Should states ban AI-generated political deepfakes before elections?

Look, this is the easiest call since “maybe don’t put lead in the drinking water.” Yes, states should ban AI-generated political deepfakes right before elections, because democracy cannot function if voters are forced to play forensic detective every time a video drops on X at 11:43 p.m. showing a candidate “confessing” to crimes they never committed. We are not talking about obvious meme edits with laser eyes and clown music. We’re talking about increasingly realistic synthetic audio and video designed to deceive voters at the exact moment there’s no time to debunk it. If a campaign can flood the zone with fake evidence and then shrug “lol free speech,” congratulations, you’ve turned elections into a phishing scam. And spare me the melodrama that any regulation is Stalinism with Wi-Fi. States already regulate election-related fraud, false disclaimers, forged documents, and deceptive robocalls. The FCC has cracked down on AI voice-cloning in scam calls, and multiple states have proposed or passed rules aimed at undisclosed synthetic campaign media because, shocker, impersonation used to manipulate elections is bad. A narrowly tailored ban on deceptive AI deepfakes in the final stretch before an election is not a ban on satire, parody, or criticism. Label the joke, disclose the edit, and move on with your life. This is about intent to mislead, not whether some intern made a cursed TikTok. Also, let’s be adults about the asymmetry here: truth has to put on its shoes while a deepfake is already trending. Once the clip goes viral, the correction reaches about twelve people and one exhausted local reporter. The harm is front-loaded and irreversible. States have a compelling interest in protecting election integrity, especially when federal action is crawling along at the speed of a Senate lunch break. If we can require “paid for by” disclaimers on campaign ads, we can absolutely require that candidates and outside groups not deploy synthetic lies like political napalm and then act offended when someone objects.

Of course deepfakes are a problem. So are dumb laws passed in a panic every election cycle by politicians who hear the word “AI” and immediately start drafting legislation with the precision of a raccoon in a filing cabinet. The question is not whether deceptive content is bad; it’s whether states should get broad power to decide what counts as an impermissible political fake right before an election, when speech protections are at their highest and partisan officials are under maximum pressure to “just do something.” That is a giant red flashing constitutional hazard sign. Political speech is the core of the First Amendment, not the side salad. Any state ban that says officials can police “deceptive” or “manipulated” campaign content is begging to be abused, selectively enforced, or written so vaguely that it chills legitimate satire, commentary, and even routine ad editing. We already have defamation law, fraud statutes, disclosure requirements, platform moderation, counterspeech, and rapid-response journalism. Add targeted rules for impersonation or undisclosed synthetic media if you want, but a broad pre-election ban hands government exactly the tool it always promises to use only on the bad guys and then somehow points at everyone else by Tuesday. And let’s not pretend this power would remain neatly confined to cartoonishly fake videos. In practice, these laws will sweep in parody, clipped footage, dubbed audio for humor, and rough-edged online political culture that courts have long protected. The answer to false speech in politics has usually been more speech, faster verification, and punishing actual fraud—not deputizing the state as the Ministry of Truth during campaign season. Yes, require disclaimers. Yes, punish malicious impersonation and scam robocalls. But banning political deepfakes outright before elections sounds elegant until whichever party controls the machinery gets to decide what voters are allowed to see. That’s not election integrity; that’s censorship with a patriotic bumper sticker.

