This chapter critically examines two competing frameworks that shape the way social robots are understood and ethically assessed: the extended mind hypothesis and the co-evolutionary perspective. While the extended mind approach treats robots as external supports that extend individual cognition, the co-evolutionary framework conceives of them as participants in relational ecologies of mutual adaptation. We argue that these epistemological alternatives prefigure divergent ethical orientations. The extended mind view sustains a protective and distrustful ethics focused on the issues of deception, substitution, and harm reduction. The co-evolutionary view grounds a relational ethics oriented toward responsibility and social sustainability—the capacity to preserve and foster human social bonds in the midst of technological change. Building on the co-evolutionary framework, we develop the proposal of a synthetic ethics: an experimental and constructive paradigm that evaluates human–robot relations not in terms of imitation or authenticity, but according to their contribution to relational coordination and collective resilience. Finally, we extend this approach to the integration of large language models into social robots, which intensifies both the risks and the possibilities for sustainability. We argue that synthetic ethics provides a timely framework for guiding this transition, emphasizing relationality, transparency, and responsibility in shaping the futures of human–robot coexistence.