We shall analyse the notion of autonomy, as applied to artificial systems, and on this basis we shall consider responsibilities for the deployment of autonomous systems, in particular, in the military domain. First, we shall distinguish three dimensions of autonomy: (i) independence, (ii) cognitive skills and (iii) cognitive architecture. With respect to cognitive architecture, we shall further discern the aspects of adaptiveness, teleology and multi-level autonomy. We shall argue that these three dimensions of autonomy are orthogonal, in the sense that a system can exhibit very different levels of autonomy under each dimension. This three-pronged approach will help us to clarify some aspects of the debate concerning autonomy of artificial systems and the regulation of their use. In particular, we shall address the responsibilities and liabilities resulting from the deployment of autonomous systems in the military domain.