With the rapid introduction of generative AI, people are outsourcing more of their intellectual labor to machines. Most ethical reflection is centred around the potentially harmful consequences of AI use on others. In this paper, we take Kant's duties to self, grounded in the required respect towards one's rational nature, as a starting point to complement existing approaches. First, we analyse 'cheating' by claiming authorship over a product or choice produced by AI through the perfect duty not to lie. Second, we argue that the imperfect duty to develop one's capacities entails a moral reason to prevent 'deskilling'. Third, we propose that agents must honour their doxastic responsibility and the need to think for themselves while forming beliefs.