Let’s not pretend AI deepfakes in elections are just the digital version of yard signs and campaign puffery. A fake image is one thing; a convincingly fabricated video of a candidate appearing to confess to corruption, insult veterans, or tell voters the wrong election date is something else entirely. We already know this isn’t hypothetical cosplay for ethics panels: in 2024, voters saw AI-generated robocalls mimicking President Biden’s voice in New Hampshire telling Democrats not to vote in the primary. That’s not "more speech." That’s fraud in a democracy costume. If your political strategy requires tricking people with synthetic lies that look and sound real, congratulations, you’re not defending liberty — you’re speedrunning election sabotage with a GPU.
The liberal case is pretty straightforward: regulate the deception, not the opinion. Require clear disclosure on AI-generated political ads. Ban deepfakes that impersonate candidates or election officials in ways meant to mislead voters about voting logistics or fabricate criminal conduct. Give platforms and campaigns rapid takedown obligations close to Election Day, because a lie that circulates for 48 hours before polls open doesn’t need to last forever to do damage. This is exactly why lawmakers in states like California and Minnesota have pushed rules on synthetic campaign media, and why members of Congress keep proposing federal transparency requirements. We label campaign finance, disclaim paid ads, and prohibit certain forms of voter intimidation already; apparently democracy somehow survived all that tyranny.
And no, this is not some wild plot to let the government become the Ministry of Memes. Satire, parody, and obvious humor can be protected — calm down, your cursed TikToks are safe. The point is that AI collapses the old common-sense barrier between "that sounds fake" and "oh wow, there’s video." Human brains are not built to do forensic media analysis at breakfast before work. In a world where synthetic content can be mass-produced by campaigns, super PACs, and foreign influence operations, refusing to set rules is not principled neutrality. It’s leaving the front door open and then delivering a TED Talk about the beauty of fresh air while the house gets robbed.
Deepfakes are a real problem, but handing the government broad power to police political speech right before elections is the kind of cure that usually comes with authoritarian side effects and a terrible FAQ page. Political speech is the most protected speech in America for a reason: the people in power are always one moral panic away from deciding their critics are "misinformation." Today the target is synthetic media; tomorrow it’s edgy satire, clipped videos, opposition research, or any ad some regulator says creates a "misleading impression." If you think bureaucrats and partisan appointees are going to neutrally referee what counts as deceptive in October of an election year, I have a pristine bridge to sell you, gently used by every agency that swore it would stay nonpartisan.
The conservative case is not "yay fake videos." It’s that existing law already covers a lot of the genuinely unlawful stuff: fraud, defamation, impersonation, and voter suppression. The Biden robocall incident in New Hampshire? Go after it aggressively. Fine. But don’t use a criminal scam as the excuse to build a sweeping censorship framework for lawful political advocacy. Disclosure requirements can sound modest until they become vague, selectively enforced speech burdens that privilege incumbents, legacy media, and well-lawyered campaigns while kneecapping grassroots speakers. And bans on "deceptive" media are catnip for abuse, because politics is full of exaggeration, editing, sarcasm, and ugly persuasion. Welcome to democracy — it’s messy, not machine-washable.
The better answer is narrower and more constitutional: punish direct impersonation and voter-interference schemes, require provenance tools where feasible, encourage platforms to authenticate official campaign content, and let rivals, journalists, and citizens expose fakes in real time. In other words, target conduct, not broad categories of speech. The moment government gets to decide which political media is too unreal for the public, every administration will be tempted to protect democracy by protecting itself. And somehow, mysteriously, the rules will always seem to break hardest against outsiders. Funny how that works.