Election advertisements already have to include a notice about who paid for and approved them. A House panel on Tuesday advanced a bill that would require campaigns in New Mexico to disclose whenever they use artificial intelligence in their ads, and would make it a crime to use artificially-generated ads to intentionally deceive voters.

The House Government, Elections & Indian Affairs Committee on Wednesday voted 7-2 to pass House Bill 182, which if signed into law, would amend the Campaign Reporting Act to require political campaigns to disclose whenever they include synthetic images, visuals and audio created by artificial intelligence in their advertising.

Co-sponsored by House Majority Leader Rep. Gail Chasey (D-Albuquerque) and four other lawmakers, the bill would also criminalize using deepfakes created to damage individuals’ reputations and make the public question what’s real and fake.

“This bill requires that if such a technique is used, whether it’s audio, visual or written, that it be disclosed to the public, so the public knows what they’re seeing,” Chasey said. “It really is about transparency with artificial intelligence.”

The bill would not entirely prohibit using artificial intelligence in campaign ads, however. Candidates must disclose any AI used to create ads. If they don’t, the proposal would make it a misdemeanor to knowingly publish a deceptive ad within three months of an election with the intent of changing voters’ behavior.

The bill defines “materially deceptive media” as an image, video or audio depicting “an individual engaged in conduct or speech in which the depicted individual did not engage.”

If the same person is convicted a second time in five years, they could be guilty of a felony.

Melanie Moses, a professor with the University of New Mexico Department of Computer Science, said these technologies can now generate images, videos and audio “almost entirely indistinguishable from reality,” are freely available, and can have a profound impact on people’s ability to understand the difference between truth and fiction.

“There’s now a point where any individual — myself included — would really struggle to simply look at an image or listen to audio and know whether it was real or just made up,” Moses told the committee. 

“The technology itself is very complex but this particular bill is dealing with something that’s actually quite simple: Do we want the voters to understand the difference between what a candidate, any of you, what any political person has actually said versus what is entirely fabricated using technology that is now available to just about everyone?” she said.

Lindsey Bachman, legislative and executive affairs director for the New Mexico Secretary of State’s Office, said under the bill, her agency would investigate a specific advertisement when they receive a complaint, and then make either a civil or criminal referral.

“We’ve seen the use of this increase in the past two years in elections around the world,” Bachman said. She said the bill would “create needed transparency to build essential trust among New Mexico’s voters in the electoral process and make voters aware of the use of AI in the democratic process generally.”

The New Mexico Secretary of State is preparing a public education campaign for elections this year about mis- and disinformation, and provide tools voters can use to understand whether something is real or representative of a candidate’s actual circumstances, Bachman said.

Reps. William Rehm (R-Albuquerque) and John Block (R-Alamogordo) questioned Chasey about whether the proposal would apply in various circumstances. Rehm said a solution is needed but HB 182 is so broad it would require disclosure for something as small as using AI to tweak a logo’s background.

“I think almost any mailer that we generate would have to have some disclaimer that AI was used,” Rehm said.

Something like a simple logo couldn’t possibly require a disclaimer, according to Moses, the computer science professor.

Chasey’s chief of staff Alisa Lauer, who sits on an interstate working group on trying to define AI in the law, said AI disclaimers on parody and satire are excluded in the bill. The text in the bill on that exclusion states, “an advertisement that reasonably constitutes satire or parody.” 

Government watchdog Common Cause New Mexico and consumer advocacy group Public Citizen support the bill.

Craig Holman, a lobbyist for Public Citizen, said deepfakes have been used to deceive voters at the federal, state and local levels around the U.S., while only five states require transparency around them. New Mexico is among 28 states considering joining them, he said.

“We can expect the 2024 election to be particularly problematic when it comes to these deepfakes, because there are very few regulations and very few disclosure requirements involved,” Holman said. “The 2024 election is going to be the first really serious deepfake election, and the federal government is very slow at acting, so it’s really up to the states to step up to the plate and address this.”

HB 182 heads next to the Senate Judiciary Committee.

Sens. Michael Padilla and Debra Sariñana have also introduced a bill which would create a work group to develop legislative proposals and policy recommendations for artificial intelligence systems. That bill has sat in the Senate Committees Committee since Jan. 22.