Commission Guidelines on the definition of an artificial intelligence system

The European Commission's guidelines on defining an AI system aim to help providers and stakeholders identify whether a computer system qualifies as an AI system under the AI Act. The definition of an AI system includes several key elements.

Firstly, the system must be developed and operate on machines, including both hardware and software, and it must be capable of functioning with varying levels of autonomy, meaning it can operate with some degree of independence from human intervention.

Another important aspect is post-implementation adaptability, which means the system can learn and improve its functionality over time after being deployed. The system must also be designed to achieve specific objectives, which can be explicit (clearly defined) or implicit (inferred from its behavior).

The system must be able to generate outputs, such as predictions, content, recommendations, or decisions, based on the inputs it receives. These outputs must have the ability to influence physical or virtual environments. Finally, the system must be able to interact with its environment, adjusting its behavior according to changing conditions.

The guidelines are non-binding and are designed to evolve over time, adapting to new technological developments and use cases. The document emphasizes the importance of evaluating each system based on its specific characteristics and avoiding a mechanical application of the definition.