The term artificial intelligence system (or AI system) is defined in Article 3(1) of the European Regulation on Artificial Intelligence (AI Act). It is a central concept in the European legal framework for AI, intended to regulate a wide range of technological applications.
An AI system is described as a machine-based system, designed to operate with different levels of autonomy (ranging from systems fully controlled by humans to those capable of making decisions independently) and can demonstrate adaptability after deployment (e.g., by learning from new data or adjusting its behavior).
These systems are capable, for explicit or implicit purposes, of inferred results from the data they receive, such as:
- predictions (e.g., assessing an individual's credit risk),
- recommendations (e.g., suggesting content to a user),
- automated decisions (e.g., accepting or rejecting an application),
- and generated content (e.g., creating images or texts).
These results can influence physical environments (e.g., autonomous vehicles, robots) or virtual environments (e.g., digital platforms, voice assistants).
This definition covers a broad technological spectrum, extending well beyond traditional computer systems. It includes machine learning algorithms, expert systems, natural language processing models (NLP), and generative models such as generative AI (e.g., large language or image models).
The aim of the European legislator is to provide a definition that is sufficiently flexible and technologically neutral to regulate all current and future uses of artificial intelligence, while ensuring a high level of protection of fundamental rights, safety, and transparency.