Sex Discrimination in AI Hiring

Hiring is often presented as the point at which merit is assessed. In AI hiring practice, it is increasingly the point at which algorithms decide what “merit” looks like. This chapter examines AI hiring systems by distinguishing between pre-selection tools and selection tools and demonstrates how each reshapes access to work and generates structural risks of sex discrimination through different technical mechanisms. Pre-selection systems widen the evidential base from the résumé to large-scale behavioural and proxy data, so that irrelevant signals can become selection criteria through correlation with predominantly male “model” employees. Selection stage systems go further by inferring emotions and personality traits from video and game data, despite contested scientific foundations and uneven performance across groups, making discrimination harder to detect, contest, or even verify. The chapter argues that existing UK equality law remains oriented toward identifiable provisions, criteria, or practices, while data protection law offers only partial control, particularly where inferences fall outside effective notice, access, and rectification. The core regulatory priorities are therefore to constrain non-job-related data points and to bring algorithmic inference within enforceable safeguards.