The risk-based approach of the AI Act
An analysis of the innovative regulatory model of the AI Act, which categorizes AI systems into four risk levels - from unacceptable to minimal - ensuring proportional obligations.
Introduction
One of the most innovative aspects of the European regulation on artificial intelligence concerns its regulatory approach.
Instead of applying a single set of rules to all AI systems, the European Union has chosen a model based on the level of risk associated with each application.
This approach was introduced in the Artificial Intelligence Act to ensure that the obligations provided by the regulation are proportionate to the potential impact of artificial intelligence systems.
The four risk categories
The regulation distinguishes four main risk levels.
The first level concerns systems considered to be at unacceptable risk, which are prohibited because they are incompatible with fundamental rights.
The second level includes high-risk systems, which can be used but must comply with very strict requirements.
There is also a category of limited-risk systems, for which lighter obligations are provided, often related to transparency.
Finally, we find minimal-risk systems, which represent the majority of AI applications currently in use and which are not subject to specific obligations.
Why this model is important
The risk-based approach makes it possible to avoid excessive regulation for applications with limited impacts.
At the same time, it allows for the introduction of stricter controls for systems that can influence important decisions in people's lives.
This balance between innovation and protection represents one of the fundamental principles of the European regulation.
Conclusion
Understanding which risk category an AI system falls into is one of the most important steps in assessing the compliance obligations provided by the regulation.
For many organizations, this classification phase represents the first concrete step in the process of adapting to the AI Act.