Register · Classify
The first step in EU AI Act compliance is a register: a list of the AI your company uses. The second is knowing which rules apply to each one. Normis does both - tick your tools from a ready-made catalogue, then get a plain-English read on every obligation, with the article reference beside it.
Who it is for
If your company uses any AI - and most do, often hidden inside everyday SaaS - then yes, an inventory is the starting point, whatever your size. You do not need a compliance officer or a consultant's fortnight to build one.
You almost certainly have AI to register if you use:
Most of what an SME uses turns out to be minimal or limited risk. The point of the register and classification is to prove that calmly, and to catch the one or two tools that genuinely need action, like a CV-screening tool that counts as high-risk.
A complete register is the foundation everything else sits on - your policy, your classifications, your evidence pack.
For each entry, a short plain-English questionnaire works out the risk class. You confirm the result, with our reasoning and the article references shown.
The public classifier is the same engine, open to anyone. Answer a handful of questions about one tool and get the risk bucket, the obligations and the article references.
How it works
Tick from the catalogue or add your own.
Per tool, in plain English.
Prohibited, high-risk, transparency-only or minimal, with reasoning and article references.
The register and classifications feed your policy and your evidence pack.
Yes - it is the practical first step for any company using AI, and it is what auditors and enterprise buyers expect to see. Without it you cannot say which rules apply or prove you have looked. The good news is most entries turn out to be low-risk.
More than you might think: standalone tools like ChatGPT and Copilot, but also AI features baked into your CRM, helpdesk, HR and marketing software. If a tool makes predictions, generates content or scores people, register it.
Almost never. Most AI an SME uses is minimal or limited risk. Classification is mainly there to give you that reassurance in writing, and to flag the rare tool - recruitment, credit, biometric uses - that genuinely needs action.
The free public classifiers (including ours) give you a one-off answer. Normis keeps a living register, stores the reasoning and rules version behind each classification, and feeds it into a policy and a sealed evidence pack you can actually show someone.
They are written against the text of the AI Act, and every rule cites the article or annex behind it so you can check it yourself. Each classification records which rules version produced it, and if the rules are updated, old records keep the version they were made under.
No. It is a plain-English system whose rules cite the article behind each answer. For a genuinely borderline or high-stakes classification, take legal advice - Normis gives you the documented starting point.
The classifier is free, no signup, no card. List a tool, get a clear answer, and start your register the moment you are ready.