The November midterm elections will be the first national election in which generative artificial intelligence is cheap, convincing, and carried in everyone’s pocket. The concerns being discussed are not hypothetical.
The Los Angeles County’s registrar has reported that AI was used during California’s June primary to manipulate publicly available footage of ballot counting to gin up fraud claims, and researchers warn that fabricated audio and video are far more persuasive than a bare assertion. Election officials now worry less about a fake candidate ad than about fake evidence, manufactured proof for whatever conspiracy theory is trending on election night.
Recommended Stories
Those concerns are legitimate, and taking them seriously is exactly why the kneejerk reflexes now sweeping statehouses — that is, to define then ban the technology — should be resisted. Thirty-one states have enacted election deepfake laws, and the midterm elections will test them all. Many will fail, either in practice or in court, because they aim at the wrong target.
Start with the practical problem. Artificial intelligence is not a thing that can be pinned down in a statute. AI is a moving family of techniques, from decision trees to the neural networks behind today’s chatbots. Define it narrowly, and bad actors simply switch tools; a law aimed at neural networks is evaded with a different algorithm that produces the same lie. Define it broadly, and Photoshop, video editing, and autocomplete are swept in. Either way, the statute is worthless before the ink dries, because the technology changes faster than any legislature ever will.
Next, the constitutional problem. Laws that prohibit a technology rather than a harm are content-based restrictions on political speech, the most protected speech there is, and courts are treating them accordingly. A federal court struck down and permanently enjoined California’s ambitious deepfake ban last year, and Hawaii’s met the same fate in January. Narrow disclosure rules have generally survived; technology prohibitions keep failing and so protect no one. Legislating against a tool invites constitutional wreckage. Legislating against deception endures.
Principle should govern: nothing that is legal without AI should become illegal because AI was used, and nothing that is already illegal becomes licit because AI did it.
Regulate the impermissible use, not the technology. The feared evils from AI in elections are old evils merely wearing new clothes. Impersonating a candidate, faking an endorsement, lying to voters about when and where to vote, hiding who paid for a political message, every one of these was unlawful before ChatGPT existed. Spreading false voting information is prosecuted under a federal conspiracy statute more than a century old, and all 50 states have their own equivalents.
Need proof? Just consider New Hampshire. Days before the January 2024 primary, thousands of voters received robocalls in an AI-cloned Joe Biden voice urging them to stay home. The consultant behind the scheme was indicted under existing state laws against voter suppression and candidate impersonation, and the FCC finalized a $6 million fine under existing caller ID spoofing rules. No AI statute was required to reach the conduct. The law focused on the deception, and the deception was enough.
So, what should lawmakers do instead? Affirm, where useful, that AI is not a loophole, so no defendant can argue that a machine-made lie is not a lie. Apply disclosure rules for political advertising neutrally, so an ad follows the same transparency requirements whether a human or algorithm produced it.
Election officials should keep publishing plain-language explainers on audits, recounts, and counting timelines before election day, so that when the fake evidence arrives, the accurate answer is already in circulation. And AI companies should enforce their own terms of service against electoral impersonation, because the private sector can move faster than any prosecutor.
THE ‘DANGEROUS’ AI MODELS ARE THE ONES SAVING US
One more dose of harsh truth. David Becker of the nonpartisan Center for Election Innovation & Research cautions that AI is not radically changing the misinformation environment, because the most powerful driver of election misinformation remains false claims from prominent people, not the software. The constant in every election panic is the human being willing to lie. That is precisely why the law has always targeted the lie, the fraud, and the impersonation rather than the printing press, the radio, the television, or the algorithm of the moment.
The threat is real. The answer is older than the technology: punish deception, whatever tool, human or machine, commits it.
Bartlett D. Cleland is senior adviser, Policy and Fellow, Innovation and Technology and a senior fellow in technology and innovation at the Pacific Research Institute. As a technology and policy expert, he has advised lawmakers, courts, regulators and served on federal and industry advisory bodies focused on technology, innovation, and communications.
