Skip to playerSkip to main content
Welcome to Verify News Global. In this deep dive, we expose one of the most sophisticated financial heists in history. We break down the terrifying $25 Million corporate scam where an employee was tricked into a fake video call featuring a deepfake of their own CFO.

This video is a wake-up call. We explore how Artificial Intelligence is being weaponized to steal identities, clone voices, and manipulate reality.

In this video, we cover:
The $25M Ghost Call: How a UK-based firm lost millions to a deepfake meeting.

The Rise of Deepfakes: Why digital deception is becoming harder to detect.

The Liar’s Dividend: How scammers use AI to make you doubt the truth.

Stay Safe: Practical steps to verify information and protect your assets.

Important Disclaimer:
This video is for educational purposes only. The goal of Verify News Global is to promote digital literacy and cyber-security awareness. Always verify sensitive information through multiple reliable online and offline sources before taking any action.

Connect with Us:
Channel: Verify News Global

Topics: AI Scams, Deepfake Technology, Cyber Security, Digital Truth.

If you found this video helpful, please SUBSCRIBE and hit the bell icon to stay protected in the digital age!

#VerifyNewsGlobal #AIScams #Deepfakes #CyberSecurity #DigitalLiteracy #ArtificialIntelligence #DeepfakeWarning #TechNews2026

