Fairness in AI systems development : EU AI Act compliance and beyond

Abstract

Rapid popularisation of artificial intelligence (AI) has accelerated initiatives for ethical AI development. In the European Union (EU), the Artificial Intelligence Act (AIA) entered into force on the 1st of August 2024, which has steered the focus in many organisations towards compliance. As the AIA is not an ethics guideline, it is reasonable to assume that measures beyond compliance are required for ethical AI systems development. To help unravel what is already covered by the AIA and what is not, this paper studies the premise that the AIA lays out for ethical AI systems development. Drawing from critical theory and using John Rawls's theory of justice, the paper shows how the AIA provides limited support for basic liberties, equality of opportunity and the least advantaged members of society, which calls for attention concerning ethical reflection in the AI system lifecycle to ensure ethical AI development. As a result, a framework is given to lay out ethical considerations AI providers should include in iterative AI development process to steer the development towards justice as fairness.