Neo-Kantian attempts of grounding morality in the necessary requirements of agency face the problem of “bad action.” The most prominent example is Christine Korsgaard’s account which considers the categorical imperative to be indispensable for an agent’s self-constitution. I will (based on a specific reading of Kant’s Groundwork) suggest a constitutive account that can solve the problem of bad action by applying the distinction between constitutive and regulative rules to the different formulas of the categorical imperative. The result is that an autonomous agent can violate the categorical imperative in so far as it amounts to a regulative rule of morality; however, an agent cannot call into question the categorical imperative as a constitutive rule of the practice of morality without losing her or his identity as a moral agent