AI Files: How AI Disinformation Spreads

A viral AI-generated video of a burning aircraft carrier sparks chaos as we're swept into the murky waters of war and misinformation. Uncover how deepfakes are reshaping our perception of reality—listen now!

Transcript

SPEAKER_04 0:00

Imagine this. You're scrolling through your feed and you see a video of the USS Abraham Lincoln, a massive American aircraft carrier, engulfed in black smoke and flames. The caption says it was struck by an Iranian missile in the Persian Gulf. It looks terrifyingly real. The lighting, the smoke physics, the panic on the deck, it's all there. But here's the kicker. None of it happened. The ship is fine. The video was generated by an AI in seconds, uploaded to a burner account, and shared a million times before the Pentagon could even issue a denial. This isn't science fiction, and it isn't a hypothetical. According to reports from the last 48 hours, this is the reality of the war we're watching unfold right now. It's Monday, March 16th, 2026. I'm Chloe, and you're listening to The Reality Check. Today we're looking at the explosion of AI disinformation that has defined the last month, from the kinetic battlefield in the Middle East to the political battlefield here in the U.S. midterm primaries. We're going to break down President Trump's recent claims about Iranian war fakes, dissect the technology making it possible, and then turn our gaze inward to how synthetic media is quietly reshaping the American left. Joining me to parse the signal from the noise is Alex, our senior analyst on digital influence operations. Alex, good to have you back. Let's start with the headline grabber. President Trump took to Truth Social a few days ago with a very specific, very fiery accusation. He claimed that Iran is, quote, militarily ineffective, but a master of media manipulation. He specifically called out images of kamikaze boats and that burning aircraft carrier I mentioned, labeling them as pure AI fabrications. He even went so far as to suggest treason charges for media outlets that spread them. Now, we know Trump uses hyperbole, but is he right on the facts here? Is Iran faking its war effort?

SPEAKER_01 2:07

On the core assertion regarding the imagery, yes, he is absolutely spot on. What we are seeing from Iranian state media and their proxy networks on platforms like Telegram and X is a textbook example of what we call evidence fabrication. The U.S. Central Command has confirmed that the USS Abraham Lincoln has not been hit, let alone set on fire. Those viral videos, some of which trended massively on Chinese social media last week, are synthetic. They are deepfakes.

SPEAKER_04 2:37

So they are creating victories out of thin air because they can't win them on the water?

SPEAKER_01 2:41

Exactly. This is asymmetric warfare in the digital age. If you look at the historical context, Iran has actually been doing this for decades, just with worse technology. Remember 2008, the Revolutionary Guard released a photo of a missile test with four missiles soaring into the sky. It turned out one of the missiles failed to launch, so they just photoshopped a fourth one over the dud. It was crude. You could see the cloned smoke pixels. But today, they don't need Photoshop. They're using generative diffusion models that can create photorealistic damage assessments of U.S. refueling planes, specifically the KC-135s Trump mentioned, that never happened.

SPEAKER_04 3:22

Okay, let's pause on the how for a second. You mentioned diffusion models. I think a lot of listeners hear AI and just think of like a magic button. But technically speaking, what is happening when someone creates a fake video of a burning ship? How does the computer know what that looks like?

SPEAKER_01 3:40

It's a fascinating, if terrifying, process. Diffusion models, which are the engine behind tools like mid-journey or the video generators we're seeing now, work on the principle of denoising. Imagine you have a crystal clear photograph of an aircraft carrier. Now imagine you slowly add static or digital noise to it until it just looks like a gray, snowy screen. The AI is trained to reverse that process. It learns how to take a screen of pure static and mathematically hallucinate a clear image out of it step by step.

SPEAKER_04 4:17

So it's sculpting an image out of static?

SPEAKER_01 4:20

Precisely. And because it's been trained on billions of images of ships, fire and smoke, when an Iranian propagandist prompts it with an American aircraft carrier on fire in the Persian Gulf, the model predicts what that specific arrangement of pixels should look like. It understands the physics of how smoke billows, how light reflects off water, and how fire chars metal. That's why these new fakes are so hard to debunk. They don't have the obvious cloned pixels of the 2008 Photoshop job. They are unique, mathematically generated images that have never existed before.

SPEAKER_02 4:58

That explains why the president's comments about the kamikaze boats struck a nerve. He called them non-existent props, but the Wall Street Journal and others reported on damage to U.S. assets. Is there a danger here that in calling out the fakes, we might dismiss the real threats? I mean, Iran does have drones. They do have boats.

SPEAKER_01 5:36

We saw a mix of this last week. There were reports of actual kinetic activity, minor skirmishes involving drones, but the AI propaganda amplified it into a major defeat for the U.S. Navy. The danger is that if a missile actually does hit a ship tomorrow, half the world won't believe it because they've been inoculated by the fakes. Or conversely, the U.S. government could dismiss a real loss as enemy AI propaganda. It creates a fog of war that is incredibly dense.

SPEAKER_04 6:10

And that fog isn't staying overseas. I want to pivot to the domestic front, because while Trump is focusing on Iran, we are seeing a very different, perhaps subtler use of this technology right here in the U.S. midterms. And interestingly, it's coming from the left. You've been tracking a fascinating story out of Chicago involving Jesse Jackson Jr. and Bobby Rush. Break that down for us.

SPEAKER_01 6:37

This is one of the most ethically complex stories of the 2026 cycle. So Jesse Jackson Jr. is running to reclaim his old congressional seat in Illinois. He released a TV ad recently that features an endorsement from former Congressman Bobby Rush. Now Bobby Rush is a real person. He really supports Jackson, but there's a catch. Rush has health issues that have severely affected his voice. He can't speak clearly enough for a broadcast commercial.

