Democracy rests on a shared relationship with truth. Citizens need to know that what they read is verifiable, that what persuades them is honest, and that the institutions adjudicating fact and fiction can be trusted. Generative AI now sits between billions of people and the world they are trying to understand. It is, increasingly, the place where knowledge is discovered, beliefs are formed, and political choices are shaped. Three forces are converging at once: 1) open knowledge ecosystems – educational institutions, public-service media, scientific publishing – are being quietly disintermediated as AI returns answers without showing their sources; 2) citizens, particularly young impressionable minds, are exposed to a new generation of systems designed to simulate intimacy, build dependency, and bypass reflection; 3) and the infrastructure on which a healthy public sphere depends is increasingly concentrated in a small number of opaque, commercial hands. This roundtable convenes a focused group to think clearly about what is at stake and what we can do about it inside a 12-month horizon. The conversation builds directly on the Cannes Declaration on the Sovereignty of the Mind, signed at the World AI Cannes Festival 2026, which calls on the international community to: limit AI designs that foster simulated intimacy and emotional manipulation; safeguard democracy against covert behavioral influence; strengthen resilient, values-aligned public-interest infrastructure; and build global red lines through technology diplomacy and cross-border cooperation.