The Last Vote: A Multi-Stakeholder Framework for Language Model Governance

TL;DR

The paper addresses the governance of powerful language models in democratic societies by reframing governance as an optimization problem with democratic integrity as the objective. It introduces a seven-category democratic risk taxonomy and a stakeholder-adaptive Incident Severity Score (ISS) to quantify and signal governance needs from diverse perspectives, paired with a four-phase, six-year implementation that scales from voluntary coordination to binding democratic oversight. The framework formalizes seven stakeholder knowledge categories, supports co-governance with citizen panels, and advocates institutionalized deliberative mechanisms, including sovereignty zones, to distribute epistemic authority. A phased roadmap aligns governance capacity with evolving socio-technical risks, supported by a monitoring-evaluation-adaptation loop that tracks democratic health, socio-economic impact, and governance performance, complemented by transparent dashboards and independent audits. Collectively, the work aims to operationalize democratic ideals in AI governance, offering concrete tools and pathways to enhance democratic resilience against manipulation, misinformation, and systemic governance threats.

Abstract

As artificial intelligence systems become increasingly powerful and pervasive, democratic societies face unprecedented challenges in governing these technologies while preserving core democratic values and institutions. This paper presents a comprehensive framework to address the full spectrum of risks that AI poses to democratic societies. Our approach integrates multi-stakeholder participation, civil society engagement, and existing international governance frameworks while introducing novel mechanisms for risk assessment and institutional adaptation. We propose: (1) a seven-category democratic risk taxonomy extending beyond individual-level harms to capture systemic threats, (2) a stakeholder-adaptive Incident Severity Score (ISS) that incorporates diverse perspectives and context-dependent risk factors, and (3) a phased implementation strategy that acknowledges the complex institutional changes required for effective AI governance.