A brand of BritixoShadow AI. Clear accountability.

Guide · reviewed 21 October 2026

Understand why a particular AI use may be high-risk.

High-risk is a legal category with specific criteria. Begin with a clear description of the system and intended use, then have the relevant reviewers assess the applicable route and exceptions.

By EUAIC editorial team · Updated

There are distinct classification routes.

Article 6 addresses certain product-related AI and the use cases listed in Annex III. The product route includes specific conditions concerning safety components or products and conformity assessment. Annex III addresses listed uses in sensitive areas; it does not classify an entire industry as high-risk.

Some Annex III uses may qualify for the Article 6(3) exception, while profiling of natural persons remains an important qualification. Any exception needs a proper rationale, not a convenient label selected by the project team.

Write down the decision and the people affected.

Describe what the system does in the actual process. Does it prepare information, make a recommendation, rank people or determine an outcome? Who relies on the output, and can they meaningfully challenge it? Distinguish the supplier's intended use from any expansion your organisation proposes.

These questions help reviewers identify the relevant facts. They are not a substitute for the statutory criteria, but they expose the assumptions that a classification decision needs to address.

Keep a reviewable classification rationale.

Record the system and release, the purpose reviewed, the role of the organisation and the evidence considered. Identify the provision or exception examined, the reviewer and any unresolved interpretation. Where the decision is conditional, state the boundaries of the approved use.

  • Factual description of the deployment.
  • Applicable classification route and reasoning.
  • Evidence, reviewer and decision date.
  • Changes that would require reassessment.

Use the result to plan the right work.

A classification decision should lead to a clear next step: a provider programme, deployer controls, further specialist review or another relevant governance route. Check current applicability dates and any product or sector requirements before setting milestones.

EUAIC can help organise the questions, evidence and review workflow. It does not issue a binding classification or conformity certificate. If a purpose changes or new facts emerge, return to the recorded rationale and assess whether the earlier conclusion still applies.

Sources & further reading

Useful answers

Before you take the next step.

Practical details to help you discuss your requirements.

Is a tool high-risk just because it uses a large AI model?

No. The legal assessment concerns the relevant system, use and criteria. Model size or a familiar vendor name does not decide the category.

Does adding a human approval step guarantee an exception?

No. Examine the actual influence on the decision and the statutory conditions. A nominal sign-off is not a substitute for that assessment.

Where can we check the current deadlines?

Use the official Article 113 text and the EUAIC AI Act guide. The current framework includes different application dates for Annex III and product-related high-risk provisions.

Start with the right questions

Tell us what you need to achieve.

Share your current setup, priorities and intended outcome. The Britixo team will help define the scope, responsibilities and next steps.

Discuss your requirementsBritixo MyAccountVerify your email. Send your enquiry. Keep your reference.

Explore services and practical guidance.

Search matching runs in your browser. Press Esc to close. Browse every page ↗