State Laws Struggle to Keep Pace With AI Election Deepfakes Ahead of Midterms

State legislatures across the United States are racing to regulate artificial intelligence in political campaigns as the 2026 midterm elections approach, but a patchwork of state laws, enforcement gaps, and constitutional challenges are limiting their impact. According to reports from Fast Company, The Conversation, and Yahoo News, 31 states had enacted election deepfake laws by June 2026.

While California and Texas enacted the nation's first election deepfake laws back in 2019, the vast majority of states rushed to pass legislation in 2024 or later. Stefani Langehennig, an AI policy scholar and assistant professor of practice at the University of Denver, outlined how these state frameworks vary widely in approach and stringency.

How do state laws handle AI election deepfakes?

A June 2026 report by the National Conference of State Legislatures found that 28 of the 31 states with deepfake laws focus on mandatory disclosures, while just three states—Maryland, Minnesota, and Texas—prohibit certain election deepfakes outright, even if an AI warning is attached.

State-level enforcement mechanisms and penalties also vary significantly:

• Louisiana (June 2026 law): Requires clear warnings on certain printed, online, and broadcast campaign ads that falsely depict candidates or urge voting behavior. Violations for state and local candidates carry fines up to $2,000, up to two years in prison, or both, enforced by local district attorneys under state attorney general supervision.
• Louisiana (May 2026 law): Requires AI disclosures at the beginning of campaign robocalls reproducing a public figure's voice, enforced by the state board of ethics with civil fines up to $2,500 for a first violation and $5,000 for subsequent violations.
• Maryland (May 2026 law / June 1, 2026 effective date): Prohibits deceptive election deepfakes when created with knowing or reckless intent to produce materially false information, alongside actual or intended harm to a voter or ballot petition. Convictions carry fines up to $5,000, up to five years in prison, or both.

How prevalent is AI in political advertising?

Political campaigns are deploying AI tools on a massive scale despite the growing web of regulations. The Wesleyan Media Project identified at least 164 political ads nationwide that were created or enhanced with AI during the 2026 election cycle by early September 2026. These ads accounted for nearly $80 million in total spending, with approximately 7 in 10 carrying no AI disclosure whatsoever.

At the federal level, instances like a 2023 Republican National Committee ad attacking former President Joe Biden utilized explicit warnings, labeling the spot as built entirely with AI imagery depicting an imagined future.

What are the enforcement hurdles and legal challenges?

Legal scholars and researchers point to substantial ambiguities in how these laws operate in practice. Maryland's statute requires proof of "actual or intended harm" to a voter, potential voter, or ballot petition, but the law does not define what constitutes that harm. Meanwhile, the Wesleyan Media Project count of $80 million in AI-enhanced spending includes general uses beyond deceptive impersonations, leaving it unclear how many specific ads violated a law or actually misled voters.

Constitutional hurdles have also emerged. In September 2026, a federal judge temporarily barred Montana from enforcing its deepfake law, ruling that the statute likely threatened a conservative PAC's First Amendment free speech rights.

Academic research further questions the efficacy of warning labels themselves. A 2025 study involving 7,579 Americans tested warning labels on misleading AI images and found that while viewers were less likely to believe the claims, the labels did little to alter users' willingness to share or engage with the content. Another recent study found that labeling headlines as AI-generated caused people to rate them as less accurate regardless of their actual truth.

This is educational information, not legal, financial, tax, or investment advice.