SPEAKER_04 7:04

So they used AI to fix it?

SPEAKER_01 7:06

They did more than fix it. They reconstructed it. The campaign used an AI voice clone trained on Rush's old speeches to deliver the endorsement. To be clear, Rush wrote the words and approved the ad. It is his intent. But the voice listeners heard on TV was synthetic. The ad includes a disclaimer, but it raises a massive question. Is it disinformation if the speaker consents? Critics are calling it deceptive because it presents a version of reality, a healthy, vigorous Bobby Rush that doesn't currently exist.

SPEAKER_04 7:39

That feels different than the Iranian boat fakes, though. One is a lie about an event, the other is what, a prosthetic voice?

SPEAKER_01 7:48

A digital prosthetic is a generous way to put it, and that's exactly the argument the Super PAC supporting Jackson Leading the Future is making. They argue this is accessibility technology. But look at the opposition. A rival group called Jobs in Democracy has attacked the ad, arguing that once you start normalizing synthetic voices in politics, you open the door to less scrupulous actors. If we accept a fake Bobby Rush voice because he consented, what happens when a deep fake of a candidate saying something racist drops two days before an election and the creator claims it was satire or art?

SPEAKER_03 8:24

We actually saw a glimpse of that dark side recently in Ireland, didn't we? The prompt mentioned incidents in Ireland and Slovakia. How do those connect to what we're seeing in Chicago?

SPEAKER_01 8:36

They serve as the warning sign. In Slovakia, just days before the election, a synthetic audio recording was released that sounded like the leader of the Progressive Slovakia Party discussing how to rig the vote. It was completely fake, likely Russian origin, but it spread during a media blackout period where it couldn't be easily debunked by the news. And in Ireland, we saw a deep fake of a left-wing candidate, Catherine Connolly, allegedly withdrawing from the race. These weren't prosthetics to help a sick man speak, they were suppression tactics designed to confuse voters and to press turnout.

SPEAKER_04 9:12

And the polling suggests Americans are terrified of this. That PBS NPR poll from earlier this month showed 85% of people are worried about AI in the election. But here's the irony, Alex. Democrats are statistically more worried about it, 86% versus 81% of Republicans, yet we're seeing Democratic campaigns like Jackson's being the early adopters of the tech. Is there a disconnect there?

SPEAKER_01 9:41

It's a massive disconnect, and it speaks to the tactical dilemma the left faces. There's a high degree of anxiety among Democratic voters about misinformation, partly a hangover from 2016 and 2020, but simultaneously, the political consulting class sees AI as a cost-saving miracle. Why pay a sound engineer $5,000 when an AI can master the audio for $50? Why fly a candidate to a chute when you can generate the B-roll? We're seeing a split where the voters on the left view AI as a threat to democracy, while the operatives on the left view it as a necessary tool to keep up. That tension is going to define November.

SPEAKER_04 10:20

I want to go back to the psychology of this. You mentioned earlier that debunking these things is hard. There's this concept of the continued influence effect. Can you explain how that applies here? Because even after the Pentagon said the carrier isn't burning, and even after the fact-checkers labeled the Slovakia tapes as fake, people still believe them.

SPEAKER_01 10:41

The continued influence effect is the psychological phenomenon where misinformation continues to shape a person's reasoning even after they have acknowledged the correction. It's not just stubbornness, it's cognitive architecture. When you see that video of the burning ship, your brain builds a mental model. The war is going badly, or the U.S. Navy is vulnerable. Later, when you hear that video is fake, you might intellectually accept the correction, but you don't necessarily tear down the mental model. You still feel the vulnerability. You still have that lingering sense of dread. The image bypasses your logic centers and hits your amygdala, your fear center. You can't fact-check a feeling.

SPEAKER_04 11:24

So the damage is done the moment we click play.

SPEAKER_01 11:26

Essentially. And that's what Iran is banking on. They don't need to convince the U.S. military that the ship sank. They know the Pentagon has satellites. They need to convince the average citizen in Tehran or Beijing or even Chicago that American power is an illusion. And in that sense, Trump's critique of the fake news media creates a chaotic feedback loop. If he tells his followers that the media is treasonous and lying, and then the media reports that an Iranian video is fake, well, a segment of the population might believe the Iranian video just because the media debunked it. It creates a post-truth environment where the most emotionally resonant video wins.

SPEAKER_04 12:05

It feels like we are entering a period where seeing is no longer believing, but believing is seeing. You believe what confirms your bias. So, Alex, as we look toward November and as this conflict in the Middle East continues, what is the antidote? Is it regulation? Is it better AI detection tools?

SPEAKER_01 12:27

Regulation is too slow. The FCC chairman, Brendan Carr, is looking at broadcast licenses, as Trump noted, but that's a 20th-century solution to a 21st century problem. You can't revoke the license of a Telegram channel. The technical solution, watermarking AI content, is helpful, but bad actors will just strip the watermarks. The real antidote, and it's an unsatisfying one, is skepticism. We have to move from a default trust society to a default verify society. When you see a shocking video, whether it's a burning ship or a politician confessing to a crime, pause. Ask what who benefits from me believing this? If we can't train our algorithms to catch the fakes, we have to train our brains.

SPEAKER_04 13:13

Default verify. It's a cynical way to live, but it might be the only way to survive the next election cycle. Alex, thank you for walking us through the smoke and mirrors today.

SPEAKER_01 13:25

My pleasure, Chloe.

SPEAKER_04 13:26

And to you listening, if this conversation made you rethink something you saw on your timeline today, do me a favor. Share this episode with a friend who needs a reality check. We'll see you next time.