Where AI is used to make or inform employment decisions, such as hiring, promotion, performance assessment, or pay determinations, there is a material risk that the system will unfairly discriminate, directly or indirectly, against persons on the basis of protected grounds.