The legal landscape around AI-generated speech is still evolving, but a recent decision in Montana offers a positive development for proponents of free speech.
Earlier this year, the Montana-based interest group Accountability in State Government sent AI-doctored political mailers to voters. One depicted a legislator spraying cash from a gas pump to display his support for a gas tax hike, and two others depicted Republican legislators sporting pride flags and pronoun pins to suggest their alignment with Democrats in the state legislature.
The legislators depicted filed suit, claiming the mailers violated a Montana state law against AI-generated “deepfakes” in political campaigns.
But on Sept. 16, a U.S District Court sided with the interest group, arguing the law likely violates the First Amendment.
The Montana law does not ban AI images outright. Instead, anyone paying for political messages they “knew or should have known” contain a “deepfake” must stamp it with a state-written disclaimer saying it “depicts speech or conduct that falsely appears to be authentic or truthful.” Violators of the law can be sued, and repeat offenders may face jail time.
Yet District Judge Susan Watters found the law likely discriminates against political speech based on both its content and its viewpoint. It treats AI images meant to damage a candidate more harshly than equally fake ones meant to flatter.
Montana is far from the only state attempting to ban deepfakes in political campaigns. Thirty states now have laws regulating deepfakes, according to Public Citizen.
“This is the third court within the footprint of the Ninth Circuit that has said these types of laws exceed the constitutional authority of states,” said Adam Schulman, senior attorney for the Hamilton Lincoln Law Institute, which represented the interest group.
State laws against deepfakes in California and Hawaii were struck down by federal judges earlier this year. A federal court, however, allowed a similar deepfake law in Minnesota to stand. The Montana decision once again brings momentum back to the free speech argument, as further challenges inevitably head to circuit courts on appeal.
When asked if the case establishes a precedent nationally, Schulman said, “I think it certainly does. There are many officials who feel they have the authority and duty to regulate misinformation more broadly.”
Yet for future campaigns, in Montana at least, AI-generated imagery and messaging will be protected under the First Amendment as much as any other doctored image has been protected in the past.
The real question here is not whether AI-generated speech is dangerous, but who gets to decide what speech is and isn’t appropriate. Schulman said many legislators view artificial intelligence as “low-hanging fruit” to go after misinformation, and that those same legislators want to “anoint themselves the arbiters of truth.”
This doesn’t mean the state’s case is completely frivolous. “Montanans deserve to know when they’re being deceived by AI-manipulated images, audio and videos of candidates,” the state’s Attorney General’s office said.
Sure, but the burden falls more on individuals than government regulators. The law on tools such as Photoshop is clear: their output cannot be separated from the user’s actual speech and expression. “No state has ever made that argument,” Schulman said.
Similarly, evolving case law suggests that absent using deepfakes for fraud or abuse, AI is considered the paintbrush, not the painter. Whatever it generates is still the protected speech of the user. “It’s just a technology for expressing an opinion,” Schulman said.
The ruling is a reassuring sign for AI companies as they continue to develop tools and innovate: their products are increasingly receiving constitutional protection. Yet the ruling is still fairly narrow, applicable only to paid political messages. New test cases across a variety of industries and applications will be necessary to establish expansive case law, and district court rulings are just the beginning.
FORGET DRONES: UKRAINE’S MOST LETHAL EXPORT IS DATA
Concerns over the abuse of new, untested technologies are real. At the same time, many of those concerns echo the same arguments that were proven overblown throughout the 20th century as other novel technologies developed.
One concern is timeless, however: regardless of the technology used, America cannot have a free society without a robust culture of free speech.
Derek Tresner is a Young Voices Contributor, small business owner, and student of political science at Liberty University’s Helms School of Government.