Deepfakes Hit 38 Countries. Newsrooms Still Don't Have a Workflow.

Full Episode Transcript

A deepfake video of Indian political commentator Dhruv Rathee spread to more than thirty-one thousand viewers in West Bengal before anyone debunked it. A detection tool flagged the audio as A.I.-generated with over ninety-eight percent confidence. The tool existed. It worked. But nobody ran it until the damage was already done.

Trusted by Investigators Worldwide

Run Forensic-Grade Comparisons in Seconds

Court-ready facial comparison reports. Results in seconds.

Get Started

7-day refund guarantee**

That gap — between having the technology to catch a

That gap — between having the technology to catch a fake and actually using it before it spreads — is the story of 2026. And it doesn't just matter if you verify media for a living. If you've ever shared a political clip, forwarded a news video to a family member, or voted after watching something online, this is about you. According to Surfshark's research, thirty-eight countries have faced deepfake incidents during elections since 2021. Those incidents touched three point eight billion people. That's roughly half the planet. And among eighty-seven countries that held elections from 2023 onward, more than a third experienced at least one deepfake event. The Rathee video in India wasn't a one-off. It was one data point in a pattern that's now global. So why are most newsrooms and investigative teams still treating verification like an afterthought instead of a standard step?

Start with what happened in India. During West Bengal's election season, a video surfaced showing Rathee — a well-known commentator — appearing to praise the B.J.P. The Quint's fact-check unit, WebQoof, ran the clip through Hive Moderation's A.I. content detector. The tool came back with a confidence score above ninety-eight percent that the speech was synthetically generated. Ninety-eight percent. And yet tens of thousands of people had already watched it, believing it was real. The detection technology performed exactly as designed. The problem was that no one in the distribution chain used it before the video went viral.

Now move to the United States. In March 2026, the National Republican Senatorial Committee released a minute-long ad targeting Democratic Senate candidate James Talarico in Texas. The ad showed Talarico appearing to speak — realistically, for the entire clip. According to TrueScreen's analysis, it was the first known political deepfake where a candidate appears to deliver a full, coherent statement that wasn't real. The ad did carry an A.I. disclosure label. But that label appeared on screen for just a few seconds, in text so small it was practically unreadable. A professional political organization, with a real budget and a national distribution network, produced and aired a synthetic video of a real candidate. That's not a prank on social media. That's a campaign tool. And for anyone who saw that ad on their phone or TV without catching the tiny disclaimer — they watched a fabricated person say fabricated things and had no reason to doubt it.

What about places where deepfakes showed up but didn't land? Researchers at Mila and McGill University studied Canada's 2025 federal election and found that about six percent of election-related images circulating online were deepfakes. Most of them were harmless or non-political. The ones that were genuinely misleading attracted almost no attention — accounting for barely a tenth of one percent of all views on X. So in Canada, the fakes mostly fizzled. But the researchers also found something important: when a fabricated image looked realistic, it drew higher engagement than obvious fakes. Reach, not existence, is what determines whether a deepfake actually changes minds. Most fakes fail quietly. The rare ones that break through are the ones that matter — and those are exactly the ones that need to be caught before they spread, not after.

The Bottom Line

Meanwhile, more than half of American adults — about fifty-eight percent — already expect synthetic lies to get worse before the next election. Public awareness is actually ahead of institutional readiness. People sense the problem. They just don't have the tools or the workflows to do anything about it. And neither, apparently, do most of the newsrooms and investigative teams responsible for catching these things. According to reporting from Pub Gen and Reality Defender, media verification is shifting from a quick gut check — does this look right? — toward something more like a repeatable forensic process. That means documentation standards, escalation pathways, and detection tools built into the editorial workflow before publication. Not after a video goes viral. Not after someone flags it on social media. Before. For investigators and compliance professionals, that's a fundamental change in how evidence gets authenticated. For everyone else, it means the next political video you see might have been checked by nobody before it reached your screen.

The shift isn't from "deepfakes are dangerous" to "deepfakes are more dangerous." We already knew that. The shift is that fabrication has become a professional, funded, deliberate campaign tactic — and verification hasn't kept pace as a professional, funded, deliberate response. The tools exist. The confidence scores are high. What's missing is the workflow that makes running them automatic.

So — deepfakes have now touched elections in thirty-eight countries, reaching billions of people. Detection tools can flag synthetic media with near-perfect accuracy. But almost no one is running those tools before a video spreads — not newsrooms, not platforms, not campaigns. Whether you investigate cases for a living or you just watched a political clip over breakfast this morning, the question is the same. Who checked it before you saw it? The full story's in the description if you want the deep dive.

Deepfakes Hit 38 Countries. Newsrooms Still Don't Have a Workflow.