Proxy Fairness under the European Data Protection Regulation and the AI Act: A Perspective of Sensitivity and Necessity

Abstract

This paper navigates the convergence of the European Data Protection Regulation and the AI Act within the paradigm of computational methods that operationalise fairness in the absence of demographic data, notably through the use of proxy variables and inferential techniques (Proxy Fairness). Particularly, it explores the legal nature of the data involved in Proxy Fairness under the European Data Protection Regulation, focusing on the legal notion of Sensitivity. Moreover, it examines the lawfulness of processing sensitive personal data for Proxy Fairness purposes under the AI Act, particularly focusing on the legal requirement of Necessity. Through this analysis, the paper aims to shed light on core aspects of the legitimacy of Proxy Fairness in the context of EU law, providing a normative foundation to this line of Fair-AI approaches.