Category

🗞
News
Transcript
00:07Imagine you're just, you know, sitting at your desk.
00:09You're logged into a totally routine video call.
00:12Right.
00:13Just a normal Tuesday.
00:14Exactly.
00:15And on the screen is your company's chief financial officer.
00:19He looks a little stressed.
00:20He's got that familiar kind of raspy cadings to his voice.
00:23And he's asking you to authorize an urgent $25 million transfer to secure a sudden acquisition.
00:31Which obviously is a massive request.
00:33Huge.
00:34But everything about the interaction feels completely normal.
00:37So you click approve.
00:39But here's the thing.
00:40Your CFO was never actually on the call.
00:42You just wired $25 million to a ghost.
00:45Yeah.
00:45It's a completely staggering scenario.
00:47And I think the most chilling aspect of it is that it's not hypothetical at all.
00:51No, it's really not.
00:52Welcome to this deep dive.
00:54Today, our mission is navigating a brand new reality for you, the listener.
00:58One where you literally cannot believe your own eyes and ears anymore.
01:01Right.
01:02Because we are looking at a landscape where the fundamental building blocks of objective,
01:07truth-like audio, video, photographic evidence, they are being compromised and weaponized at an industrial scale.
01:13It's wild.
01:14So to help us map out this minefield, we're drawing on two incredibly eye-opening sources today.
01:19First, we have an industry report from editor and publisher by Rob Tornow.
01:24Which takes a really hard look at how this tech is just eroding media trust across the board.
01:29Yeah, exactly.
01:29And second, we have a stark corporate risk analysis published on LinkedIn by Prashant Doom.
01:35That one really details the severe financial and operational threats companies are facing.
01:41So, okay, let's unpack this because we really need to understand the sheer scale of what we are dealing with
01:47here.
01:47Well, the scale is the entire point, really.
01:49We have to move past this completely outdated notion that deep fakes are just, you know, niche internet jokes or
01:54people making silly mashup videos in their basements.
01:57Right.
01:57Like the early YouTube days.
01:59Exactly.
01:59That era is over.
02:01Generative media is now an active, highly sophisticated battleground.
02:05And the targets are your wallet, your media consumption, and basically your baseline trust in reality itself.
02:11I think to really grasp the threat, we have to look at how absurdly fast the mechanics of this technology
02:17have evolved.
02:18I mean, in the editor and publisher piece, Tornow points out that just a few years ago, the absolute pinnacle
02:24of AI video was that viral, honestly laughable clip of Will Smith eating spaghetti.
02:30Oh, my gosh.
02:31Yes.
02:31The spaghetti video.
02:33It looked like a fever dream.
02:34Like the physics were entirely wrong.
02:35The pasta was clipping right through his face.
02:37His jaw was morphing into the bowl.
02:39It was obviously a joke.
02:40Total meme.
02:41Right.
02:41But you fast forward to today, and we are looking at the hyper-realistic real-time deception from Doom's corporate
02:48risk analysis.
02:49Let's look at that $25 million heist at a UK architecture firm.
02:53Yeah.
02:54The terrifying part of that specific scam wasn't merely that they cloned the CFO's voice.
02:58I mean, voice cloning is bad enough.
03:00But it's that the attackers likely used real-time face swapping software.
03:04Wait, real-time.
03:05So someone was actually there on the camera.
03:07Yes.
03:07An actual human being was sitting on the other end of that video call, moving their head and moving their
03:13lips.
03:13But the AI was tracking their facial landmarks.
03:16So the eyes, the nose, the jawline.
03:19And it was literally mapping the CFO's face onto theirs in real time.
03:24That is insane.
03:26It operates very similarly to a highly advanced, weaponized social media filter.
03:31And while that's happening visually, a secondary audio processing tool is simultaneously altering the pitch and the timbre of the
03:37scammer's voice to match the CFO's exact acoustic fingerprint.
03:41Wow.
03:42It makes me think about the evolution of CGI in movies, you know?
03:45Yeah.
03:45Like 20 years ago, you went to the theater and you knew exactly when the monster was coming.
03:49Oh, absolutely.
03:50The lighting was off.
03:51The rendering was clunky.
03:52Exactly.
03:52The shadows were wrong.
03:53Your brain just inherently knew it was a special effect.
03:56But today, the technology is so seamless, filmmakers use CGI for things you don't even notice.
04:02Like changing the weather in a scene.
04:03Changing the weather, extending a city skyline, or entirely reconstructing a historical room.
04:09You have literally no idea you're looking at special effects.
04:12Yeah.
04:12And that level of invisible manipulation is exactly what is happening to our daily communication now.
04:17It's a great analogy.
04:19But if the tech is moving this fast, I mean, aren't we just completely outmatched?
