Nathaniel Persily is the James B. McClatchy Professor of Law at Stanford Law School and Co-Director of the Stanford HAI Cyber Policy Center. He works on election law, social-media and platform governance, and the intersection of AI with democracy, and is co-editor of the Digitalist Papers (Vol. 1 and Vol. 2). His essay "Misunderstanding AI's Democracy Problem" (Digitalist Papers Vol. 1) develops the Liar's Dividend thesis as applied to AI.
Background and election-law scholarship
Persily's broader scholarly focus is election law and election administration, including writing on redistricting, voting rights, campaign finance, and voter ID. He frames the AI-democracy work as an extension of this longer-running concern, treating AI as the latest incarnation of technology-mediated questions about election integrity.
AI and democracy
In "Misunderstanding AI's Democracy Problem" (Digitalist Papers Vol. 1), Persily argues that direct persuasion by deepfakes is mostly a tail problem, and that the more consequential harm is a broad erosion of trust in authentic content — the dynamic associated with the Liar's Dividend. He holds that fact-checking and provenance approaches such as C2PA are necessary but insufficient, on the grounds that technical mitigation alone cannot address the emotional and psychological dimensions of mistrust.
The essay sets out four macroprinciples for AI governance: that societies cannot "tech their way out" of the problem (watermarking alone fails); that enforcement and administration matter more than stated principles; that a civil-society auditing ecosystem must support government; and that public compute is a prerequisite for outside accountability.
Platform-transparency research
Persily's earlier, pre-AI-era research on social-media platform transparency anchors his AI-era recommendations. He was the principal architect of the Facebook Open Research Initiative (2018), which sought to give academic researchers access to platform data for election-integrity research. He cites the experience of that program, including its difficulties, in arguing that government-mandated transparency with vetted researcher access is essential.
Positions and tensions
Persily advocates operationally focused rules — mandatory transparency, public compute for research, and clear accountability for AI-platform decisions with election relevance — while rejecting both broad deregulation and what he characterizes as naive reliance on watermarking. He argues that the empirical record of recent elections (India, Indonesia, the EU, the UK, and France) undercuts narratives positing an existential deepfake threat.
His position runs against that of John Cochrane, whose Digitalist Papers Vol. 1 essay "Just Relax" takes the opposite regulatory posture (Vol. 1 "Just Relax"). He partially agrees with Eugene Volokh's Vol. 1 essay "Generative AI and Political Power" on user-sovereignty concerns but disagrees on the role of government regulation (Vol. 1 "Generative AI and Political Power").
Relationships
- supports: The Digitalist Papers (Stanford, Volumes 1–2) / Liar's Dividend — his essay and the concept it develops.
- supports: AI in Elections and Democratic Institutions / Synthetic Media / Deepfakes / Synthetic Content: Exploring the Risks, Technical Approaches, and Regulatory Responses — adjacent concepts.
- supports: ChatGPT, Can You Solve the Content Moderation Dilemma? — related territory.
- related: Stanford HAI — institutional affiliation.
- related: AI Snake Oil — Narayanan and Kapoor (2024) — parallel empirical argument against deepfake-panic framings.