The immediate risk of AI is not a sensational conflict with malevolent superintelligence and killer robots. It is the risk of AI destabilizing the current political-economic equilibrium, radically altering the balance of power in society, and tearing up the social fabric. This talk describes how such a process might unfold and proposes policies and institutions for managing the political and economic risks of AI. Competition policy emerges as a critical tool for risk management, as does strengthening democratic and egalitarian institutions. Without appropriate policies, there will be no AI-driven growth take-off and the scale of inequality that would emerge would dwarf anything experienced in the twentieth century.