As many leading voices in debates about ethics and AI recognize, AI creates risks for democracy and we need to democratize AI. But what, exactly, are these risks, and what does it mean to democratize AI? Drawing on his forthcoming book Why AI Undermines Democracy And What To Do About It, Mark Coeckelbergh offers an overview of some of the risks and mobilizes the tradition of political-philosophical republicanism to argue that and why AI should be developed in a more participatory way and serve the common good, enabled by education and regulation that create possibilities for non-domination and communication.