EU AI Act: Does it apply to you?

Which obligations of the EU AI Act can apply to you

The self-check above shows in a few clicks whether the EU AI Act (KI-Verordnung, KI-VO) applies to you and what to do. The points below are an overview of the possible obligations, from the labelling duty for chatbots and AI images to the high-risk area. For the detailed explanation, see the guide to the EU AI Act.

• Possible high-risk case (staff selection and similar)

  You may be operating in the high-risk area under Annex III of the AI Act. The more demanding obligations (risk management, data quality, human oversight, documentation) only apply from 2 December 2027, thanks to the postponement through the Digital Omnibus. That is enough time to prepare calmly. A short conversation helps to classify your case cleanly.

• Worth checking: does your AI decide about people?

  Whether the high-risk area applies to you comes down to one question: does an AI make decisions about people (hiring, assessment, lending) or substantially prepare them? If so, the high-risk framework applies, though only from December 2027. A short conversation clarifies this quickly.

• Label your chatbot

  Anyone who writes to or speaks with your bot must be able to tell that an AI is answering, unless that is obvious anyway. A clear notice is enough (Art. 50).

• Label deceptively real AI images

  Content that looks like real photos or recordings (so-called deepfakes) must be visibly labelled as AI-generated. A note such as ‘Created with AI’ on the content and in the alt text is enough. The machine-readable marking in the background is done by the AI tool itself. When in doubt, label it (Art. 50).

• Drawings and icons: nothing to do

  Clearly recognisable illustrations, icons, diagrams or graphics do not need to be labelled. They are obviously artificial, no one mistakes them for a real photo.

• Rule of thumb for AI images

  The line is simple: if the image looks like a real photo, it needs an ‘AI-generated’ note. If you can tell it is AI (drawing, icon, graphic), none is needed. When in doubt, label it.

• Build and document AI literacy

  Your team should be able to work with AI, and that belongs in writing (Art. 4). Four steps: take stock of AI use, set rules (which data goes into which tool, who checks what), train people by role, and document the whole thing. The regulation does not require a mandatory certificate, an AI officer or knowledge tests.

This check gives initial guidance and does not replace legal advice. It reflects the status of August 2026 (after the Digital Omnibus). For a binding assessment of your specific case, talk to us or your legal adviser.