In a newly released campaign ad, California Governor Gavin Newsom stands beside a strange and ominous machine—a robotic Democratic Party apparatus—and introduces it to his would-be successor, former Health Secretary Xavier Becerra. The machine, Newsom explains, was “created to make sure that nothing ever changes.” All Becerra has to do to enjoy its full protection, he says, is submit to it. Becerra flashes a grin and replies, “I can do that.” It’s a biting, satirical moment, and it comes courtesy of Steve Hilton, the Republican candidate running against Becerra in November’s gubernatorial election. But the ad has a twist that sets it apart from traditional political attacks: it’s almost entirely generated by artificial intelligence. A two-minute-plus disclaimer at the end reveals that the video was “generated or substantially altered using artificial intelligence.” For Hilton’s campaign, the ad is a creative experiment, a way to mock the Democratic establishment with a futuristic flourish. But for many lawmakers and watchdogs, it’s also a warning about a fast-evolving political landscape in which AI-generated content is flooding social media, raising uncomfortable questions about what voters can trust and whether existing laws are capable of keeping pace with a technology that can imitate reality with unnerving precision.
Hilton’s ad is far from an isolated case. Across the country, political campaigns and their supporters are experimenting with AI in ways that range from playful parody to outright deception. A viral AI-generated video depicted Kentucky Republican Representative Thomas Massie holding hands and dining with Democratic Representatives Alexandria Ocasio-Cortez and Ilhan Omar, an image designed to provoke outrage among his constituents. Another ad featured a fake likeness of New York Governor Kathy Hochul telling viewers to “Keep it unsafe and unaffordable with Kathy Hochul,” a clearly fabricated statement intended to damage her politically. These examples have caught the attention of federal lawmakers, including U.S. Senator Adam Schiff and Silicon Valley Representative Ro Khanna, who recently reintroduced the AI Ads Act in Congress. Schiff’s bill would update existing federal election law, which already prohibits fraudulent political ads, to explicitly include content generated by artificial intelligence. In a video promoting the bill, Schiff compiled clips from Hilton’s ad alongside the other examples, arguing that the rise of AI-generated campaign material represents a “runaway challenge” that will only get worse as the technology becomes more sophisticated. Under the proposed law, the Federal Election Commission would handle complaints about fraudulent AI political ads in federal races. But even as the bill gains attention, it remains to be seen whether it can overcome the significant legal and philosophical hurdles that any attempt to regulate political speech inevitably encounters.
California has already tried to get ahead of the AI wave. In recent years, the state has passed a slate of laws aimed at regulating deepfakes—AI-generated images or videos that depict people saying or doing things they never said or did. State law also requires campaign ads to disclose when they’ve been created or substantially altered using AI. But enforcement has been uneven, and the boundaries of these laws are still being tested. Earlier this year, an AI-generated video surfaced in the Los Angeles mayoral race, portraying candidate Spencer Pratt as Batman and his rival, Mayor Karen Bass, as the Joker. The video raised alarms for some lawmakers, who worried it violated state disclosure requirements. But a more significant hurdle came when a federal judge struck down a California law that would have required social media platforms to restrict AI-generated political content ahead of elections. The judge ruled that the law ran afoul of the Communications Decency Act, a federal statute that shields online platforms from liability for what their users post. Adding to the frustration, President Donald Trump issued an executive order last year that established a national framework for AI policy, explicitly aiming to limit state-by-state regulation. The order also created an AI Litigation Task Force within the U.S. Department of Justice to challenge state laws that conflict with the president’s vision. For state lawmakers like Assemblymember Marc Berman, a Palo Alto Democrat who authored the ill-fated California law, the situation is deeply frustrating. “It’s been very frustrating,” Berman said. “I think the reality is what we need is strong federal law to regulate in this space.”
