This Element presents challenges of attributing responsibility for AI harms through political theories of structural injustice, where fault is broadly diffused and difficult to establish. I argue that AI systems feature complex structural entanglements between humans and machines-in computational architectures, organizational workflows, and societal data collectives-that pose problems for conventional liability where direct causal links are necessary for blame. I develop two modes of structural responsibility that address unjust AI outcomes from these entanglements. Thin structural responsibility traces where power and accountability lie in the processes of developing unjust AI systems. Thick structural responsibility enlists deployers, users, and broader publics alongside developers in collectively repairing unjust social conditions that produce and rely on AI systems. I show that many Responsible AI initiatives remain thin by demanding disclosures without systemic repair. Incorporating thicker reforms makes responsibility for just AI a living political practice rather than a managerial task. This title is also available as open access on Cambridge Core.