Like the GDPR for personal data, the EU AI Act is a European regulation aimed at governing how companies design, deploy and use artificial intelligence systems. The objective is clear: to enable innovation while limiting risks for users and citizens. Europe is the first to implement such a regulatory framework, in particular by classifying AI systems using a so-called 'risk-based' approach, divided into four risk levels:
1. Minimal risk, such as certain simple chatbots.
2. Limited risk, subject to transparency obligations.

3. High risk, the most regulated and most common in sectors such as banking or insurance. For example, an AI system assigning a credit
risk score can introduce biases, particularly if trained on non-representative data (men/women, minorities, etc.).
4. Unacceptable risk, strictly prohibited, such as person scoring or behavioural prediction systems in the manner of Minority Report. A central challenge of these systems is that of the black box: when it is not understood how a decision was made by the AI. This raises major questions of explainability and traceability between input data and the results produced.