I discuss Kant’s conception of conscience in the context of his conceptions of two phenomena central to Kant’s practical philosophy: ordinary moral-cognition and self-deception. Conscience in Kant has the role of a reflective watch-dog: It watches over or reflects about how cautiously an agent uses her rational capacities when she reasons about concrete matters of duty. In my paper I discuss a problem that has not yet received its due attention. Conscience can warn agents every time they are about to transgress the moral law and it retrospectively reproaches agents for transgressions, yet Kant believes that there is also a natural propensity to self-deception. Self-deception, however, is only possible if agents can successfully hide from themselves the moral implications of some of their actions or present them as morally innocent. I argue that Kant’s model of an internal court of conscience is inadequate to account for the pervasive threat self-deception poses. I propose a modified model instead, which I call “internal-panel-model”.