There’s a video of Rebecca Cooke, who is running for Congress in Wisconsin's 3rd District, on X saying that she’s campaigning simply to “take advantage of as much corruption as possible.” She also says she wants to “give every illegal alien free healthcare.” But the speech never happened. Instead, it’s a fake video her opponent, GOP Rep. Derrick Van Orden (WI-03), made using artificial intelligence.
The video is just one of a series of fake ones Van Orden has been running to lob attacks at Cooke. “Derrick Van Orden can’t run on his record, and is instead spending his time and energy producing AI videos putting words in my mouth, pretending that I said things that I didn’t and that I support policies I do not,” Cooke said in a statement.
It’s not an isolated incident. For the first time, American voters are being inundated with AI-generated political ads. Researchers from the Wesleyan Media Project, which tracks federal spending in political advertising, found that in this cycle alone, campaigns and outside groups have spent approximately $80 million on AI ads.
And in an era where it’s getting increasingly difficult to tell what’s real and what’s fake online, AI-generated content is proving to be harmful to voters.
“The most salient danger would be that a voter would not know clearly that what was in a political ad or what was being said by a candidate was not something they actually said or did,” Mike Franz, the co-director of the Wesleyan Media Project told The Caucus.
Cooke’s lawyers sent Van Orden a cease and desist letter. “Apparently unable to run a campaign based on the issues, you have made AI deepfakes a centerpiece of your attacks on Ms. Cooke, posting myriad such videos,” the notice said.
“We are monitoring your X account and your future communications misusing AI to defame Ms. Cooke. And, if you persist in running a campaign against an AI deepfake strawman rather than reality, we will evaluate all legal options to address that misconduct,” it continued.
While anti-AI sentiment cuts across partisan lines, Democrats are more pessimistic about its effects on society. The Wesleyan Media Project also found that the vast majority of AI ads and content are being made either by Republicans or by GOP groups.
The problem is more pronounced on social media, where more and more content is being seen by voters. That could cause them to make a decision based off of a fake content, like the videos Van Orden is posting about Cooke.
“What seems to be the most likely scenario where that would show up might be on digital platforms, like Facebook, Instagram, or ads on YouTube,” Franz said. “They’re highly targeted to particular constituencies, you can target your ads to reach your supporters…so it wouldn’t be something that would easily be policed.”
Some states have already made attempts to clear up any confusion in political ads. In 2019, the Texas state legislature voted to ban all AI-generated ads 30 days before an election. But Lt. Gov Dan Patrick (R-TX.), still ran spots that contained fake images of his opponent, Democrat Vikki Goodwin, in a shirt that says “taxes” in the same styling of University of Texas’ signature font and colors.
In response, Goodwin filed a criminal complaint. One of the ads attacking Goodwin also included state Rep. James Talarico, who’s running against the state’s attorney general Ken Paxton. It was certainly not the first deepfake of Talarico. In June, a Republican super PAC released an ad depicting a AI-generated Talarico singing a pro-transgender song to the tune of “My Favorite Things,” from the movie The Sound of Music.
The same situation is occurring across the country. In Ohio, a conservative super PAC released an anti-Sherrod Brown ad, the Democratic Senate candidate, in which he overstays his welcome at a party.
And in New Mexico, where lawmakers recently passed a measure that requires disclosures on AI-generated ads, the state’s Democratic Party has also lodged a complaint in the statehouse against their Republican colleagues.
The Democratic Party of New Mexico said images posted by state Republicans that include local candidates and Deb Haaland, who’s the Democratic candidate for governor, is "materially deceptive.”
"Voters should not be misled by public officials or their official offices about what is or isn't real," Sara Attleson, the state’s Democratic Party chair said in the complaint.
It’s unclear if a hodgepodge of laws across the country can help voters discern the truth when watching political ads.
“Some states do require campaigns to disclose and some do require candidates or groups to state in the ad that it was generated with AI, but some states don’t,” Franz said. “And there’s no federal level requirement. And so I do think voters are having a hard time.”
And a disclaimer alone may not even be enough. “It’s definitely not a cure-all for this,” Franz said. “It’s just more a way to flag it and people can make up their own minds from there. So, the jury's out on its effectiveness.”