The EU AI Act: Four Tiers of Risk

The EU AI Act (Regulation 2024/1689) is the first comprehensive AI law — and its architecture is a risk pyramid, not a prohibition:

• Unacceptable risk — banned outright (Art. 5): a short list of practices considered incompatible with EU values
• High risk — heavily regulated: a defined set of use cases that trigger the Act's full compliance machinery
• Limited risk — transparency only: systems like chatbots and deepfakes that must disclose what they are
• Minimal risk — unregulated: the vast majority of AI — spam filters, game AI, most recommenders — carries no new obligations

The practical consequence: most AI applications are untouched; the entire compliance weight concentrates on the high-risk tier and on general-purpose AI providers.