AI Deepfakes Are Flooding the 2026 Midterms and No One's Stopping Them

TL;DR:

The FEC is stuck in a 3-3 partisan deadlock. No federal rules exist for AI-generated political content
28 states passed political deepfake laws, but most only require disclaimers and some have already been struck down in court
Virginia Republican John Reid debated an AI deepfake of his opponent for nearly an hour after she refused real debates
Andrew Cuomo's campaign posted a racist AI-generated attack ad showing his opponent eating rice with his hands, then blamed a junior staffer
The FCC banned AI voices in robocalls, but that rule doesn't cover social media, streaming, or digital ads
Experts call it a "Wild West." Voters are on their own to sort real from fake

The Federal Regulators Who Won't Regulate

The Federal Election Commission could regulate AI in political ads. It has the legal authority. Democrats in Congress have been demanding it for two years. But the FEC won't do it.

Here's why: the commission has six members: three Democrats, three Republicans. Every AI regulation proposal has deadlocked 3-3. That means no rules pass. The agency evaluates complaints "case-by-case," which is regulator-speak for "we'll get back to you after the election."

Public Citizen, Protect Democracy, the Brennan Center, and the Campaign Legal Center have all filed formal petitions urging the FEC to clarify that existing fraud prohibitions apply to deepfakes. The FEC's response? They acknowledge the requests exist. Then nothing happens.

Senator Adam Schiff led a coalition of 50 Democratic Congress members supporting the Public Citizen petition. Still nothing.

Commissioner Ellen Weintraub has been on the FEC since 2002 and was elected chair in December 2025. She's now fighting to keep her position after Trump attempted to remove her. But even when Democrats were in the White House, the commission couldn't agree on AI rules. Partisan gridlock isn't a bug. It's how the agency works by design.

The Wild West Is Already Here

While regulators argue about jurisdiction, candidates are using deepfakes in real races.

The AI Debate That Wasn't

In October 2025, Virginia Republican John Reid wanted to debate his Democratic opponent, State Senator Ghazala Hashmi, in the lieutenant governor race. She declined.

So Reid created an AI deepfake of Hashmi and debated it for nearly an hour. Live-streamed the whole thing. The AI avatar mimicked Hashmi's appearance and voice, responding with answers scraped from her public statements and policy positions.

Hashmi's campaign called it "a desperate move straight out of Donald Trump's playbook." Reid was transparent about using AI. The fake didn't look perfectly real. But that misses the point. A candidate just demonstrated you can manufacture your opponent's policy positions and argue against them on camera. What happens when the technology gets better?

Experts predict 2026 will see far more AI content in campaigns. The visual and audio quality has improved dramatically. Detection tools haven't kept up.

Cuomo's Racist AI Ad

Andrew Cuomo's mayoral campaign embraced AI from the start. Ads showed deepfakes of himself operating a subway and washing skyscraper windows. Gimmicky but harmless.

Then came October 2025. Cuomo's campaign released an AI-generated attack ad titled "Criminals for Zohran Mamdani" targeting his opponent. The ad depicted a deepfake of Mamdani eating rice with his hands. It showed a Black man wearing a keffiyeh robbing a store.

The campaign posted it Wednesday night, then quickly deleted it after immediate backlash. They blamed a junior staffer. Mamdani won by a landslide.

Following the controversy, New York Governor Kathy Hochul announced she wants to ban AI-generated images from state politics entirely. Whether that survives a First Amendment challenge is another question.

The State Law Patchwork

With the feds paralyzed, states have scrambled to fill the gap. As of January 2026:

28 states have laws addressing deepfakes in political communications
38 states passed some form of AI legislation in 2025
47 states have enacted deepfake laws of some kind since 2019

That sounds like progress. But dig into the details and the picture darkens.

Most state laws don't ban deepfakes. They require disclaimers. That means you can still run an AI-generated ad showing your opponent doing something terrible, as long as there's fine print saying "materially altered." How many voters read the fine print?

Many laws only apply within a certain window before an election, typically 30 to 60 days. Campaigns can run deepfakes all summer, then switch to "real" content right before voting.

And some laws have already been challenged in court. Parts of California's AB-2839, requiring disclaimers on AI-altered political content, were partially struck down as unconstitutional.

Key State Laws

Texas SB753: Prohibits publishing deepfake videos intended to influence voters within 30 days of elections
Maryland SB0141: Criminalizes AI deepfakes spreading election misinformation, authorizing content removal, fines, and prison sentences
Missouri SB 509: Requires clear disclaimers on AI-generated political ads; legally holds humans responsible for AI-generated content
California AB-2355, AB-2655, AB-2839: Require disclaimers on AI-altered political ads, but portions have been struck down or challenged

The FCC's Narrow Victory

There's one area where federal regulation worked: robocalls.

In February 2024, the FCC ruled that AI-generated or voice-cloned audio counts as "artificial or prerecorded voice" under the Telephone Consumer Protection Act. That means robocalls using AI voices need prior express consent. Campaigns can't just deepfake a candidate's voice and blast it to millions of phones.

This matters because we've already seen it happen. During the 2024 New Hampshire primary, voters received calls from a faked version of President Biden's voice urging them not to vote. The FCC rule gives enforcement teeth against that specific tactic.

But here's the limit: the FCC only regulates telecommunications. The rule doesn't cover social media. It doesn't cover streaming. It doesn't cover digital ads. It doesn't cover anything you see on your screen.

A deepfake robocall is illegal. The same deepfake on X, YouTube, or Facebook? No federal law touches it.

The Liar's Dividend

The deepfake threat isn't just fake content passing as real. It's real content being dismissed as fake.

Security researchers call this the "liar's dividend." Once voters know deepfakes exist and are common, politicians can dismiss genuine evidence of misconduct as AI-generated. Got caught on video saying something problematic? Claim it's a deepfake. Your opponent has proof of corruption? Deepfake. Eyewitness video of an incident? Could be AI.

This erodes trust in all political information. When nothing can be verified, everything becomes deniable. The flood of synthetic media doesn't just deceive. It creates permission for everyone to deny reality.

What Voters Should Watch For

With no federal protection and state laws full of holes, you're largely on your own. Here's what to look for:

Check the source: Where did this video or audio first appear? Campaign content on official channels is more likely legitimate. Random viral clips with no origin deserve skepticism.
Look for tells: AI video struggles with hands, earrings, backgrounds, and consistent lighting. Audio deepfakes often have unnatural pauses or breathing patterns.
Wait for confirmation: Breaking video of a candidate saying something outrageous deserves 24 hours for verification before you share it.
Check for disclaimers: Many states require "materially altered" labels. Their absence doesn't prove authenticity, but their presence confirms AI involvement.
Cross-reference: Did the candidate really say that? Check their official campaign for a response. Major news outlets will flag confirmed deepfakes.

What Happens Now

Congress has bills sitting in committees. The NO FAKES Act would establish right-of-publicity protections against unauthorized use of someone's likeness or voice in deepfakes. It has bipartisan support. It hasn't passed.

State attorneys general are increasingly aggressive. California AG Rob Bonta has already sent cease-and-desist letters to AI companies over deepfake pornography. That enforcement posture could extend to political content.

But the midterm primaries are already starting. Congress returns February 23. Even if they act immediately (unlikely), any new law won't take effect for months.

For the 2026 midterms, this is it. Deepfakes are legal in most contexts. The FEC is deadlocked. State laws are inconsistent and challenged in court. The FCC only covers phone calls.

Voters will see synthetic media. Some will be obvious. Some won't. Sorting real from fake is now part of civic participation in America.

Sources

Published: February 20, 2026