Redefining democracy for the age of AI: AI governance and the fiduciary turn in the architecture of knowledge

This paper advances a constitutional re-foundation of democracy for the age of artificial intelligence. It argues that democratic legitimacy no longer rests on procedural participation or informational abundance but on fiduciary–epistemic trust—the moral architecture that sustains truthful, reciprocal knowing. Artificial intelligence challenges this foundation not merely through misinformation but through algorithmic clientelism: the systemic conversion of epistemic autonomy into managed dependence within opaque infrastructures of mediation. Integrating fiduciary theory (Frankel 1983; Smith 2023; Gold & Miller 2014; Kahl 2025i), epistemic justice (Fricker 2007; Medina 2013), and hybrid-governance scholarship (Ofir & Levine-Schnur 2025; Lim & Lim 2025), the paper situates AI developers and institutions as epistemic fiduciaries whose legitimacy depends on duties of candour, care, impartiality, and accessibility. It introduces a unifying framework—Fiduciary Epistemic Governance (FEG)—that redefines AI accountability as a constitutional ethic of truth stewardship rather than risk management. The analysis culminates in a set of institutional innovations: Epistemic Fiduciary Entities (EFEs), Public Epistemic Trusts (PETs), and Epistemic Audits, each designed to embed fiduciary openness and relational accountability within AI governance. These mechanisms instantiate what the paper terms the Fiduciary Constitution of Democracy 2.0—a democratic order re-founded on fiduciary reciprocity between human and artificial knowers, where truth is preserved not by control but by trust. Beyond its policy applications, the paper contributes to fiduciary and political theory alike: it extends fiduciary law from property and management into epistemic governance and reconceives democracy itself as a fiduciary-epistemic constitution of knowledge—a living architecture of candour, care, and mutual responsibility.