Obligation guides

EU AI Act risk classification

Classification is a decision tree on facts — territorial reach first, then prohibited practices, product-safety high-risk, listed high-risk uses, transparency duties, then minimal-risk.

Run the AI Regulatory Requirements Check →Obligation Atlas →

Territorial scope is not a risk tier

If the EU AI Act does not reach the deployment, the use case is unassessed under the Act — not proven harmless. A contingent high-risk reading can still matter if EU people are added later.

Annex I versus Annex III

AI that is a safety component of a regulated product (medical devices, machinery, vehicles) can follow the Annex I path. Listed use cases such as employment, credit, and insurance pricing follow Annex III.

Article 27 FRIAs attach to specific Annex III deployer shapes — not automatically to every high-risk label, and not to Annex III point 2.

Use the free check for the shape

The AI Regulatory Requirements Check returns a tier label and a one-line reason where an EU nexus exists. It does not paste article text. Compass is where the cited classification and obligations live.

← Obligation Atlas · Start the free check