As artificial intelligence (AI) is set to disrupt all areas of life, militaries worldwide are investing heavily in the research and development of military AI applications. Among the most controversial are autonomous weapons systems (AWS), which are capable of selecting and engaging targets without human intervention. AWS range from existing technologies, such as autonomous anti-radar devices, to speculative future capabilities, including drone swarms operating in complex urban combat environments. There are clear reasons why regulating AWS may be desirable. Ethically, removing human agency from lethal decisions risks eroding core humanitarian principles, such as moral judgement and human dignity in armed conflict. Legally, delegating target selection and use of lethal force to machines without human intervention raises questions of accountability and compliance with International Humanitarian Law. Strategically, the widespread proliferation of AWS may increase the risk of inadvertent escalation and accelerate a global arms race, driven by the opacity of underlying algorithms and competitive pressure among States. Against this backdrop, this chapter adopts an interdisciplinary approach, combining international relations, arms control, AI ethics, international law, and computer science. It evaluates the limits of multilateral regulation and argues that a dual-track approach may serve as a necessary last resort: combining national legislation and industry self-regulation to achieve coalition-building among like-minded States. Taking into account the inherent properties of AWS and lessons from historical arms control, the chapter explores how convergence around responsible practices may be achieved through real-world application, even in the absence of a binding international agreement.