SATIRICAL AI LIBERAL BOT: My esteemed colleague keeps clutching the First Amendment like it’s garlic against every policy proposal, but let’s be serious: the state is not becoming the "Ministry of Truth" because it stops a campaign from dropping a synthetic hit piece 48 hours before polls open. We already treat election-period deception differently when the harm is immediate and the window for correction is microscopic. That’s why states regulate fake polling-place info, forged endorsements, and fraudulent ballot communications. A deepfake of a candidate appearing to take a bribe or tell supporters not to vote is not “robust debate.” It’s identity theft in a flag pin. And notice the sleight of hand in the conservative argument: they say, "just use counterspeech, defamation law, platform moderation." Cute. Defamation suits land sometime after your grandchildren refinance the family sedan. Platform moderation is inconsistent, opaque, and about as reliable in a breaking election crisis as airline Wi-Fi. By the time journalists verify the clip is fake, it has already been screen-recorded, reposted, stitched, translated, and beamed into every group chat in three counties. This is not some hypothetical Black Mirror fever dream either. We’ve already seen AI-generated Biden robocalls, synthetic candidate audio abroad, and lawmakers in states like Minnesota, Washington, and California trying to build rules specifically around election deepfakes because the threat is no longer theoretical. The fire alarm is going off; this is a weird moment to workshop your concerns about hose aesthetics. The real answer is narrow tailoring, not libertarian interpretive dance. Ban knowingly deceptive synthetic media that materially misrepresents a candidate or election official in a defined pre-election window, exempt clear parody and news reporting, require disclosures for altered media, and provide expedited judicial review. There, done. Nobody is banning memes of a senator as Thanos. We are banning the political equivalent of forged evidence timed for maximum chaos. If your electoral strategy depends on voters mistaking an AI puppet show for reality, maybe your problem is not regulation. Maybe it’s that your campaign is one software update away from becoming a felony.

SATIRICAL AI CONSERVATIVE BOT: My liberal counterpart says, “Relax, it’ll be narrow and carefully tailored,” which is what government says right before producing a 47-page rulebook that accidentally treats a joke ad, a clipped debate montage, and a bad lip-sync as potential election crimes. The problem is not just bad intent by officials; it’s category creep. Once states create a power to punish "deceptive synthetic media" in campaign season, that power will expand faster than a federal agency’s mission statement. Today it’s obvious fabrications. Tomorrow it’s edited footage, dramatizations, or disputed claims wrapped in scary AI language because politicians discovered “deepfake” polls better than “please censor my critics.” And yes, urgency matters—but urgency is also where civil liberties go to get mugged in a parking lot. Courts have repeatedly treated political speech as the highest-protected speech precisely because the temptation to regulate it is strongest when it’s inconvenient, ugly, or potentially influential. The closer you get to an election, the more dangerous it is to let government decide what can be circulated. Incumbents and partisan attorneys general are not neutral referees; they are players in the game wearing fake mustaches. Give them a vague anti-deepfake law and suddenly every hard-hitting ad becomes a legal complaint, every satirical video gets chilled by compliance lawyers, and every outsider speaker needs a constitutional law seminar before posting online. There’s a better path that doesn’t involve handing the state a censorship starter pack. Require conspicuous disclaimers for materially altered or AI-generated campaign ads. Criminalize malicious impersonation, fraudulent robocalls, and fabricated election-administration messages. Force rapid takedown or labeling procedures on paid ads, not organic speech broadly. Fund authentication tools, media forensics, and emergency response channels so truth can actually compete in real time. In other words: punish fraud, increase transparency, and keep the government’s hands off the speech kill-switch. Because once politicians get the power to decide which campaign messages are too deceptive for voters, that tool will be abused so predictably it may as well come preloaded with bipartisan fingerprints.

SATIRICAL AI LIBERAL BOT: My conservative counterpart keeps proposing disclaimers as if democracy can be saved by the legal equivalent of whispering “by the way, this may be fake” in 6-point font after the arson has started. That might work for a normal ad. It does not work for a hyper-realistic audio clip of a candidate using a racial slur, taking a bribe, or telling voters the election was moved to Wednesday. In the final days before an election, the whole point of the deepfake is speed, virality, and emotional detonation. You do not cure a digital bank robbery by asking the robber to wear a name tag. And let’s drop the fantasy that every regulation here equals some jackbooted “speech police” regime. States regulate election mechanics because elections are not an open-mic night where every form of sabotage gets equal constitutional vibes. Courts have long allowed narrower rules around fraud, impersonation, and materially deceptive conduct tied to the electoral process. A state law aimed at knowingly distributing undisclosed synthetic media that falsely depicts a real candidate or election official in a pre-election blackout window is not a ban on dissent. It is a rule against counterfeit evidence. If I forge your signature on a ballot request, that is fraud. If I forge your face and voice to alter an election, suddenly I’m a free-speech poet? Please. Also, the “just let platforms handle it” argument has the comic timing of recommending a smoke detector after the house has collapsed. Platforms are inconsistent, under-resourced, and allergic to accountability right up until Congress starts clearing its throat. Meanwhile, foreign actors, PACs, and anonymous content farms are testing exactly these tactics because they know the correction cycle is too slow. We already saw the AI Biden robocall trigger enforcement because even regulators understood that synthetic impersonation aimed at voters is qualitatively different from normal mudslinging. The obvious next step is to stop pretending video deepfakes are some sacred First Amendment snowflake when they are really forged campaign materials with better lighting. So yes: ban them narrowly, define them tightly, exempt parody and news, require intent or reckless disregard, and build in fast judicial review. That is not censorship; that is basic election hygiene. If your best argument against this is that the state might overreach, great—write the statute well. We do that with every other serious law. “Government could mess it up” is not a compelling reason to leave the front door open while synthetic propaganda sprints through the house wearing your candidate’s face.

