency obligations. |
| Possible Annex III signal | Pause self-serve launch, gather evidence, and ask for specialist review. |
| Clear high-impact decision support | Build human oversight, documentation, monitoring, and customer controls before release. |
| Prohibited-practice signal | Stop the use case and redesign before testing with real users. |
Use the documentation template
Official sources
Last reviewed: July 3, 2026.
- EU AI Act Article 6: high-risk classification rules
- EU AI Act Annex III: high-risk use areas
- EU AI Act Article 9: risk management system
- EU AI Act Article 14: human oversight
- European Commission: guidelines on high-risk AI systems
Updated review note
Review and responsible-use note
EU AI Act High-Risk Checklist is maintained as a practical page for AI compliance self-assessment. Use the result to organize internal review, collect evidence, and decide when a qualified legal or compliance specialist should review the system.
What it does not replace
These tools provide operational checklists and first-pass triage, not legal advice, certification, or a guarantee of EU AI Act compliance.
How to verify the result
The highest-risk scenarios on this site are framed as compliance reviews, not as instructions to build or deploy sensitive AI systems. Useful references include EU AI Act official text and European Commission AI Act overview.
Practical value note
How this page supports the site
EU AI Act High-Risk Checklist gives visitors context about AI Compliance Kit, the workflow boundaries, and how the site's AI compliance review pages should be used. It supports trust by explaining purpose, limitations, contact paths, and review expectations.
Who this is for
EU AI Act High-Risk Checklist is written for founders, product owners, compliance teams, and operators preparing AI systems for review. It assumes the visitor wants to complete a practical task, not browse a decorative landing page.
What to save
The useful result should be a checklist, report, policy draft, questionnaire, or review note that can be copied, downloaded, printed, compared, or used as a next-step working document.
What to check
The output is an operational preparation aid. It should be reviewed against the actual system, data flow, jurisdiction, and qualified legal or compliance advice before production use. Keep the original source material and record the assumptions used for the generated result.
Use this page to understand the site before relying on any generated output or publishing a changed file, policy, or technical configuration. If the result affects a public page, customer-facing workflow, policy decision, or uploaded file, review one sample manually before repeating the workflow in bulk.