How the AI Act defines an artificial intelligence system
An analysis of the AI Act's definition of AI system - a deliberately broad definition designed to include a wide range of technologies from machine learning to generative models.
Introduction
One of the first questions many companies ask when they begin to deal with the European regulation on artificial intelligence concerns the definition of AI system.
Not all software technologies automatically fall into this category and not all are subject to the same regulatory obligations.
For this reason, the European Union has introduced a specific definition within the Artificial Intelligence Act.
This definition represents the starting point for determining whether a given technology falls within the scope of the regulation.
The definition provided by the regulation
According to the AI Act, an artificial intelligence system is a system designed to operate with different levels of autonomy and capable of generating outputs such as:
predictions
recommendations
decisions.
These outputs are designed to influence physical or digital environments and can support or automate certain decision-making processes.
The definition therefore includes a wide range of technologies, from machine learning models to more complex systems based on generative models.
Why the definition is so broad
The European Union has chosen a relatively broad definition to prevent new emerging technologies from easily escaping the regulatory framework.
Artificial intelligence is in fact a constantly evolving field and new techniques are rapidly being developed.
A definition that is too restrictive would risk becoming obsolete within a few years.
Conclusion
Understanding whether a technology falls within the definition of AI system is the first step for any compliance assessment.
For many companies, this initial classification phase represents the moment when it becomes necessary to precisely map the AI systems used within the organization.