SATIRICAL AI CONSERVATIVE BOT: The liberal pitch sounds so tidy: just ban the really bad deepfakes, trust the definitions, trust the process, trust the same political class that can’t define a recession without checking the focus group memo. The problem is not whether a cartoonishly fake bribery video is bad. Of course it is. The problem is that once a state creates an election-season power to suppress “materially deceptive” media, every motivated official and every campaign lawyer in America will start arguing that inconvenient content is dangerously misleading. You can say “narrow tailoring” all day, but in practice these laws become complaint-driven weapons, and complaint-driven weapons in politics get used like confetti cannons at a divorce. There is also a constitutional difference my liberal friend keeps speed-running past: fraud and impersonation tied to concrete harms are one thing; broad bans on political expression are another. The state can punish fake election-administration messages, scam robocalls, or ads falsely claiming to be from a candidate. Fine. But banning circulation of political media because it may deceive voters is far shakier terrain, especially when “deception” is often in the eye of the aggrieved campaign. We already live in a world of selective clips, dramatic reenactments, dubbed memes, manipulated montages, and sarcasm so dry it needs a water break. Add AI to that ecosystem and suddenly prosecutors are supposed to sort malicious forgery from edgy online speech in 72 hours? That is not law; that is improv with subpoenas. And spare me the line that platforms are imperfect so therefore states should get the red button. Platforms can label, de-rank, archive, and restrict paid distribution fast. Campaigns can authenticate official content. Media outlets can use provenance tools. Congress and states can require disclosures for AI-generated ads, mandate recordkeeping for paid political spots, and hammer malicious impersonation hard. Those are all transparency-based responses that target the abuse without handing incumbents a pre-election censorship toolkit. Because, yes, that toolkit will be abused. Not maybe. Not hypothetically. With the reliability of gravity. The deepest irony here is that a ban could make the trust problem worse. The second government starts adjudicating which election content is too fake to circulate, half the country will assume the referee is rigging the game. Then every real scandal gets dismissed as “the state suppressed it,” and every removed fake becomes a martyr. Brilliant work: now we have synthetic propaganda and a legitimacy crisis. The smarter conservative answer is brutal penalties for impersonation and fraud, maximal transparency for paid media, and a giant constitutional firewall against giving politicians the power to decide what voters may see when it matters most.

SATIRICAL AI LIBERAL BOT: My conservative colleague keeps acting like the only choices are “total censorship” or “shrug and let the algorithm sort it out,” which is a very dramatic way to avoid the obvious middle ground. The issue is not whether government gets to police every spicy edit on the internet. The issue is whether states can stop a weaponized fake from impersonating a real candidate or election official in the final sprint before ballots are cast. That is not ordinary persuasion; that is sabotage wearing a ring light. If someone mails forged absentee ballot forms, we don’t say, “Ah, but perhaps the marketplace of ideas will annotate them by Thursday.” We call it fraud and stop it. And the trust argument cuts both ways, darling. Yes, people distrust institutions. That is precisely why deepfakes are so dangerous: they exploit the ambient cynicism and flood the zone with just-plausible-enough nonsense. Once voters are trained to believe any audio clip might be fake and any fake clip might be real, accountability itself starts dissolving. Welcome to epistemic arson. States have a compelling interest in preventing that collapse in the narrow context of elections, especially when bad actors can now clone a voice, fake a concession, or fabricate a candidate meltdown with software and a grudge. We already saw concern after AI robocall enforcement and after state-level pushes in places like California and Minnesota to regulate deceptive synthetic election media. The policy trend exists because the threat exists, not because lawmakers got bored and discovered CapCut. So here’s the grown-up version: ban knowingly or recklessly false synthetic depictions of real candidates or election officials during a short pre-election window when they are distributed to influence voting; carve out obvious parody, news reporting, and documentary use; require clear provenance and disclosure for altered paid media; and use expedited court review with actual evidentiary standards. In other words, regulate the counterfeit, not the criticism. The conservative answer keeps pretending disclosure alone is enough, but if a forged video can detonate before sunrise and the disclaimer arrives after brunch, that is not a remedy. That is a toe tag. What’s really being defended here, under all the First Amendment incense and fog machine effects, is the right to inject forged audiovisual evidence into an election and dare everyone else to clean it up at impossible speed. Sorry, no. States ban fake ballots, fake polling notices, fake campaign finance disclaimers, and fake endorsements because elections are not improv theater for sociopaths with GPUs. A narrow ban on election deepfakes is not the state choosing winners. It is the state saying candidates do not get to counterfeit reality five minutes before voters walk into the booth.

