In order to be counted as autonomous, a weapons system must perform the critical functions of target selection and engagement without any intervention by human operators. Human rights organizations, as well as a growing number of States, have been arguing for banning weapons systems satisfying this condition – that are usually referred to as autonomous weapons system (AWS) in this account – and for maintaining meaningful human control (MHC) over any weapons systems. This twofold goal has been pursued by leveraging on ethical and legal arguments, which spell out a variety of deontological or consequentialist reasons. Roughly speaking, deontological arguments support the conclusion that by deploying AWS one is likely or even bound to violate moral and legal obligations of special sorts of agents (military commanders and operators) or moral and legal rights of special sorts of patients (AWS potential victims). Consequentialist arguments substantiate the conclusion that prohibiting AWS is expected to protect peace and security, thereby enhancing collective human welfare, more effectively than the incompatible choice of permitting their use. Contrary to a widespread view, this paper argues that deontological and consequentialist reasons can be coherently combined so as to provide mutually reinforcing ethical and legal reasons for banning AWS. To this end, a confluence model is set forth that enables one to solve potential conflicts between these two approaches by prioritizing deontological arguments over consequentialist ones. Finally, it is maintained that the proposed confluence model significantly bears on the issue of what it is to exercise genuine MHC on existing and future AWS. Indeed, full autonomy is allowed by the confluence model in the case of some anti-materiel defensive AWS; it is to be curbed instead in the case of both lethal AWS and future AWS which may seriously jeopardize peace and stability.