Real ads, fake images: AI enters Michigan governor's race

In election years, Michiganders are no strangers to being flooded with campaign ads. Billboards, TV and radio spots, social media posts and more promoting candidates and movements fill whatever medium campaign operatives think will reach the most prospective voters.

And in a growing number of races across the country, campaigns are turning to generative artificial intelligence to bolster, or even just entirely create, their ads. Using AI can be quicker and cheaper than producing a traditional ad and does allow for some creative liberties — but in Michigan, state law mandates campaigns disclose when they use AI in political materials.

Take a few recent social media posts from state Sen. Aric Nesbitt, a Porter Township Republican among those running in the Tuesday, Aug. 5, statewide primary election with hopes to become Michigan's governor: On June 9, he posted an AI-generated campaign video on social media featuring the likeness of several politicians, including himself, Democratic gubernatorial candidate and Secretary of State Jocelyn Benson, Gov. Gretchen Whitmer, Republican U.S. Senate candidate Mike Rogers and others. Another video posted on Nesbitt’s Facebook page includes likenesses of Whitmer, Benson and Fay Beydoun, a former nonprofit leader who was charged with a series of crimes in May related to a $20 million state grant.

Also included in both videos is a likeness of Attorney General Dana Nessel. At the tail end of each spot, the Nessel-likeness informs viewers that the video was created entirely by artificial intelligence.

That’s because Michigan is among the states that has taken steps to put guardrails around the use of AI in political campaign materials. In 2023, Whitmer signed legislation that requires politicians and their campaigns to disclose when AI is used in political advertisements.

Specifically, political ads are required to clearly disclose if they were generated using artificial intelligence. And within 90 days of an election, it’s a crime to distribute “materially deceptive media” of an individual that attempts to sway voters without including a clearly visible or intelligible disclaimer that the ad depicts events that did not occur.

Nesbitt defended the videos. He said in a recent interview with the Detroit Free Press Editorial Board the ads are meant to be a bit of fun and a way to reach voters across more forms of media.

“Unlike some others, we made sure there’s all the disclaimers on there, that people know the words, the actions weren’t real and that it was there for entertainment value,” Nesbitt said.

Nesbitt isn't the only gubernatorial candidate who has used AI in a recent ad, either. Bloomfield Hills businessman and Republican candidate Perry Johnson also disclosed the use of AI in one of his political ads in late May.

What happens if you don't disclose?

Violators of Michigan’s relatively new AI-disclosure ads face criminal penalties — a first violation is a misdemeanor punishable by up to 90 days in jail, a $500 fine or both. A second violation within five years is a felony punishable by up to five years' imprisonment, a fine of $1,000 or both.

A 2023 report published by the Brennan Center for Justice at New York University noted that generative AI is growing in prominence for political operatives — AI platforms can be used to write fundraising emails, sort through potential donor candidates and, as mentioned above, generate campaign content.

At the time Michigan’s AI-disclosure measures were signed into law in 2023, the state became just the fifth to have legal requirements for political ads to say wheher they were generated using AI. Today, 30 states have some sort of law regulating the use of AI content and deepfakes in political messaging, according to the National Conference of State Legislatures.

While the use of AI-generated content by political campaigns is becoming more commonplace, AI literacy hasn't kept up at a proportionate pace, said Cliff Lampe, a professor at the University of Michigan's School of Information.

The lack of comprehensive understanding around AI-generated content — voters being able to discern from what's real and what's not — makes legal guardrails like Michigan's less effective, he added.

"Literacy is the most important thing," Lampe said. "There's a lot of catch-up to do."

At the federal level, there currently isn’t legislation in place regulating the use or disclosure of AI in political materials. A 2024 paper published by the nonpartisan Congressional Research Service found while federal advertising statutes that require political materials to include disclaimers about who paid for them do apply to AI-generated ads, they don’t necessarily require those disclaimers to include any disclosure that the ad was made by AI.

Free Press staff writer Clara Hendrickson contributed to this report.

You can reach Arpan Lobo at alobo@freepress.com