Artificial Intelligence Systems in the European Union: Guidelines and Architectures for Compliance-by-Design
Résumé
The Artificial Intelligence (AI) Act, published in July 2024, represents a pioneering legislative effort to regulate artificial intelligence within the European Union (EU). Aimed at fostering innovation while ensuring safety and funda-mental rights, the AI Act seeks to create a comprehensive legal framework for developing, deploying, and using AI systems. However, the requirements, guide-lines, and best practices mandated in the AI Act are spread amongst various articles and annexes. They are intermingled with legal and sometimes financial issues, making their reading cumbersome for engineers and computer scientists. In addition, the AI Act is not selfcontained, as it delegates several issues to other EU legislation, like the Cyber Resilience Act, including technical details. In this paper, we propose an indepth analysis of both acts, from which we extracted a clear list of technical and organizational requirements that can be used, primarily by developers, to produce systems compliant with these regulations. The requirements are presented in tables, organized according to their categories, and have a workflow to navigate them quickly. We also provide generic architectures to help developers design fully compliant solutions with the AI Act and Cyber Resilience Act.
Origine | Fichiers produits par l'(les) auteur(s) |
---|