Owing to an old Cartesian intuition, the concepts mechanism and autonomy are commonly juxtaposed in an antagonistic way. This intuition has been a long-standing mainstay for those looking to articulate what is distinctive about human cognition and rationality, as well as a serious theoretical obstacle for those looking to reproduce it. We aim to show that these concepts are not antagonistic, and therefore that the Cartesian intuition is a faulty one. Part of our argument turns on the premise these two concepts are integrated by their common focus on control systems, which are composed of various types of mechanisms that make it possible to achieve and maintain autonomy. In particular, we suggest that—contra the Cartesian intuition—degrees of autonomy are positively correlated with the degrees of complexity of control systems. We end with a proposal to the effect that research in AI would do well to incorporate biomimetic strategies that invoke models of allostatic mechanisms as a way of understanding how to enhance autonomy in artificial systems. Robots can achieve a better autonomy by replacing the homeostatic regulatory system (already implemented as we’ll see) with a new form of allostatic mechanism of control called here “allostats”.