2 December 2027 — stand-alone high-risk AI systems (Annex III) must be compliant
EU AI Act risk levels and self-check
The four EU AI Act risk levels and the questions that place a system in one of them. Automated classification is switched off on this site, so this page does not classify your system.
4
Risk Levels
47
Obligations for High-Risk
€35M or 7%
Revenue max penalty
The 4 Risk Levels
Unacceptable Risk
Banned — social scoring, real-time biometric ID, manipulation
High Risk
HR, credit, healthcare, law enforcement, critical infrastructure
Limited Risk
Chatbots, deep fakes, emotion recognition — transparency required
Minimal Risk
Spam filters, games, inventory — no specific requirements
Automated classification is switched off
This page does not send a description to an AI model and does not return a risk level. To place a system yourself, answer these questions and compare the answers with the four risk levels above.
- Which sector is the system used in?
- Does it make or assist decisions about individuals?
- How automated is the decision, and who reviews it?
- Could errors cause significant harm?
- Is it public-facing, B2B or internal?
Your answers give a first reading, not a legal classification. The final determination needs the text of Regulation (EU) 2024/1689 and qualified counsel.