States' efforts to regulate election deepfakes keep colliding with the First Amendment.
asserted
efforts → regulate → Amendment
On Wednesday, U.S. District Judge Susan Watters issued a preliminary injunction preventing Montana from enforcing a 2025 law that restricts AI-generated political advertisements.
asserted
that → issue → advertisements
The law had been challenged by Accountability in State Government, a group run the former state lawmaker Dan Bartel; the organization had sent mailouts containing doctored images of political figures holding pride flags and wearing pronoun badges.
asserted
organization → challenge → badges
Under Montana law, it is illegal to sponsor or distribute an election communication within 60 days of voting if you "knew or should have known [it] is a deepfake"—though "satire or parody" is exempt.
asserted
satire → sponsor → voting
And so the targets of Bartel's mailouts filed complaints.
asserted
targets → file → complaints
State Rep. Jennifer Carlson (R–Manhattan), for example, said a mailer showing her holding an AI-generated pride flag was created "with the intent to injure the reputation of the candidate or otherwise deceive a voter."
asserted
mailer → say → voter
The complaints were dismissed on August 20, with Christopher Gallus, commissioner of political practices, concluding that the mailers were sufficiently "humorous or outlandish" to qualify as satire.
asserted
mailers → dismiss → satire
The next morning, Watters heard arguments on Bartel's request to block enforcement of the law altogether on the grounds that it violates the First Amendment.
asserted
it → hear → Amendment
This week, Watters found that Bartel and his group are likely to succeed on their First Amendment challenge.
asserted
Bartel → find → challenge
Montana's statute covers synthetic media created "with the intent to injure the reputation" of a candidate or political party, meaning that, as Watters put it, the law "treats identically deceptive media differently based solely on whether its message is weaponized to damage or deployed to elevate a candidate."
asserted
message → cover → candidate
Adam Schulman—an attorney at the Hamilton Lincoln Law Institute, which is representing Bartel in the case—says that this was the statute's fundamental flaw.
asserted
this → represent → case
"You could have AI deepfakes that were fine as long as they were complimentary," Schulman tells Reason.
uncertain
Schulman → have → Reason
"You could have Donald Trump dressed in a Pittsburgh Steelers uniform deepfake [and that would be] fine because it's not injuring the reputation of a candidate."
uncertain
it → have → candidate
Schulman also argues that the law asks regulators to make inherently subjective judgments about political imagery.
asserted
law → argue → imagery
Montana's statute relies in part on how a "reasonable person" would understand the material, but that becomes difficult when satire and parody are involved.
asserted
satire → rely → material
"People have different artistic sensibilities," says Schulman.
asserted
Schulman → have → sensibilities
Plenty of states have attempted to restrict AI-generated political speech and discovered the constitutional difficulty of doing so.
asserted
Plenty → attempt → difficulty
After the conservative YouTuber Christopher Kohls posted an AI-produced parody of then–Vice President Kamala Harris in the 2024 election, California passed Assembly Bill 2839, prohibiting the distribution of communications "that contains certain materially deceptive content" produced "with malice."
asserted
that → post → malice
In 2025, a federal court ruled the law was unconstitutional and permanently enjoined California from enforcing the law.
asserted
law → rule → law
Similarly, Hawaii passed Act 191 in 2024, which restricted "materially deceptive" election media.
asserted
which → pass → media
The Babylon Bee and Dawn O'Brien, a Hawaiian resident, challenged the law as a violation of free speech.
asserted
Bee → challenge → speech
In January 2026, U.S. District Judge Shanlyn Park granted a summary judgment finding the law "unconstitutionally vague," in part because it "employs ambiguous and subjective standards."
asserted
it → grant → standards
As new technologies develop, calls for more government regulation inevitably ensue.
asserted
calls → develop → regulation
The push to regulate AI-generated political speech reminds Schulman of COVID-era efforts to police "misinformation."
asserted
push → regulate → misinformation
Montana has a month to appeal the decision.
asserted
Montana → have → decision
But given how things have gone for other states, it might not want to bother.
uncertain
it → give → states
The constitutional limits on regulating AI-generated political speech may not be fully settled, but the early cases have been a warning against attempts to restrict protected political expression.
uncertain
cases → regulate → expression