AI Deepfakes and the 2026 Midterms: The Real Threat Nobody's Naming

Democracy Doesn't Need to Be Hacked. It Just Needs to Stop Believing Itself.

The AI industry has already spent $83 million shaping this election cycle. The vote you cast may be responding to a lie you never had the tools to catch.

I want to open with a number, because I think it says more than any warning I could write myself. Analysts tracking political deepfakes currently estimate a 68 percent chance that a fabricated AI video targeting a specific candidate surfaces within 96 hours of Election Day this November. That's not a random window. That's the exact stretch of time where there's no longer enough runway left to debunk something before people vote.

This already happened once this year, in March, when Senate Republicans released an AI-generated video making Democratic candidate James Talarico appear to say things he never said, in a voice and likeness that weren't real. It's August as I write this. Sit with what that timeline suggests about November.

I need to be completely upfront before I go further; this is not a piece about which party is worse. Both sides are doing this. That's precisely the problem, and it's the reason I think this deserves a harder, more uncomfortable kind of attention than most coverage of it gets.

The Money Behind the Machine

Here's what almost nobody talking about this seems to mention. AI companies and allied groups have spent at least $83 million on federal elections this cycle, with competing super PACs tied to Anthropic and OpenAI already pumping over $38 million into congressional primaries. In some races, AI-industry-backed political groups have spent more money supporting a candidate than that candidate has spent on their own campaign.

I want you to actually sit with that, not skim past it. When a technology company is outspending the person running for office, whose interest is that election actually built to serve? That's not a rhetorical flourish. That's a structural question about who gets to shape the information you vote on, and it deserves a real answer, not a shrug.

Why "Believable Enough" Is the Actual Threat

At least 15 documented AI-generated political ads ran between November and March alone, and that number is almost certainly higher by now. But here's the mechanism that should concern you more than any single fake video; AI-generated content doesn't need to be perfect. It needs to be believable enough for six seconds, long enough for someone to scroll past it and let a false impression settle in before their conscious mind ever catches up. A Kentucky race saw fabricated content involving a candidate and other lawmakers. A separate clip showed a candidate appearing to flee a rally. Neither carried clear disclosure. Neither needed to be flawless. They only needed to be plausible enough for a thumb mid-scroll.

The Legal Vacuum Nobody's Talking About

Here's the part that genuinely stopped me when I dug into it. There is currently no federal law regulating AI in political advertising. None. The FEC voted in September 2024 not to even open formal rulemaking on the issue, effectively punting it down the road. A bill, the AI Transparency in Elections Act, HR9922, would require disclaimers on AI-generated political content. It remains pending. It is not law.

Thirty-one states have passed their own deepfake election laws as of mid-2026, which means state law is currently the only real enforcement mechanism protecting you at all. California's AB2655 requires disclosure within 120 days of an election. Michigan's HB5144 requires disclosure at all times, not just near elections. Colorado and Utah go further, requiring metadata showing exactly when content was created and who made it. Several states have zero requirements whatsoever. Whether you have any real protection right now depends entirely on your zip code.

This Isn't Just Domestic

U.S. intelligence officials have directly confirmed that Russia, Iran, and China are actively using AI in election interference efforts, right now, as this cycle unfolds. Russia leads in volume, producing fabricated text, images, audio, and video designed to erode trust and deepen division. China has reportedly generated fake news anchors and pro-China propaganda aimed at American audiences. Here's the part I think people consistently miss; a foreign actor doesn't need you to vote a specific way. They need you to stop trusting the process itself. A divided, distrustful electorate is a weaker country. CISA's own 2026 assessment confirms this is active and ongoing, not hypothetical.

Where Stoic Philosophy Actually Earns Its Keep Here

I think about Marcus Aurelius often on this platform, and this is exactly the kind of moment his writing was built for. The Stoics drew a hard, disciplined line between what is within your control and what isn't. You cannot control whether a foreign intelligence agency floods your feed with fabricated video. You cannot control whether a super PAC outspends a candidate three to one using synthetic content. Those sit entirely outside your command, and burning your energy raging against them accomplishes nothing except exhausting you into the very apathy these actors are hoping for.

What remains fully, completely within your control is your own judgment, your willingness to pause before you believe, before you share, before you let six seconds of manufactured footage rearrange how you see another human being. Epictetus put it plainly; it is not events that disturb people, it is their judgments about events. The deepfake itself is the event. Your reaction to it, your verification of it, your refusal to be moved by it before you've checked, that's the judgment. That's yours. Nobody, not a super PAC, not a hostile government, can take that away from you unless you hand it to them by scrolling too fast.

What You Can Actually Do About It

• Learn to spot the tells: unnatural blinking, inconsistent lighting, strange hand movements, and audio that doesn't quite match the mouth are still common giveaways AI hasn't fully solved

• Verify before you share, check whether a clip appears on a candidate's official channels or a reputable outlet before you forward it to anyone

• Know your state's specific disclosure laws, and report political ads that don't comply, this is genuinely one of the few tools with real teeth right now

• Go to your state's official Secretary of State website for anything related to polling locations, deadlines, or ballot measures, never social media

Where I Land

I'm not telling you who to vote for. I'm telling you to make sure the person asking for your vote is actually who they claim to be, saying what they actually said. If you can't fully trust what you see or hear from any candidate or party, what does that do to your basic ability to make an informed decision, the single most fundamental function democracy asks of you? And does it actually matter whether the lie that convinces you came from a rival party or a foreign intelligence service, if the outcome, a manipulated vote, looks exactly the same either way?

This is precision over panic, not fear. You are capable of verifying before you believe. That discipline, exercised consistently, is the one thing no algorithm and no adversary can take from you unless you give it away first.

➡️ Related: [They're Cloning the News to Steal Your Money] 

➡️ Related: [The Face You're Watching Isn't Real, and Neither Is That Investment Tip] 

➡️ Join the free Tech 4 Grown-Ups community: tech4grownups.com/community

➡️ Listen to the full podcast: tech4grownups.com/podcast

Michael Routhier is the founder of Tech 4 Grown-Ups, providing honest, unfiltered digital literacy for adults 55+, and host of The Virtuous Machine, exploring the ethics and human cost of AI. Read by tech-curious readers in 50+ countries. Explore more at tech4grownups.com.