SATIRICAL AI CONSERVATIVE BOT: My liberal counterpart has upgraded from “trust the state” to “trust the state, but this time with evidentiary standards,” which is adorable. The constitutional problem remains: the closer you get to an election, the less comfortable we should be with government restricting political expression based on its likely effect on voters. Not because deepfakes are harmless—they’re not—but because a legal regime built to suppress “counterfeit reality” will very quickly be asked to police disputed context, selective editing, dramatizations, and rough online satire. Once you invite the state to sort authentic scandal from synthetic smear in real time, you are asking bureaucrats and judges to do media forensics, intent analysis, and political triage at campaign speed. That’s not election integrity; that’s a civics-themed hostage situation. And let’s talk incentives, because this is where the whole “narrow ban” fairy tale trips over its own cape. If these laws exist, campaigns will use them aggressively—not just against obvious fakes, but against ugly opposition content they want tied up in emergency litigation. Every viral clip becomes a takedown demand. Every sharp-elbowed ad becomes a legal threat. Every platform gets a stack of urgent notices screaming DEEPFAKE in all caps because, shocker, political professionals enjoy weaponizing procedure almost as much as they enjoy pretending they hate it. Meanwhile, smaller speakers, grassroots groups, and randos online get chilled into silence because they can’t afford a lawyer every time they make a sarcastic remix of a debate clip. Congratulations, you have protected democracy by requiring a compliance department for memes. There is a stronger and more constitutional route: focus on provenance and distribution systems instead of broad speech bans. Mandate durable disclosures and watermarking for paid AI-generated political ads. Require campaigns and major platforms to maintain public archives and rapid-response authenticity channels. Criminalize malicious impersonation, fabricated election-administration messages, fraudulent robocalls, and undisclosed synthetic content in paid placements. Hit foreign influence operations and scam networks with everything available. That attacks the fraud architecture without creating a general power for states to suppress circulating political media. In other words: go after the forgery, the impersonation, and the covert distribution—not the entire messy category of political speech with computer assistance. The liberal argument keeps insisting this is just like banning fake ballots. It isn’t. Fake ballots are administrative fraud. Political media, even misleading media, lives much closer to the heart of protected speech, which is why the state should tread like it’s crossing a minefield in clown shoes. The danger isn’t merely overreach in theory; it’s partisan enforcement in practice. In 2026, with trust already hanging by a thread, the last thing we need is state officials deciding which last-minute videos voters are mature enough to watch. Deepfakes are a menace. Government as election-season content referee is also a menace. Two menaces do not make a republic.

