AI Act classification · step 1 / 5
Link with the EU
The AI Act, the European regulation on artificial intelligence, applies to a Swiss company only if the EU market is touched. Describe an AI use case, then tick what applies: in five steps you get its risk level and your obligations.
The AI Act also concerns Swiss companies
The European AI regulation does not stop at the EU border: it targets anyone placing an AI system on the European market, or whose systems produce results used in the EU. For a Swiss SME the question is therefore not “are we in the EU?” but “are our customers, our users or our systems' outputs?”. The classifier asks these questions in the order of the regulation, use case by use case, because it is the use case, not the company, that is classified.
The four risk levels
- Prohibited practices (Art. 5) manipulation exploiting vulnerabilities, social scoring, emotion recognition at work and so on: to be stopped, with no SME exception.
- High risk (Annex III) recruitment, credit, insurance, education and so on: heavy requirements (risk management, documentation, human oversight), applicable from 2 December 2027.
- Transparency (Art. 50) chatbots, generated content, deepfakes: telling the user that they are talking to a machine or that the content is artificial, since 2 August 2026. Our ready-to-paste notices cover these cases in four languages.
- Minimal risk the majority of office uses: no specific AI Act obligation, but the Swiss FADP still applies to the processing of data.
Frequently asked questions
My company is 100 % Swiss, with no EU subsidiary: am I concerned?
Possibly: the AI Act is extraterritorial (Art. 2). It applies if you place an AI system on the EU market, or if the results produced by your systems are used in the EU; a chatbot reachable by European customers is enough. And Switzerland is preparing its own regulation (draft expected by the end of 2026).
Which deadlines concern an SME?
Since 2 August 2026 the transparency obligations of Art. 50 apply: flagging chatbots and deepfakes as such (machine marking of content, Art. 50(2), has a transition until 2 December 2026 for systems already on the market). The “high-risk” obligations of Annex III were postponed to 2 December 2027 by the “Digital Omnibus” package.
Is this legal advice?
No. The classifier provides dated, sourced information landmarks, each step citing the article it applies. It structures your analysis and saves you time; binding decisions go through your legal adviser.
What do I do with the result?
Each classification can be saved, if you wish, in your AI system register, kept locally in your browser, exportable and printable. It is the inventory an auditor, a large client or an authority will ask you for (Art. 26 AI Act, ISO/IEC 42001).