How to Detect and Measure the AI Dangers to Democracy

AI amplifies existing democratic vulnerabilities rather than creating entirely new ones. The paper finds that across information ecosystems, elections, and public administration, AI systems primarily exacerbate pre-existing democratic problems rather than introducing fundamentally novel threats.

Accountability gaps emerge when citizens cannot monitor AI systems making decisions on their behalf. Using principal-agent theory, the authors identify that democratic breakdown occurs when voters and citizens delegate authority to AI systems and their providers without adequate ability to oversee operations or understand outputs—creating a "delegation problem" that undermines democratic control.

Institutional assessability is essential for maintaining democratic integrity. The framework proposes that the central condition for protecting democracy is ensuring that AI systems and their decision-making processes remain sufficiently transparent and measurable so that democratic institutions can evaluate their impact and hold them accountable.

A systematic framework can help prioritize and compare AI risks to democracy across different domains. By operationalizing NIST's seven characteristics of trustworthy AI into measurable indicators, the paper provides a method to compare risks in information ecosystems, elections, and public administration—helping policymakers identify where democratic control is most likely to fail.