Discriminatory algorithms. A proportionate means of achieving a legitimate aim?

Within the EU legal framework, cases of indirect discrimination are justifiable when suitable and proportionate means are necessary to achieve a legitimate aim. However, the technical reality attached to machine learning (ML) environments challenges this legal postulate from multiple angles. Previous academic writings suggest that, in most scenarios, discriminatory outputs produced by an AI system would meet not only the requirements for indirect discrimination but also the requirements for its justification. This research confirms those views by analysing the relevant case law of the Court of Justice of the European Union (CJEU) to determine the current standing of the prohibition of indirect discrimination. The analysis is complemented by a case study in which it is discussed whether scenarios of proxy discrimination caused by an automatized recruitment tool would pass the CJEU test.