Context: In previous papers, I suggested six rules proposed by Varela, Maturana and Uribe as a validation test to assess the autopoietic nature of a complex dynamic system. Identifying possible non-biological autopoietic systems is harder than merely assessing self-organization, existence of embodied boundaries and some observable autonomous behavioural capabilities: any rigorous assessment should include a close observation of the “intra-boundaries” phenomenology in terms of components’ self-production, their spatial distribution and the temporal occurrence of interaction events. Problem: Under which physical and components’ relational conditions can some social systems be properly considered as autopoietic unities compliant with the six rules? Results: Dynamic systems can be classified according to “degrees of autonomous behaviour” that they may acquire as a result of the emergence of organizational closure (i.e., autonomy. Also, the different “degrees of attainable systemic autonomy” depend on the “degrees of autonomy” shown by a system’s dynamic components. For human social systems, a necessary balance between individuals’ autonomy and the heteronomous behaviour brought about on people by social norms (laws, culture, tradition or coercion) sets limits to the “degree of systemic autonomy” that human organizations may acquire. Therefore social systems, defined as dynamic systems composed of physical agents, could not attain the high “levels of systemic autonomy” ascribable to autopoietic systems without constraining the autonomy of agents to “levels” that are incompatible with spontaneous human behaviour. Also, social organizations seen as composed of physical agents interacting in physical space cannot be construed as autopoietic systems. Alternatively, if seen as composed of “process-like” entities, where agents participate as actors within processes, some social systems could be described as autopoietic wholes existing in the abstract space in which we distinguish interactions between processes, provided that we can assess compliance with the rules for some specific cases. Implications: These conclusions contribute to the debate on the possible autopoietic nature of some human social systems and to grasping the opportunity to shift focus to the more interesting issue of the “degrees of systemic autonomy” that human organizations could acquire (if needed) without imposing unbearable limitations on the autonomy of human actors. Also, the conceptual framework of this explanatory approach could be used in practical terms to assist the development of new dynamic modelling languages capable of simulating social systems.