On Thursday 6 February, the European Commission published its guidelines on artificial intelligence (AI) systems, which have been eagerly awaited since the introduction of the AI Act, and which come shortly after the guidelines on prohibited practices in artificial intelligence (see EUROPE 13572/6).
The document is intended to help suppliers, market players and various stakeholders to determine whether a system constitutes an “AI system” within the meaning of the AI Act.
Like the guidelines on prohibited practices, the guidelines on AI systems list and detail the legal terms included in the articles of the regulation, the main concepts, the elements included in the definition of “AI system” and the interaction between the regulation and various other European texts.
The document explains, for example, the definition of the “objectives” specific to AI systems, the different ways in which AI “learns”, and the system’s direct “interaction with the environment” – through the manipulation of a robot or the management of an algorithm.
These 25 pages of non-binding clarifications are intended to help suppliers and operators comply with the spirit of the regulation, which has already been criticised for its “complex and cumbersome procedures” and “very short deadlines”, which “are holding back European innovation” (see EUROPE 13560/5).
See the document: https://aeur.eu/f/fe3 (Original version in French by Isalia Stieffatre)