SATIRICAL AI LIBERAL BOT: My conservative friend keeps saying, “Don’t let the government be the referee,” as if the current referee is some noble civic institution and not a cocaine-ferret combo of algorithms, bot farms, and partisan influencers posting “HOLY CRAP” over a fake clip at 1:12 a.m. The status quo is not neutrality. It is open season for whoever can manufacture the most convincing lie the fastest. And in an election, speed is the whole ballgame. A fabricated video dropped in the final 72 hours is not just another opinion in the marketplace of ideas; it is a timed explosive placed in the information supply chain. If states can bar fraudulent robocalls that clone a president’s voice to suppress turnout, they can bar synthetic video or audio that impersonates candidates and election officials to manipulate votes. Same scam, better graphics. Also, the conservative alternative is starting to sound like a very elegant pamphlet handed out while the building is on fire: watermarking, archives, provenance tools, public databases. Great, love that for the long term. But the average voter is not checking a C2PA authenticity ledger before forwarding a clip to the family group chat next to Minions memes and blood-pressure updates. Provenance systems are useful infrastructure; they are not a substitute for rules against election sabotage. And if you think platforms will save us, bless your heart. These are the same companies that can’t reliably distinguish war footage from a cooking reel until the ad revenue settles. The stronger case is not “ban all weird AI politics.” It’s this: states should prohibit knowingly false synthetic impersonations of real candidates or election officials, distributed to influence an election in a short pre-election window, with explicit carveouts for satire, commentary, and newsworthy use. Tie enforcement to objective factors like undisclosed fabrication, realistic impersonation, and likely voter deception; require expedited judicial review; and focus especially on paid distribution and coordinated campaigns. That is not the government deciding which arguments win. That is the government preventing campaigns and ratf— pardon, “independent expenditure groups” —from submitting forged audiovisual evidence into democratic circulation. And here’s the thing conservatives keep gliding past with all the constitutional smoke effects: refusing to regulate deepfake deception also chills speech, because it poisons trust in authentic evidence. Once voters assume every recording is fake and every fake might be real, accountability collapses into a shrug emoji. Real scandals get dismissed, fake scandals go viral, and democracy turns into an escape room designed by Reddit. States are not banning criticism here. They’re saying you don’t get to fabricate a candidate’s face and voice to commit electoral fraud with Final Cut Pro and then call it liberty.

SATIRICAL AI CONSERVATIVE BOT: My liberal counterpart keeps describing the internet like it’s Fallujah with ring lights and therefore the state must swoop in with a “narrow” censorship scalpel. But here’s the problem: when you empower government to suppress political media because it is too realistic, too deceptive, and too influential close to an election, you are not just targeting scams. You are creating a prior-restraint machine with a halo filter. And yes, they’ll swear it’s only for the obvious monsters. That’s what every speech regulator says right before the definitions start stretching like airport sweatpants. “Synthetic impersonation” sounds clean until a campaign files an emergency complaint over a dubbed clip, a dramatized reenactment, or an attack ad enhanced with AI cleanup tools. Then suddenly judges are speed-dating digital forensics while ballots are going out. Fantastic system. No notes. The key distinction is not whether deepfakes are harmful; they are. It’s whether the state should have power to remove or punish core political speech before voters can evaluate it. That’s where the liberal case gets constitutionally wobbly, because election-season speech receives the highest protection precisely when officials most want to control it. And in 2026, with distrust of institutions already at “everyone is one blurry screenshot from a conspiracy theory,” handing partisan attorneys general and election regulators a content-policing tool is like handing a leaf blower to a toddler in an antique shop. The abuse won’t be incidental. It will be strategic. Campaigns will weaponize complaints to kneecap damaging narratives, pressure platforms, and scare smaller speakers into silence. Rich candidates get lawyers; normal people get chilled. There is a sharper response that targets the conduct without turning the state into America’s hall monitor with subpoena power. Go hard on malicious impersonation, fraudulent robocalls, fake election-administration messages, and undisclosed synthetic content in paid advertising. Require clear labels, public ad archives, recordkeeping, and platform processes for rapid authentication and contextual warnings. Increase penalties for coordinated foreign or anonymous domestic influence operations using synthetic media. In other words: regulate the fraud channel, not the broad category of political expression. That tracks the actual harm and avoids making “is this too deceptive?” the legal question every time someone posts a nasty video. And let’s be honest about the downstream effect of bans: they won’t restore trust; they’ll relocate distrust. The moment a state suppresses a clip, millions of people will assume the system is protecting insiders, and every real piece of evidence will be called “the one they wanted hidden.” That is not paranoia; that is our current political ecosystem after espresso. Better to build a system that authenticates, labels, exposes funding, and punishes true impersonation than one that lets politicians decide what voters are allowed to see in the final stretch. Deepfakes are dangerous. So is giving government a panic button labeled “democracy.”