Let’s do the obvious thing that Congress treats like advanced wizardry: set one clear national standard for AI deepfakes in politics before the internet turns every swing state into a YouTube comment section with a PAC budget. If an AI-generated ad can make it look like a candidate confessed to bribery, endorsed a coup, or personally invented inflation, that is not just "more speech" floating around the marketplace of ideas. That is industrial-scale deception aimed at voters on a deadline. And because federal elections are national events with digital content crossing state lines in seconds, a patchwork of 50 different rules is basically a gift basket for bad actors. They will upload from the weakest state, target the closest race, and dare everyone else to catch up.
A federal baseline does not have to mean some dystopian Ministry of Memes. It can mean narrow, content-neutral rules focused on disclosure, provenance, and fraud: label materially deceptive AI political ads, ban knowingly false deepfakes impersonating candidates or election officials close to Election Day, and require platforms to maintain transparent reporting channels. This is not radical. We already regulate campaign finance disclosures, robocalls, and deceptive voting information because elections are one of the few arenas where "lol just fact-check it later" is a truly terrible public policy. Recent pushes in Congress and the Federal Communications Commission’s scrutiny of AI-generated robocalls show the government already knows the problem is real.
And yes, free speech matters. But the First Amendment is not a coupon for fraud. Courts have long allowed regulation of impersonation, defamation, and deliberate deception in specific contexts. If anything, failing to act invites exactly the kind of chaos authoritarian-minded grifters love: flood the zone with fake clips, muddy reality, then shrug and say truth is just vibes now. Very chic, very postmodern, very bad for democracy.
States can still innovate on top of a federal floor, just like they do with consumer protection and election administration. But when the threat is fast, borderless, and targeted at national trust itself, pretending every state should freestyle its own AI rules is less "federalism" and more "good luck, America."
Washington’s answer to every new technology is the same: panic first, regulate broadly, and swear this time the bureaucracy will be precise. Deepfakes are a real problem, but handing the federal government sweeping power over political speech right before major elections is the kind of idea that should come with a warning label. Once D.C. gets authority to define what counts as "deceptive AI," that power will not stay neatly confined to obvious fraud. It will expand into gray areas like parody, satire, selective editing, and issue advocacy, all filtered through agencies and administrations that just happen to have political preferences. Fantastic. What could possibly go wrong?
States are better positioned to act quickly and tailor rules to local conditions, and many already have. Places like Texas, California, and Minnesota have moved on election-related deepfake laws, offering a live laboratory of approaches instead of one national speech code imposed from above. That matters because the legal and cultural line between fraud and protected expression is not simple. A clumsy federal law could easily chill legitimate campaigning, citizen activism, and even journalism, especially for smaller groups that cannot afford teams of lawyers to decode whatever the latest compliance memo says this week.
There is also a basic constitutional humility missing from the federal push. Elections are administered largely by the states, and the First Amendment places political speech at the very top of protected expression. If someone uses AI to impersonate an election official or spread false voting instructions, existing fraud laws, defamation law, consumer protection tools, and targeted state statutes can already do real work without creating a broad federal censorship architecture. The danger is not imaginary: recent years have given both left and right plenty of reasons to distrust centralized moderation systems and official truth-policing, especially when "misinformation" labels have sometimes been slapped on disputed claims before the facts were fully known.
The smarter path is targeted enforcement against clear harms, platform transparency, and state-led experimentation—not a one-size-fits-all federal regime written by people who still hold hearings asking if TikTok connects to home Wi-Fi. Deepfakes are dangerous, yes. But so is giving the same political class Americans already distrust a giant new lever over campaign speech. If democracy is the goal, maybe don’t protect it by centralizing control over what voters are allowed to hear.