The debate over AI-generated political ads ultimately boils down to a fundamental conflict between innovation and protection, and at the heart of that conflict lies the First Amendment. Proponents of using AI in campaigns argue that it is a form of political speech protected by the Constitution, no different from a cartoonist’s pen or a satirist’s script. Hector Barajas, a spokesperson for Hilton’s campaign, defended the decision to use AI, calling it “one of the most powerful creative tools available today.” When asked about restrictions on deepfakes, Barajas emphasized transparency, drawing a sharp line between innovation and deception. The greater concern, he argued, isn’t campaigns that openly disclose what they’ve created, but anonymous actors using AI to spread misinformation without accountability. That sentiment echoes the views of free speech advocates like David Inserra, a fellow at the Cato Institute, who cautions that AI is ultimately “a tool of expression.” Inserra points to Supreme Court rulings that fiercely protect political speech, even when it’s uncomfortable—like the 1989 decision that deemed flag burning a protected form of protest. Under that reasoning, restrictions on AI-generated political ads could be seen as an unconstitutional curb on citizens’ ability to express themselves about their government and their elections. Inserra compares AI-generated ads to political cartoons that parody and satirize public figures, suggesting that the proper remedy for misleading content lies not in preemptive regulation, but in libel laws that allow individuals to seek damages after harm is done.
Critics see the issue very differently. To them, the free speech argument misses the point entirely. Assemblymember Berman put it succinctly: “The First Amendment allows you to say what you want to say, it does not allow you to put your words into my mouth. And that’s literally what this technology allows—for somebody else to create a very, very realistic looking video of me saying things or doing things that I never said or did.” That distinction—between expressing your own views and hijacking someone else’s voice—is at the core of the concern. AI-generated deepfakes don’t merely add another opinion to the marketplace of ideas; they poison the well by making it impossible to know what is real and what is manufactured. Ramesh Srinivasan, a professor of information studies at UCLA, warns that the steady spread of AI-generated content will erode trust in everything we see and hear. “It’s going to make us distrust everyone and everything,” he said. “Both are sledgehammers to democracy because you have to have not just a common basis of understanding of reality or what the truth is, but you also have to have some basic notion of trust in what you see.” Srinivasan, who served as a surrogate for Bernie Sanders’ 2020 presidential campaign, worries about a future in which AI dynamically creates personalized content tailored to each individual, trapping people in “non-overlapping worlds” where they can’t even agree on the basic facts of a shared reality. He advocates for at least requiring watermarks on AI-generated content, and believes that the industry must be regulated not to stifle innovation, but to ensure that innovation serves the public good rather than undermining it.
Looking ahead, it’s clear that AI-generated content is not a passing trend—it’s becoming a major media strategy for campaigns and political action committees, and its use is likely to expand dramatically in the coming cycles. The Hilton ad itself was produced by filmmaker Charles Curran, the same creator behind the Pratt Batman ad that grabbed national attention before Pratt was eliminated in the June primary. Curran didn’t respond to requests for comment, but his work illustrates how quickly AI-generated political content is moving from the fringes into the mainstream. Even Berman acknowledged that the depictions in Hilton’s ad looked “obviously fake,” but he expressed concern about what the future holds as AI models become more advanced and produce hyper-realistic content that is increasingly hard to distinguish from reality. The legal and regulatory landscape remains uncertain. Schiff’s AI Ads Act would remove the requirement that fraudulent misrepresentation be “damaging” to a candidate or party, which could make it easier to police deceptive AI content. But it would still face the challenge of balancing the Constitution’s robust protections for political speech with the need to protect voters from deception. In the absence of strong federal action, states like California will continue to struggle against federal preemption and court challenges. The result is a patchwork of laws, executive orders, and unresolved debates that leaves candidates, platforms, and voters in a state of flux. What’s ultimately at stake is more than just the outcome of any single election—it’s whether democracy can function when reality itself becomes negotiable, and whether innovation can be harnessed without sacrificing the trust that binds a society together.