Why first attempts stall.
What your first assessment leaves you with.
Three first passes.
A 30-person agency listing its AI tools for the first time.
- Eleven tools found; seven were on nobody's list.
- Nine are limited or minimal risk: Art. 50 and Art. 4 only.
- Two need a closer look: a client scorer and an HR screener.
A scale-up that thought it had no EU AI Act exposure.
- The bought recruitment tool lands in Annex III point 4.
- Deployer duties apply: oversight, logs, the worker notice.
- Their own product feature makes them a provider for it.
A public body preparing for scrutiny.
- Art. 27 binds: a fundamental-rights assessment is required.
- It builds on the existing DPIA and feeds the DPIA view.
- Affected people informed under Art. 26(11), the notice dated.
What follows the first pass.
Questions about your first EU AI Act risk assessment.
Where does a first assessment actually begin?
With an inventory; the policy comes later. You cannot classify what you have not listed, and most organisations find more AI than they thought.
How long does it take?
The free check asks a short set of questions per system; a first pass over a small estate takes an afternoon. The evidence behind the obligations takes longer, and that is where deadlines bite.
What if most of our systems turn out to be low risk?
That is common, and useful: if eight of ten systems carry only Article 50 transparency and Article 4 literacy, the effort goes to the two that carry more.
Do we need to classify systems we are only piloting?
The EU AI Act attaches duties to putting a system into service, so a genuine internal pilot sits differently. Recording it now costs little and keeps the register right when it goes live.
What comes out at the end?
A register of systems, each with a role, a risk class and obligations citing their articles, plus a classification record you can reopen and a dated decision trail.