Congress Revives an AI Election-Deception Bill as $137 Million in AI Ads Flood the Midterms

On October 19, Rep. Shontel Brown, an Ohio Democrat, filed H.R. 10841, the Securing Elections from AI Deception Act, in the House of Representatives. It is a near-identical revival of a bill she first introduced in June 2024, which never got a floor vote, and the new version carries 30 cosponsors. The timing is the story: the bill arrived two weeks before a midterm election in which political advertisers have already spent more on AI-generated content than in any previous U.S. campaign cycle, almost none of it governed by any federal rule.

The bill would make it a federal crime, punishable by a fine or up to a year in prison, to intentionally use AI to deprive someone of the right to vote — including AI-generated falsehoods about polling times, locations, voter eligibility, ballot counting or candidate endorsements. Separately, it would require anyone who publishes AI-generated or AI-altered media simulating a candidate's voice, image or video, or conveying voting information, to attach a disclaimer: either the text "AI Disclaimer: This output was generated by artificial intelligence" or a circled-AI symbol sized like a copyright mark. The Federal Trade Commission would write the implementing rules and enforce violations as unfair or deceptive practices, with disclaimer requirements taking effect 60 days after the FTC finalizes them.

The ads are already here, and most aren't labeled

The bill's reintroduction follows new spending data that undercuts the case that the market is policing itself. Political scientists at the Wesleyan Media Project — Travis Ridout of Washington State, Erika Franklin Fowler of Wesleyan, and Michael Franz of Bowdoin — reported on October 6 that they had tracked roughly $80 million in spending on nearly 170 ads using AI so far in the 2026 cycle, drawing on AdImpact spending data and their own ad-coding team. Republican candidates and allied groups accounted for 80% of those ads and 83% of the spending. Only 31% of the ads carried any AI disclosure. By October 16, outlets citing updated New York Times and Wesleyan figures put the total past $137 million, with more than 80% of those ads still Republican-produced.

The ads themselves are not subtle. In New York, an undisclosed AI ad depicted Mayor Zohran Mamdani and Gov. Kathy Hochul promoting a fake pill called the "political delusion bill," with a narrator warning of side effects including "chronic wallet shrinkage." In Tennessee, Sen. Marsha Blackburn appeared in an AI-generated video dancing with fellow Republican gubernatorial candidates to a Kendrick Lamar track, drawing a rebuke from Democratic nominee Jerri Green. In Kentucky, Rep. Thomas Massie said after losing his primary that AI-generated ads had falsely depicted him dating Reps. Ilhan Omar and Alexandria Ocasio-Cortez — a claim he attributed to pro-Israel donor groups spending against him. Fowler told the Times that without a disclaimer, "it would be hard to know that A.I. was there" at all.

Those disclaimers, where they exist, come entirely from state law, and the coverage is spotty. The National Conference of State Legislatures counts 29 states with AI-deepfake election statutes in effect, ranging from Minnesota's and Texas's pre-election blackout windows to Colorado's and Utah's detailed creator-and-edit-history disclosures. But courts have permanently enjoined California's and Hawaii's versions on First Amendment grounds, and the Wesleyan researchers found almost no correlation between whether a state required disclosure and whether an ad actually carried one: 32% of ads disclosed in states without a law, versus 29% in states with one. A synthetic-video ad targeting Minnesota Democratic Senate candidate Peggy Flanagan ran in May despite that state's ban, with no clear enforcement response.

A federal bill with no floor vote, against a deadline that keeps passing

Congress has had chances to set a national floor and hasn't taken them. Brown's bill failed to advance in 2024, and no companion Senate measure accompanying this revival has surfaced yet. Axios reported in August that House Democrats were privately planning to push election-specific AI legislation "next year if they regain power" — an implicit admission that nothing was moving through the current Republican-controlled chamber before this election. The FTC, which the bill designates as enforcer, has not indicated it has rulemaking capacity queued up for a bill that hasn't passed.

Meanwhile, the security dimension of the problem has become urgent enough that federal agencies are acting without waiting on legislation. The FBI and the Cybersecurity and Infrastructure Security Agency issued a joint alert on October 16 warning that election officials, candidates and political organizations are "attractive targets" for hackers using AI to craft more convincing phishing emails and texts. The alert, routine in format, was notable mainly for what it didn't address: it covers AI as a tool for intrusion, not AI as the content of deceptive political advertising — leaving the ad-disclosure gap entirely to Brown's stalled bill and a patchwork of state laws that researchers say voters can't rely on uniformly even where they exist.

The practical result for the 2026 midterms is that the legal status of an AI-generated attack ad depends almost entirely on which state airs it, whether anyone identifies it as synthetic, and whether a sponsor chooses to say so. With early voting already underway in much of the country, H.R. 10841 is very unlikely to become law before ballots are counted. What it mainly establishes, on the record, is that the authority asking Congress to act identified the same gap researchers are now pricing at well over $100 million.