Minnesota Law Banning Deepfakes for Political Candidates Challeng

Research basis: Minnesota Statutes Section 609.771, Minnesota House, Minnesota Secretary of State, AP, Reuters, Eighth Circuit materials, Brennan Center, Public Citizen, R Street, University of Minnesota Law, Business Insider, MPR, and MediaPost.

Note: This article is for general informational purposes only and is not legal advice.

Minnesota’s political deepfake law has become one of the most closely watched battles in the fast-growing fight over artificial intelligence, elections, and free speech. The law was designed to stop realistic fake videos, images, and audio from being used to damage political candidates or influence election results. That sounds simple enoughuntil you remember that American politics already runs on satire, memes, attack ads, bad impressions, and the occasional campaign commercial that treats nuance like a suspicious package.

The result is a major legal collision: election integrity on one side, First Amendment concerns on the other, and generative AI standing in the middle wearing a fake mustache. Minnesota wants to prevent voters from being tricked by synthetic media that appears real. Critics argue the state has written a law broad enough to chill parody, humor, platform moderation, and ordinary political commentary. The challenge matters far beyond Minnesota because many states are trying to regulate political deepfakes before AI becomes the campaign intern nobody hired but everyone has to manage.

What Minnesota’s Political Deepfake Law Actually Says

Minnesota Statute 609.771 targets the use of “deep fake technology to influence an election.” In practical terms, the statute covers video, audio, images, photographs, or other technological representations that are realistic enough that a reasonable person could believe they show someone saying or doing something they never said or did. The content must also be substantially created through technical means rather than ordinary human impersonation. In other words, a bad comedy sketch with a wig is not the same thing as an AI-generated candidate confession that looks like it was filmed on a real phone.

The law makes it a crime to disseminate, or agree to disseminate, a deepfake when several conditions are met. The person must know or act with reckless disregard that the material is a deepfake. The material must be shared without the consent of the person depicted. It must be intended to injure a candidate or influence an election result. And it must occur during sensitive campaign windows, including within 90 days before a political party nominating convention or after the start of absentee voting before a primary or general election.

The penalties can be serious. In ordinary cases, a violation can lead to jail time or a fine. More severe penalties can apply for repeat violations or content intended to cause violence or bodily harm. The statute also includes election-related consequences for candidates convicted of violating it, including potential forfeiture of nomination or office. That is not a slap on the wrist; that is the legal equivalent of being escorted out of the campaign bus while the engine is still running.

Why Minnesota Passed the Law

Minnesota enacted the measure in 2023, at a moment when lawmakers across the country were waking up to a new campaign problem: AI tools had become cheap, fast, and good enough to create convincing synthetic media. A fake robocall, a forged candidate video, or a fabricated audio clip could spread widely before journalists, election officials, or campaigns had time to correct it.

The fear is not imaginary. Political deepfakes can be used to make a candidate appear to confess to a crime, insult voters, announce a fake withdrawal, or give false voting instructions. In close elections, even a few hours of confusion can matter. Election misinformation does not need to fool everyone; it only needs to muddy the water long enough for people to shrug and say, “Who knows what’s real anymore?” That shrug is where democracy starts misplacing its car keys.

Supporters of Minnesota’s law argue that the statute is aimed at fraud-like deception, not ordinary political speech. They say the law focuses on synthetic media that falsely depicts real people, lacks consent, and is used with intent to harm a candidate or alter an election outcome. From that perspective, the law is not trying to ban jokes, criticism, or harsh political messaging. It is trying to stop realistic digital forgeries from becoming campaign weapons.

The Free Speech Problem: Political Lies Are Complicated

The constitutional challenge begins with a stubborn American principle: political speech receives the highest level of First Amendment protection. The government generally cannot punish speech simply because it is false. Defamation, fraud, true threats, and certain other categories can be regulated, but the Supreme Court has long been wary of laws that let officials decide what counts as political truth.

That creates a hard question. Is a realistic AI-generated video of a candidate saying something false more like fraud, or more like political commentary? What if it is satire? What if it is labeled as parody? What if it is obviously absurd to some viewers but not to others? What if a platform hosts it without knowing the creator’s intent? These are not law school hypotheticals anymore. They are Tuesday afternoon on the internet.

Critics say Minnesota’s law risks sweeping in protected speech because it turns on concepts such as intent to influence an election and whether a reasonable person would believe the content is real. Political speech is almost always intended to influence elections. Satire often works by imitating reality. Attack ads are designed to harm a candidate’s reputation. The line between deception and persuasion can get blurry, especially when politics already treats context like optional seasoning.

The Kohls and Franson Challenge

One major challenge came from political commentator Christopher Kohls, known online as “Mr Reagan,” and Minnesota state Representative Mary Franson. Kohls created an AI-generated parody video involving Vice President Kamala Harris. The video was labeled as parody and included a disclaimer saying that sounds or visuals were significantly edited or digitally generated. Elon Musk later shared the video on X without the same parody context, and Franson reposted Musk’s post.

