States split on deepfakes: NCSL warns many laws face First Amendment tests
Loading...
Summary
NCSL told Nevada lawmakers most states opt for disclosure requirements for AI-generated political media, while a minority use prohibitions; recent cases in California and Hawaii have resulted in injunctions against broadly worded laws.
The National Conference of State Legislatures told Nevada’s interim committee that states have taken different routes to regulate AI-generated campaign content — with the majority using disclosure requirements and a few adopting election‑period prohibitions.
Adam Cook, policy associate with NCSL’s elections and redistricting program, told members that as of 2026 roughly two dozen states have enacted laws addressing deepfakes in campaigning, most requiring an on‑media disclosure such as "this media was generated by AI." A smaller number of states (for example, Texas and Minnesota) restrict publication of deceptive AI media within a set period before an election.
Cook highlighted constitutional hurdles: courts have already enjoined laws in California (Coles v. Bonta) and Hawaii (Babylon v. Lopez) for imposing overly broad or vague restrictions that failed strict scrutiny in the view of the district courts. Those rulings suggest states must narrowly target demonstrable harms such as voter intimidation or coercion rather than regulating based on speculative effects to electoral prospects.
Cook also described a few legislative designs that may help laws withstand challenge: limiting who can sue (for example, only candidates or officials who suffered concrete harm), exempting clear satire or journalistic republication, including metadata disclosure requirements, and combining platform reporting obligations with avenues for injunctive relief.
On election administration uses of AI, NCSL said most current uses are operational (drafting text, ballot proofing, chatbots for procedural manuals) and recommended guardrails: funding for secure tools, human review of outputs, and policy frameworks that inventory and restrict sensitive data use.
Lawmakers asked about foreign actors and platform detection capabilities; Cook said state laws typically do not address foreign actors directly, and platform detection tools are emerging but not yet a substitute for tailored statutes and enforcement frameworks.
