Getting your Trinity Audio player ready…
An incident in early June depicting an AI-generated image of Lt. Gov. Peggy Flanagan is just one example of how AI is popping up in campaigns. But is that just the tip of the iceberg?
The pro-U.S. Rep. Angie Craig TV ad made by North Star Dawn PAC shows Flanagan standing atop a pile of cash holding her hand out for more money. The two are competing in the Aug. 11 Democratic primary for U.S. Senate
Whether the ad actually violated Minnesota’s existing law that bars the dissemination of political deepfakes within 90 days of a nominating convention is still being disputed. While the ad wasn’t made by a campaign itself, it’s not clear to what extent Minnesota political campaigns are using AI — if at all
Dr. Manjeet Rege, an AI expert and professor of software engineering and data science at the University of St. Thomas, said AI is being used in elections more than people realize
“AI is already woven through elections in ways most people never see or are not aware (of),” he said. “So the largest use is behind the scenes in campaign operations, where it would draft ad copies and speeches, summarize opposition research, sort donor and voter lists.”
Rege said the “intentional harms are the obvious ones,” meaning deepfakes created specifically to deceive, such as fabricated audio of false endorsements or AI-altered images
“The unintentional side worries me just as much, and it gets far less attention,” he said. “A lot of harm can happen with no bad intent at all, like when chat or a AI summary confidently tells a voter something about a wrong polling place or the wrong deadline, because that kind of basic logistics needs more accuracy than these tools reliably can deliver today.”
An issue lawmakers may revisit
Minnesota’s only major law specific to AI use in elections is the 2023 deepfake law. Free speech protections can sometimes complicate efforts to regulate AI-generated content, as well as lawsuits from Big Tech
Sen. Zach Duckworth, R-Lakeville, an author of Minnesota’s AI nudification technology ban, said he would bet that taking a closer look at AI use in campaigns would be a bipartisan issue if lawmakers chose to revisit it
“We have enough trouble as it is making sure that people are getting accurate information in terms of being able to formulate their opinions and the decisions they want to make when it comes time to vote,” he said. “The last thing we need are campaigns or … outside third parties … using AI to purposely mislead voters and produce outcomes that are based on anything other than fact. That should always be concerning and alarming to everybody.”
Duckworth said that, to be fair, he could see some “innocent” ways campaigns might use AI. For example, a small business owner running a campaign might use AI to make fliers
Sen. Erin Maye Quade, DFL-Apple Valley, who has worked on many of Minnesota’s AI laws, says campaign uses like speech writing aren’t necessarily concerning from a legal perspective, but they are concerning from a “human perspective.”
“Minnesotans don’t want this in their elections,” she said
Campaigns asked about AI use
Research shows humans are becoming better at spotting AI, Maye Quade said, and that generally, it’s “unsettling” to spot it
“You don’t want someone to look at something of yours and be unsettled by it. If you want them to vote for you, that’s actually not going to urge people to vote for you. And so, from a human perspective, I think it’s just stupid,” she said. “Legally, if people want to be lazy and use AI to write speeches instead of saying their own words, like, by all means, they can do that.”
Several high-profile campaigns — Craig, Michele Tafoya, and gubernatorial candidates Kendall Qualls and House Speaker Lisa Demuth — did not provide details to Forum News Service about whether they use AI in their campaign materials, how they use it if they do, or whether they have guidelines on AI use
A spokesperson for U.S. Sen. Amy Klobuchar’s campaign for governor said Klobuchar “has long warned about the deceptive use of AI-generated images, audio, and video in our elections, in scams, and in nonconsensual sexual images.”
“She worked across the aisle to pass her bipartisan TAKE IT DOWN Act, signed into law last year, to criminalize the publication of non-consensual intimate images, and leads bipartisan legislation to ban deceptive AI content in political ads,” the spokesperson said. “Her campaign follows the same standards her bipartisan bills would make law.”
Lexi Byler, a spokesperson for Flanagan’s campaign, said the use of “deceptive AI in political campaigns is a threat to public trust.”
“We’ve already seen that firsthand in Minnesota when a super PAC supporting Congresswoman Angie Craig aired an attack ad with an AI-generated deepfake of Lt. Gov. Peggy Flanagan,” Byler said. “Campaigns and super PACs should be able to agree on one basic principle: don’t use AI to manipulate voters or fabricate images, audio, or video of your opponents.”
How much AI tools could be used in campaigns — and whether the public will know about that use — remains to be seen. Regardless, said Rege, the AI expert with the University of St. Thomas, “the line between optional and unavoidable is disappearing.”
“That is because AI is no longer a separate tool that you just choose to switch on. It is being built directly into the apps campaigns already rely on,” he said
