Increased time spent by users in virtual environments globally coupled with the choice architectures created by artificial intelligence (AI) systems within these virtual environments means that the AI systems can ‘persuade’ users to take certain actions or abstain from them. In this article, we address this issue through the lens of ‘dark patterns’. Dark patterns are features of the user interface that shape the actions of the users in different ways, getting them to act in ways that they might not act otherwise and that are not necessarily beneficial for them. Recently, regulatory action has been undertaken in various jurisdictions to contain the menace of dark patterns, including in India. But can these measures prevent dark patterns that are powered by AI systems? In this article, we argue that they cannot, as AI-powered dark patterns work at a deep level of behavioural change, making users’ choices appear uninfluenced by manipulation. The laws pertaining to unfair trade practices, transparency and labelling obligations are also not sufficient to deal with these problems. We conclude with an appeal for a holistic approach to AI-powered dark patterns that engage in persuasive behaviour.