Andrew Sorota, Josh Hendler

AI is becoming the primary interface through which people form political beliefs and participate in democratic governance, a shift comparable to how the printing press and broadcast media reshaped society. The authors identify three critical layers of risk: the epistemic layer (how people learn what is true), the agency layer (how AI agents act on users' behalf in civic matters), and the collective layer (how millions of agents interacting could produce unintended polarization). Without deliberate design choices, this transition could strain already fragile democratic institutions, though thoughtful implementation could address longstanding problems like civic disengagement and polarization.