04:23Why is this suddenly absolutely everywhere?
04:26Well, the short answer is that the protective guardrails are quietly being dismantled by the industry itself.
04:32Really?
04:32Yeah.
04:32In the early days of generative AI, the major players' companies like Google and OpenAI, they placed strict, hard-coded
04:41limitations on their tools.
04:43To keep people from making fake news.
04:45Exactly.
04:45They explicitly prohibited users from prompting the AI to create videos of public figures or copyrighted materials precisely to avoid
04:53this kind of chaos.
04:54But I'm guessing the market pressure to dominate the space kind of changed that calculus.
04:58Significantly.
05:00Torno points out that when OpenAI launched their new social video app, Sora, the initial architecture basically allowed users to
05:08generate content using celebrities and protected intellectual property with minimal friction.
05:14Oh, wow.
05:14So they just took the brakes off.
05:16Pretty much.
05:16The compute power was essentially handed over without the traditional filters.
05:20It resulted in such a wild free-for-all that Sam Altman actually had to scramble to rewrite those constraints.
05:26Probably to stave off a massive tsunami of copyright lawsuits, I'd imagine.
05:31Primarily, yes.
05:32But the underlying capability is now out in the wild.
05:35And because those guardrails vanished, the barrier to entry just dropped to zero.
05:40You no longer need a Hollywood budget or a massive server farm.
05:43You literally just need a smartphone.
05:46Just a phone and an internet connection.
05:47Right.
05:48Which brings us to the inevitable result of handing over studio-level special effects to the entire internet.
05:53We get what the sources call AI slop.
05:56AI slop.
05:57It's such a perfect term for it.
05:58It really is.
05:59Torno quotes Vox's Brian Walsh, who had this incredibly vivid, honestly visceral description of it.
06:05He compared taking a TikTok-style social media platform and mixing it with AI-generated content to taking heroin and
06:13mixing it with heroin.
06:14Wow.
06:14It is a really intense comparison, but neurologically speaking, it actually makes perfect sense.
06:19How so?
06:19Well, social media recommendation algorithms are already designed to be highly addictive, right?
06:24They surface high-engagement, emotionally resonant content to keep you scrolling.
06:29Right.
06:29The infinite scroll.
06:30Exactly.
06:31So when you flood that system with generative AI, you are injecting an infinite brain-rotting stream of garbage content
06:40that can be manufactured at zero cost.
06:43It perfectly feeds the algorithm's demand for constant novelty.
06:46And a massive portion of that novelty content is deeply political.
06:50Now, before we get into this, I need to be really clear with you, the listener.
06:53Our sources highlight examples from across the political spectrum.
06:56We are simply reporting these case studies to show how the tech is being used without endorsing any political viewpoints
07:01whatsoever.
07:02Yes, that is a crucial distinction.
07:04We're looking at the mechanics here.
07:05Exactly.
07:05Because what's really striking in Torno's report is that this isn't isolated to one political ideology.
07:12Looking at the case studies, whether it's Andrew Cuomo using AI to depict himself as a subway conductor in an
07:18ad with a tiny, barely visible disclaimer, by the way.
07:22Oh, very tiny.
07:23Or Eric Adams utilizing AI imagery before suspending his campaign.
07:28Or Donald Trump sharing an AI-altered video putting a sombrero on Representative Hakeem Jeffries.
07:33Or Elon Musk sharing a manipulated campaign ad featuring Kamala Harris.
07:38We're seeing this technology deployed across the entire spectrum.
07:41They're all basically beta testing the exact same tactic.
07:44Right.
07:45The focus for us here isn't the politics.
07:47It is purely the mechanics of the deception.
07:48What's fascinating here is the sheer variety of the testing.
07:51Some political actors are using it for campy online humor to drive engagement metrics.
07:56You know, just trying to go viral.
07:57Yeah, the meme wars.
07:58Exactly.
07:59But others are trying to artificially manufacture a relatable blue-collar aesthetic without actually, you know, staging a real photo
08:07shoot.
08:07But every single instance contributes to the exact same underlying problem, which is the normalization of fabricated reality in our
08:16civic discourse.
08:17And it is not just playing out at the national presidential level where you have teams of forensic digital analysts
08:25scrutinizing every single pixel.
08:26The accessibility of this tech means it's hitting local communities, which is honestly arguably much scarier.
08:32Oh, without a doubt.
08:33The local level is highly vulnerable.
08:35Yeah.
08:36Torno cites an example from a suburban Baltimore school in 2024.
08:39The school principal was suddenly placed on administrative leave because an audio recording went incredibly viral.
08:45In the clip, the principal was allegedly using just horrific racist and anti-Semitic slurs.
