The company’s federal lawsuit filed this week also contends that the 2023 state law is preempted by a 1996 federal statute that shields social media from being held responsible for material posted on their platforms.

“While the law’s reference to banning ‘deep fakes’ might sound benign, in reality it would criminalize innocuous, election-related speech, including humor, and make social-media platforms criminally liable for censoring such speech,” the company said in a statement. “Instead of defending democracy, this law would erode it.”

Minnesota’s law imposes criminal penalties — including jail time — for disseminating a deepfake video, image or audio if a person knows it’s fake, or acts with reckless disregard to its authenticity, either within 90 days before a party nominating convention, or after the start of early voting in a primary or general election.

It says the intent must be to injure a candidate or influence an election result. And it defines deepfakes as material so realistic that a reasonable person would believe it’s real, and generated by artificial intelligence or other technical means.

“Elon Musk funneled hundreds of millions of dollars into the 2024 presidential election and tried to buy a Wisconsin Supreme Court seat,” said the law’s author, Democratic state Sen. Erin Maye Quade.

“Of course he is upset that Minnesota law prevents him from spreading deepfakes that meant to harm candidates and influence elections. Minnesota’s law is clear and precise, while this lawsuit is petty, misguided and a waste of the Attorney General Office’s time and resources,” her statement said.

Democratic Minnesota Attorney General Keith Ellison’s office, which is legally obligated to defend the constitutionality of state laws in court, said in a statement that it’s “reviewing the lawsuit and will respond in the appropriate time and manner.”

The Minnesota law was already the subject of a constitutional challenge by Christopher Kohls, a content creator, and GOP state Rep. Mary Franson, who likes to post AI-generated parodies of politicians. That case is on hold while they appeal to overturn a judge’s denial of their request to suspend the law.

The attorney general’s office argues in that case that deepfakes are a real and growing threat to free elections and democratic institutions, that the law is a legitimate and constitutional response to the problem, and that it contains important limitations on its scope that protect satire and parody.

X, formerly known as Twitter, said it’s the only social media platform challenging the Minnesota law, and that it has also challenged other laws it considers infringements of free speech, such as a 2024 California political deepfakes law that a judge has blocked.

X said in its statement that its “Community Notes” feature allows users to flag content they consider problematic, and that it’s been adopted by Facebook, YouTube and TikTok. The company’s lawsuit said its “Authenticity Policy” and “Grok AI” tool provide additional safeguards.

Alan Rozenshtein, a University of Minnesota law professor and expert on technology law, said in an interview Friday that it’s important to separate the free-speech issues from whatever one thinks about the controversial Musk.

“I’m almost positive that this will be struck down,” Rozenshtein said.

There’s no exception under the First Amendment for false or misleading political speech, even lies, he said. And the potential for criminal penalties gives social media companies like X and Facebook “an incentive to take down anything that might be a deepfake. … You’re going to censor a massive amount to comply with this law.”

Deepfakes aren’t good, but it would be nice to get evidence that they’re causing actual problems before imposing such limits on free speech, the professor said. And while it’s easy to focus on the supply of misinformation, the large demand for it is the problem.

“People want to be fooled, and it’s very bad for our democracy, but it’s not something I think can be solved with a deepfakes ban,” he said.