- Michigan is one of 31 states that allows for using generative AI in political advertising, with some exceptions
- As new ads and technology test the law, some feel more guardrails may be necessary
- So far, AI-generated ads have sparked just one campaign finance complaint
LANSING — “Michigan law requires me to tell you that this video was generated completely by artificial intelligence,” an AI-generated version of Attorney General Dana Nessel said at the end of a recent political advertisement.
“It has been manipulated by technical means and depicts speech and conduct that did not occur.”
The June ad from Republican gubernatorial candidate Aric Nesbitt, who has since dropped out of the race, shocked some viewers by using AI to show Nesbitt driving a tractor towards a fallen Democratic Gov. Gretchen Whitmer. “Justice is coming,” it said.
Democrats just wanna have fun- at your expense.
2026 is the year we pull the plug on their nonsense and TAKE MICHIGAN BACK! pic.twitter.com/DKGrfnJo8Z
— Aric Nesbitt (@aricnesbitt) June 9, 2026
It’s among a series of AI-generated ads testing the boundaries of a new state law that aimed to put guardrails on how the emerging technology is used to influence political campaigns.
Michigan is one of at least 31 states to regulate political deepfakes or other forms of AI-assisted advertising. But two years into the new state law, the quickly improving technology has backers wondering if more can be done to help voters separate fact from fiction.
“AI is doing a lot for improving research, for improving efficiencies … but it’s just moving so very rapidly that we have to protect from any bad actors, misuse or just using it recklessly,” sponsoring state Rep. Penelope Tsernoglou, D-East Lansing, told Bridge Michigan.
Related:
In Michigan, AI has so far been used most prevalently in the state’s Republican gubernatorial primary, where candidate Perry Johnson recently published a six-minute online ad that showed himself chasing down rival John James in an effort to force James to debate him. (They since debated twice, last week.)
The ad ended with an AI-generated Johnson, 78, tackling James on a golf course. Johnson then removed James’ face — secretly, a mask — to reveal Secretary of State Jocelyn Benson, the Democratic frontrunner who James, Johnson or fellow Republican Mike Cox could face in November.
“Well, that explains his votes,” Johnson said, at times looking more like Al Pacino.
What’s on the books
Tsernoglou and state Rep. Matthew Bierlein, R-Vassar, were two of the driving forces behind Michigan’s AI deepfake law, which Whitmer signed in late 2023.
The law allows AI use in political ads so long as backers disclose its use in the making of a video, graphic or audio message. Failure to do so could be a civil infraction, punishable by a fine of up to $250 for a first offense and up to $1,000 for subsequent offenses.
The law also bars distributing deceptive AI-generated material within 90 days of an election, as well as knowingly disseminating false information.
Ilana Beller, director of AI governance and tech policy at the nonprofit democracy group Public Citizen, said Michigan’s law surrounding generative AI use in political ads is “really strong” compared to other states.
But since the law took effect in 2024, just one campaign finance complaint has been filed alleging AI use was not properly disclosed in an ad, Department of State officials told Bridge.
While state legislators intended the law to apply to federal campaigns, federal candidates may not be subject to its provisions.
Because AI disclosure is required under state law in the Michigan Campaign Finance Act, “candidates for federal office are excluded from the definition of candidate” under the law, said Department of State Press Secretary Sam May.
Michigan has not seen a huge number of “fully faked” political ads this cycle, but AI is more regularly “creeping into the generation of text, graphics — those kinds of things,” said John Sellek, founder and CEO of the Lansing-based Harbor Strategic Public Affairs.
“It’s just the natural progression of technology and the need for speed in a campaign.”
First Amendment protections
Even some candidates campaigning against data centers are using AI, which is fueling the boom in data center construction.
Johnson, the Republican gubernatorial candidate, has called for a one-year moratorium on data centers. In an AI-generated video published online last month, he accused Benson of benefiting from data centers because of ties her husband has to the industry.
Any attempts to bar AI use altogether in political ads could be tricky and possibly run afoul of free speech protections, said Beller, the AI governance director for Public Citizen.
“There are incredibly strong First Amendment protections around political speech, and … those protections make it such that there are only certain measured ways to address this type of content,” Beller said.
“Satire and parody are protected no matter what, so if the political deepfake constitutes satire or parody, it will not be regulated by law.”
That’s the case with Nesbitt, a Porter Township Republican and former 2026 gubernatorial candidate, who was an early adopter of fully AI-generated ads.
In June alone, his campaign released three separate but fully AI-generated ads targeting Benson, all of which carried the necessary disclaimers.
During one, the Statue of Liberty came to life and seemingly crushed New York City Mayor Zohran Mamdani and Democratic US Senate candidate Abdul El-Sayed as they worked to help Benson win Michigan’s gubernatorial election.
In another, an AI-generated Nesbitt crashed a tractor through what seems to be a Benson fundraiser before driving it through a seven-layered cake iced with the words “Fraud, Fraud & Abuse.”
Nesbitt dropped out of the race in later June after President Donald Trump endorsed James.
But his AI ad that appeared to depict him preparing to run over Whitmer with a tractor has the law’s original sponsor considering whether to revisit the legislation as AI capabilities grow.
“We hadn’t even considered the fact that the AI ads would be inciting political violence,” Tsernoglu told Bridge. “That’s something I would want to take a second look at.”
In Michigan’s 7th Congressional District, a political ad attacking Democratic candidate Bridget Brink sparked a complaint over alleged nondisclosure of AI use.
Michigan Values PAC has denied using AI in the ad, which appears to superimpose footage of Brink onto a folder sitting on President Donald Trump’s desk.
While the Michigan Department of State is not applying the AI disclosure law to federal candidates, it is reviewing whether the law “might apply to any types of spending by any entities on federal races,” said May, the spokesperson.
The bureau, she added, “investigates every campaign finance complaint in accordance with its interpretation of the law.”
Related
Republish This Story
