Power2Idea AI

EU AI Act High-Risk Systems: Annex III signals and how to triage

A practical guide to Annex III high-risk signals and how to triage an AI system using deterministic questions and evidence.

Thu Jan 22 2026 • EU AI Act, High-risk, Annex III, Compliance, Risk

Quick triage

Run your system spec here:

  • /eu-ai-act

What “high-risk” means in practice

Most teams don’t need the full legal program on day 1. They need an early signal:

  • Are we in a domain likely covered by Annex III?
  • Do we need a high-risk compliance track?

Common Annex III signals (v1)

Examples of question buckets:

  • Biometrics
  • Critical infrastructure
  • Education / vocational training
  • Employment / recruitment / worker management
  • Essential services / benefits
  • Law enforcement
  • Migration / asylum / border control
  • Justice / democratic processes
EU AI Act High-Risk Systems: Annex III signals and how to triage — architecture diagram 1
A practical guide to Annex III high-risk signals and how to triage an AI system using deterministic questions and evidence.
Mermaid source
flowchart LR\n  Q[Annex III questions] --> A[Yes/No/Unknown]\n  A --> E[Deterministic evaluation]\n  E --> R[High-risk signal?]

What to collect as evidence

To reduce “UNKNOWN” outcomes, explicitly state:

  • Deployment context (where and by whom)
  • What decisions are made (or supported)
  • Whether the system identifies or categorizes people
  • Whether it affects access to services / employment / education

Next guide

  • Prohibited practices → /policy-guides/eu-ai-act-prohibited