For months, the Ethics Commission has discussed what to do about AI-generated deepfakes of Oklahoma politicians appearing in TV ads, mailers and other forms of campaign messaging.
Should the commission require disclosures of AI political content? Or completely ban its use during campaigning?
"If we are going to consider the disclosure approach, which I think we are, that's why we're talking about it," Executive Director Lee Anne Bruce Boone said. "The first step we've got to think about is really defining the problem that we're trying to solve."
The problem is Oklahomans are seeing ads and hearing claims that are untrue, she said, and there is no telling how that might be impacting their choices at the ballot.
But regulating the issue isn't as simple as slapping together a rule mandating disclosure.
"Are we trying to disclose any use of artificial intelligence? Or are we trying to really disclose somebody who is doing a realistic fabrication that depicts a candidate engaging in speech or conduct that never occurred?" Bruce Boone said. "I think you have to think about that distinction and what should trigger a disclosure and how it might be factual, how it might be neutral."
There are also constitutional lines that can't be crossed, she said, and despite precedent set in cases like Buckley v. Valeo (1976) and Citizens United v. Federal Election Commission (2010), which established existing campaign finance and electioneering disclosure laws, it's not clear where those lines are drawn for this specific issue.
"There's no definitive Supreme Court decision yet that really addresses compelled AI labels in political advertising," Bruce Boone said. "We've got some latitude between Buckley and Citizens United about traditional campaign finance disclosure, which we already do, but the unresolved question is where that AI label falls between those two bodies of law."
The goal, Bruce Boone said, is to protect voters' right to know and be informed, while also protecting the First Amendment right of free speech. What that looks like in an AI-world is an open legal question for the U.S. Supreme Court, she said. And while the Supreme Court hasn't yet heard a case addressing the issue, the ethics commission is monitoring lower-court litigation across the country.
Bruce Boone pointed out two cases that illustrate how states may be able to regulate AI-generated political messaging, or not.
Kohls v. Bonta et al in the U.S. District Court of Eastern California
Several satirical content creators led by Christopher Kohls (known as "Mr. Reagan on social media) and the social media platform X sued California Attorney General Rob Bonta and state election officials over the state law AB 2839 in 2024.
The measure, inspired by a deepfake Kohls posted of former Vice President Kamala Harris during her presidential run, banned political content depicting candidates or elected officials doing or saying something they never did 120 days out from an election. It also required political AI content in the form of images, video, or audio to have a disclaimer saying: "This [Blank] has been manipulated."
While there were no explicit punishments for violating the law, it allowed parties harmed by the content to sue for damages. The plaintiffs fighting the law argued it violated their right to free speech, pre-emptively chilling creators who criticize politicians while staying lenient on those who depict deepfakes of themselves, including politicians.
In August 2025, the court agreed with the plaintiffs, granting the permanent injunction they sought and effectively negating the new law.
"While there are serious concerns about deepfakes and AI affecting elections, California's AB 2839 represents a law that is well-intentioned but constitutionally infirm," the court order reads. "When it comes to political expression, the antidote is not prematurely stifling content creation and singling out specific speakers but encouraging counter speech, rigorous fact-checking, and the uninhibited flow of democratic discourse."
Kohls et al v. Ellison et al in the U.S. District Court of Minnesota
This case, also launched in 2024 by Kohls and Minnesota Republican State Rep. Mary Franson, revolves around Minnesota's 2023 HF 1370, banning deepfakes of politicians published:
- 90 days before an election
- Without their consent
- With the intent to injure them or influence an election outcome
In 2024, the Minnesota legislature amended the law to include a prohibition 90 days before a political party's nominating convention and after the start of the state's absentee voting period.
Kohls made essentially the same argument he made in California alongside Rep. Franson. They asked the Minnesota court to stop the law, saying it violated their free speech rights.
The court found that Kohls, despite being a parodist and political commentator, lacked standing to sue because his content falls outside the parameters of the specific statute. On the other hand, the court said Franson did have standing but took too long to seek relief. So, the law stayed in effect.
The plaintiffs appealed the case to the Eighth Circuit, which kicked it back to the district level, where it remains in active litigation.
What does this mean for the future in Oklahoma?
The big takeaway for Oklahoma regulators, according to Bruce Boone's presentation during the meeting, is that "a legitimate governmental objective does not eliminate the need for narrow drafting when political speech is involved."
Furthermore, the courts haven't defined what language counts as a constitutional violation, meaning Oklahoma will likely enter the regulatory fray without pre-existing guidance, like more than 30 states already have.
"At this stage, I'm not recommending a particular rule," Bruce Boone said to the commission. "I'm trying to give you guidance on how the Constitution addresses it and leave it open for discussion and any future questions."
The Ethics Commission has until the second day of the regular legislative session if they mean to propose language for new rules to lawmakers in the next year. The session starts on Feb. 2, 2027.