Kohls and Franson sued Minnesota officials, arguing that the deepfake statute violated the First and Fourteenth Amendments. They sought a preliminary injunction to block enforcement of the law. A federal district court declined to grant that immediate relief. The court concluded that Kohls lacked standing because his labeled parody videos were outside the statute’s scope, and that Franson had waited too long to seek preliminary relief.

The Eighth Circuit later agreed, at least for purposes of preliminary relief. The appellate panel reasoned that because Kohls labeled his videos as parody, they were not realistic representations that could reasonably be understood as statements of actual fact under the statute. The court also upheld the denial of Franson’s request for a preliminary injunction because of delay. In April 2026, the appellate court declined to revive the challenge, leaving the earlier ruling in place. That does not mean every constitutional question has vanished, but it does mean the law survived an important early test.

The AI Hallucination Twist Nobody Ordered

The case also produced a moment so on-the-nose that a screenwriter might be told to tone it down. In defending the law, Minnesota relied on expert material about misinformation and AI. One expert declaration contained citations that turned out to be fabricated by artificial intelligence. A federal judge excluded that expert testimony, noting the irony of an AI misinformation expert relying too heavily on AI in a case about the dangers of AI misinformation.

That episode became a cautionary tale inside a cautionary tale. It did not end the case, and the court still declined to block the law, but it illustrated one of the central problems of the AI era: synthetic content can appear polished, authoritative, and confident while being completely wrong. If courts, campaigns, lawyers, journalists, and voters do not verify what AI produces, the machine can walk everyone into a wall while politely announcing, “Destination reached.”

X Corp. Enters the Fight

The controversy expanded when X Corp., the social media platform owned by Elon Musk, sued to challenge Minnesota’s law. X argued that the statute violates free speech protections and conflicts with Section 230, the federal law that generally shields platforms from liability for user-generated content. X’s concern is that platforms could face pressure to remove ambiguous political material rather than risk criminal liability.

That argument matters because modern political speech happens largely on platforms. A law aimed at deceptive AI content may affect not only the people who create deepfakes but also the services that host, recommend, label, downgrade, or remove them. If platforms believe the safest legal move is to delete anything remotely risky, protected parody and commentary could disappear along with harmful deception. That is the classic moderation dilemma: leave too much up and you may mislead voters; take too much down and you may censor debate.

At the same time, defenders of the law argue that platforms should not be passive bystanders when realistic synthetic media threatens elections. They say platforms already make countless moderation decisions and have tools such as labels, community notes, warning screens, and reduced distribution. The key policy question is whether criminal law is the right instrumentor whether disclosure rules, civil remedies, platform transparency, and rapid correction systems would do the job with fewer constitutional risks.

Minnesota Is Part of a Larger National Trend

Minnesota is not alone. States across the country have moved quickly to regulate election-related deepfakes. Some states prefer disclosure rules, requiring synthetic political media to be labeled. Others, including Minnesota, have taken a prohibition approach for certain deceptive uses. The disclosure model is often seen as less risky under the First Amendment because it allows speech while giving voters context. The ban model is stronger medicine, but stronger medicine usually comes with a longer warning label.

The national surge is easy to understand. Lawmakers do not want to wait until a fake candidate video goes viral the night before an election. Public confidence in elections is already fragile, and AI-generated media can exploit that fragility. A realistic fake does not merely mislead; it can also create a “liar’s dividend,” where real damaging evidence is dismissed as fake. Once voters believe anything can be fabricated, accountability itself becomes easier to dodge.

Still, state laws must fit inside constitutional guardrails. Courts are likely to ask whether a law is narrowly tailored, whether it targets actual deception rather than broad political persuasion, whether it protects satire and parody, and whether it gives speakers clear notice of what is prohibited. A law that tries to solve every AI problem at once may end up solving none of them after a judge takes out the red pen.

What the Law Gets Right

Minnesota’s law gets one big thing right: timing matters. Election deepfakes are most dangerous when released close to voting, when there is little time to investigate, debunk, and distribute corrections. A fake video in January may be ugly; a fake video the weekend before Election Day may be explosive. By focusing on election windows, the statute tries to address the period when deception can do the most damage.

The law also tries to distinguish AI-generated impersonation from ordinary criticism. It does not ban all negative ads, all false claims, or all synthetic content. Its target is realistic media showing people saying or doing things they did not say or do. That focus is important because voters process video and audio differently from text. A fake quote in a post may be challenged quickly, but a realistic video can feel emotionally true before the brain has time to ask for receipts.

Where the Law Could Run Into Trouble

The hardest problem is parody. Political satire has always relied on exaggeration, impersonation, and absurdity. If a law does not clearly protect labeled parody, creators may avoid making lawful commentary because they fear prosecution. The Eighth Circuit’s interpretation helped Minnesota by treating labeled parody as outside the statute’s reach, but critics argue the text does not expressly create a parody exception. That debate is likely to continue.

