31 States Have AI Election Deepfake Laws, But Gaps and First Amendment Challenges Persist
Summary
- • 31 states have AI election deepfake laws as of June 2026; 28 require disclosure, only 3 prohibit outright
- • Most disclosure laws allow deepfakes to keep circulating legally — just with an AI warning label attached
- • Federal judge temporarily blocked Montana's deepfake law on First Amendment grounds in September 2026
- • 164 AI political ads found in 2026 cycle; ~$80M spent; approximately 70% carried no disclosure at all
Details
31 states have AI election deepfake laws as of June 2026 NCSL report
The National Conference of State Legislatures' June 2026 report counted 31 states with laws regulating AI-generated election deepfakes. California and Texas were pioneers, enacting their first such laws in 2019; most other states followed in 2024 or later.
28 states require AI disclosures; only MD, MN, TX prohibit certain deepfakes
Of the 31 states, 28 have disclosure-only laws — deepfakes can circulate with a warning label. Only Maryland, Minnesota, and Texas prohibit certain deepfakes outright, even when they carry an AI disclosure.
Louisiana's June 2026 law requires AI warnings but excludes federal candidates
Louisiana requires AI warnings on campaign ads falsely depicting candidates, but explicitly excludes candidates for federal office — a significant coverage gap. Violations carry up to $2,000 in fines and up to two years in prison.
Maryland's May 2026 law prohibits deceptive deepfakes regardless of AI label
Maryland bans AI-fabricated content that falsely depicts a person in a realistic way; adding an AI disclosure does not make a prohibited deepfake lawful. Violations carry up to $5,000 in fines and up to five years in prison. Removal authority took effect June 1, 2026.
Federal judge blocked Montana's deepfake law in Sept 2026 on First Amendment grounds
A federal court temporarily blocked Montana's election deepfake law in September 2026, finding it likely posed a First Amendment threat to a conservative PAC's free speech rights — highlighting constitutional limits on stricter deepfake prohibition approaches.
164 AI political ads in 2026 cycle; ~$80M in spending; ~70% with no disclosure
The Wesleyan Media Project identified at least 164 AI-created or AI-enhanced political ads by early September 2026, accounting for nearly $80 million in ad spending. About 7 in 10 carried no AI disclosure, though not all necessarily violated a law.
Warning labels reduce belief but don't curb sharing of AI-generated content
A 2025 study of 7,579 Americans found that AI warning labels made people less likely to believe misleading content, but did not significantly reduce their willingness to engage with or share it. A separate study found labels also cast doubt on accurate AI-generated content.
Details on the state-by-state landscape of AI election deepfake laws, enforcement challenges, and 2026 election data.
What This Means
The patchwork of 31 state AI election deepfake laws has significant structural weaknesses: most only require disclosure labels, allowing fabricated political videos to remain in circulation with a warning, and several exclude federal candidates entirely. A federal court's First Amendment block of Montana's stricter prohibition-based law signals that harder regulatory approaches face serious constitutional barriers. With the 2026 midterms underway and nearly $80 million already spent on AI-generated political ads — 70% without any disclosure — voters are being exposed to AI-fabricated content faster than law can respond. The research evidence on disclosure effectiveness also suggests that even where laws work as intended, warning labels may not meaningfully change voter behavior.