08:50I remember reading about this.
08:51The damage to his reputation and the resulting community outrage was instantaneous.
08:56Overnight.
08:56But an investigation eventually revealed the audio was a complete fabrication.
09:01It was AI-generated.
09:02And the architect behind it wasn't some sophisticated foreign hacker collective.
09:07It was the school's own athletic director.
09:10Which is just crazy.
09:11He created the fake as revenge because he was under investigation by that very principal for allegedly embezzling school funds.
09:19It perfectly illustrates how you don't need a massive supercomputer to destroy someone's life or career anymore.
09:26Just a grudge, an internet connection, and like five minutes of audio training data.
09:31That's really all it takes now.
09:32But here is the twist that I found genuinely mind-bending in the sources.
09:36Because fake audio and fake video are becoming so ubiquitous, there is a secondary psychological effect taking hold.
09:43And it might actually be more dangerous than the fakes themselves.
09:46Ah, you were referring to the liar's dividend.
09:48Yes.
09:49The liar's dividend.
09:50Torno shares a personal anecdote that perfectly captures this feeling.
09:54A friend sends him a video of former NFL head coach Jim Mora.
09:57It's an old clip.
09:58It's grainy, low resolution.
10:00And Mora is just passionately cursing out reporters.
10:03Standard old school sports footage.
10:05Yeah, it's wild footage.
10:07But Torno, a man who literally analyzes digital media for a living, looks at it and genuinely cannot tell if
10:14it is real.
10:15He has this headache-inducing moment of profound doubt.
10:19The mechanism causing that doubt is so fascinating.
10:22The low resolution of older video actually acts as a form of camouflage for AI.
10:27Really?
10:28How does that work?
10:28Well, the natural visual artifacts of old VHS tapes or early digital cameras like the blur, the pixelation, the dropped
10:35frames, they perfectly mimic the rendering flaws and glitches of early generative AI.
10:40Oh, that makes sense.
10:41Right.
10:42So the human eye struggles to differentiate between analog decay from an old tape and digital fabrication from an AI
10:48model.
10:48Wow.
10:49Now, Torno eventually verified that the Mora video was completely real by digging into old newspaper archives.
10:55But the fact that he doubted it at all is the core issue here.
10:58And this brings us to that crucial study published in the Journal of Politics by Professors Burari, Lucas, and Munger.
11:03Yes, this study is the absolute key to understanding the true societal impact of this technology.
11:09What they discovered is that deepfakes, on their own, might not actually deceive the broader public on a massive civilization
11:16-ending scale.
11:17Because eventually someone figures it out.
11:19Exactly.
11:20Most people eventually figure out the fake, or journalists rigorously debunk it.
11:25But the sheer existence of so much fake content achieves something far more insidious.
11:31It systematically discredits authentic media.
11:34It's like tossing a single, counterfeit $100 bill into a busy cash register.
11:39The real danger isn't just that someone successfully spends the fake bill.
11:42The danger is that now the cashier doesn't trust any of the real bills in the drawer.
11:47That is a brilliant way to put it.
11:49They are questioning the authentic currency.
11:51Right.
11:51And the studies show this exact psychological mechanism across the board.
11:55They show Republicans a very real, completely authentic video of Donald Trump messing up Apple CEO Tim Cook's name.
12:02And the Republicans were highly likely to dismiss it as a deepfake.
12:05And the researchers observed the exact same cognitive dissonance playing out on the other side.
12:10They showed Democrats a real, authentic video of Barack Obama appearing to make a post-election deal with Russian President
12:18Vladimir Putin.
12:19And the Democrats overwhelmingly dismissed the real footage as an AI fake.
12:25So wait, let me make sure I'm locking on to the underlying logic here.
12:28The primary danger isn't necessarily that we're all going to fall for the fake video.
12:33The real danger is that we use the existence of fake videos as an incredibly convenient psychological escape hatch to
12:41just ignore the truth whenever it challenges our worldview or makes our side look bad.
12:46That is the liar's dividend in its purest form.
12:48When anything can be faked, nothing has to be believed.
12:51Wow.
12:52People are just offloading their cognitive dissonance onto the technology.
12:55If a corporate executive is caught on tape admitting to a crime, their immediate legal and public relations defense will
13:01simply be, well, that is an AI voice clone.
13:03Right. You can't prove it's me.
13:05Exactly.
13:05If a politician is filmed doing something corrupt, they just claim it's a deepfake.
13:09And because the public knows deepfakes exist, a large portion of the population will readily give them the benefit of