Another issue is intent. Proving that someone intended to injure a candidate or influence an election can be difficult, especially online, where users repost content for many reasons: agreement, mockery, outrage, confusion, or because the algorithm whispered, “Engagement, my precious.” A vague or unpredictable intent standard can chill speech because people cannot confidently know what conduct crosses the line.

Finally, platforms raise a separate concern. If a platform does not create a deepfake but hosts it, recommends it, or fails to remove it, how much responsibility should it bear? Section 230 questions, state criminal enforcement, and platform knowledge requirements are all legally complicated. The more a law appears to pressure platforms into broad censorship, the more intense the challenge becomes.

Practical Examples: What Could Be Covered?

Imagine a realistic AI audio clip released two days before an election that makes a mayoral candidate appear to tell supporters not to vote. If the creator knows it is fake, lacks consent, and intends to influence the result, that is exactly the kind of conduct Minnesota’s law is designed to address.

Now imagine a clearly labeled cartoon video showing a candidate as a talking raccoon promising free cheese to every household. That may be weird, but weird is not illegal. It is not realistic, and it is plainly commentary. American politics has room for talking raccoons, whether it deserves them or not.

The difficult case is a slick AI video that looks realistic but includes a tiny disclaimer, or a parody label that disappears when the clip is reposted. That is where courts, platforms, and campaigns will wrestle with context. A label that is obvious on YouTube may be missing in a screenshot. A joke that is obvious to the creator’s followers may fool voters seeing it cold in a group chat. Deepfakes travel faster than context, and context is usually wearing flip-flops.

Experience-Based Guide: How Campaigns, Voters, and Platforms Should Respond

For campaigns, the first practical lesson is to prepare before a deepfake appears. Every serious campaign should have a rapid-response plan for synthetic media. That means designating a verification team, maintaining direct contacts with local reporters and election officials, preparing public statements in advance, and monitoring major platforms without turning the campaign office into a caffeine-powered panic room. If a fake video drops near Election Day, the campaign that already has a plan will move faster than the campaign trying to remember who has the login password.

Second, campaigns should authenticate their own media. Posting original files, using trusted channels, watermarking official videos, and consistently labeling AI-assisted content can help voters distinguish real campaign materials from synthetic ones. This will not stop every fake, but it raises the cost of deception. Think of it like locking your car: it does not end car theft, but it encourages the thief to go bother someone less prepared.

For voters, the best habit is emotional patience. Deepfakes are designed to trigger instant outrage, shock, or disgust. When a clip makes you want to share it immediately, that is the moment to slow down. Check whether trusted news outlets have reported it. Look for the original source. See whether the candidate, campaign, or election office has responded. Be especially skeptical of dramatic content released very close to voting, when corrections have less time to catch up.

Voters should also treat platform labels as helpful but not magical. A warning label can provide context, but labels can be missing, wrong, delayed, or stripped away when content is reposted. Community notes and fact checks are useful tools, not seat belts made of titanium. The safest approach is to combine platform signals with outside verification.

Journalists have a special role. Newsrooms should avoid amplifying deepfakes while debunking them. A headline that repeats the false claim too loudly can spread the deception further. Responsible coverage should explain what is fake, how it was verified, who is affected, and what voters should know. The goal is not to give the fake a second life with better lighting.

Platforms should invest in friction, not just deletion. In many cases, slowing the spread of suspected synthetic media, adding context, preserving evidence, and notifying affected campaigns may be more effective than a simple remove-or-ignore choice. A platform that deletes too broadly invites censorship concerns; a platform that does nothing becomes a free storage unit for election fraud cosplay.

Finally, lawmakers should learn from Minnesota’s challenge. The best deepfake laws will be narrow, clear, and focused on deceptive impersonation that causes concrete election harm. They should protect satire, parody, news reporting, and good-faith platform moderation. They should also encourage fast civil remedies and public corrections, because elections operate on clocks, not law review timelines. A perfect ruling six months after Election Day is like a weather report delivered after the picnic has floated away.

Conclusion

The challenge to Minnesota’s political deepfake law is not just a local legal dispute. It is an early test of how American democracy will handle synthetic media in the age of generative AI. Minnesota has a legitimate interest in protecting voters and candidates from realistic digital forgeries. Critics have a legitimate concern that broad bans on political content can chill satire, criticism, and platform speech.

The strongest path forward is not pretending deepfakes are harmless, and not pretending every fake image requires a criminal prosecution. The answer is precision: laws that target knowing, harmful, realistic deception; platforms that add context quickly; campaigns that prepare before crisis hits; journalists who verify before amplifying; and voters who pause before sharing. Political deepfakes may be new technology, but the civic habit they test is old-fashioned: don’t believe everything you see, especially when it arrives wrapped in outrage and asks you to click “share.”