13:15the doubt.
13:16It actively erodes the very concept of shared objective evidence.
13:21Which is terrifying, because if we can't agree on basic recorded facts, the entire foundation of how we function as
13:28a society starts to crumble, which leads to the obvious next question for anyone listening.
13:34How do we fight back?
13:35I mean, why can't we just build an AI to catch the AI?
13:39We have antivirus software for our computers.
13:42Where is the algorithmic antivirus for our eyeballs?
13:45It is a very logical assumption, but the reality of the engineering is much more complicated than that.
13:50Researchers at Cornell University are calling this a new battleground, and it really comes down to how these AI models
13:55are built in the first place.
13:56Okay, lay it on me.
13:57They often use what are called generative adversarial networks, or JANs.
14:00Right, JANs.
14:01Essentially, one part of the system generates the fake, and another part of the system tries to detect it.
14:06And they learn from each other in real time.
14:08So by the time you train a detection algorithm to catch a specific type of AI artifact,
14:14the generative AI has already used that detection data to smooth the artifact out in the next version.
14:19Man, so it's not just an endless game of whack-a-mole.
14:22The mole is actively learning how to dodge the mallet.
14:25Precisely.
14:26And relying heavily on software can actually introduce new vulnerabilities.
14:30Experts at the Toe Center for Digital Journalism warn about something called automation bias.
14:35Automation bias.
14:36What's that?
14:36It's where detection tools create a false sense of security.
14:39If a glitchy detection tool flags a fake video as real, journalists and the public might just blindly trust the
14:46software's output and amplify the misinformation.
14:48It effectively weakens our own critical thinking muscles because we're outsourcing our skepticism to a machine.
14:54Though, to give credit where it's due, Tornow did find one tool that actually worked quite well in his testing.
14:59It's called the Hiya DeepFig Voice Detector.
15:02It's a Chrome plug-in, and it analyzes audio in real time.
15:05But I'm curious, how does it actually know the difference if human ears can't tell?
15:09It's analyzing the acoustic artifacts that fall completely outside of normal human perception.
15:15Like stuff we literally can't hear.
15:17When an AI generates speech, it often struggles with the subtle physics of human anatomy.
15:22The micropauses for breath, the natural resonance of a vocal cord, the specific frequency distribution of a consonant, the plug
15:31-in measures those anomalies and assigns an authenticity score from 1 to 100 based on those physical impossibilities.
15:38And it seems highly effective, at least for now.
15:40Tornow tested it on 10 real videos and 10 deepfakes, and it scored a perfect 20 out of 20.
15:45That's impressive.
15:47Yeah.
15:47It even caught that AI-altered Kamala Harris ad that Elon Musk shared, giving it a 1 out of 100
15:53and flagging it as a likely deepfake.
15:55But as you said, the tech will adapt, so if software can't permanently save us, we have to patch human
16:00behavior.
16:01Doom's piece had a brilliant suggestion for a human defense playbook, starting with pre-agreed code words.
16:07It sounds almost archaic, like a digital speakeasy.
16:11But if your CFO calls asking for $25 million, you just ask for the password.
16:15It is archaic, and that's exactly why it works so beautifully.
16:19It breaks the reliance on the screen.
16:21If you have elderly parents or if you work in finance and regularly authorize transfers, establish a specific word or
16:28phrase offline.
16:29Right. Don't text it to them.
16:30No. Do it in person.
16:32It is a remarkably simple, zero-tech defense that completely shatters the AI impersonation.
16:38You should also really watch out for timing manipulations.
16:41What do you mean by timing manipulations?
16:42Bad actors know that chaos is their best friend, so you need to be hyper-alert if a scandalous video
16:48or a dire financial request drops right before a major event, like the night before an election, or an hour
16:54before a company's quarterly earnings report.
16:56The goal is to force you to react before the truth can put its boots on.
17:00And a massive part of forcing that reaction is weaponizing our feelings.
17:05Bad actors are exploiting our amygdala.
17:07They know that if they can trigger a high arousal, emotion outrage, terror, deep empathy, sudden urgency, they effectively bypass
17:16the brain's critical thinking centers.
17:18That is the most crucial vulnerability we have as humans.
17:22If a video makes you instantly furious or instantly terrified, you must recognize that emotional spike as a giant red
17:28flag.
17:28Take a breath.
17:29Yes.
17:29That physiological reaction is exactly what the creator is relying on to bypass your rational processing.
17:35To combat that, Torno brings up digital literacy expert Mike Caulfield's SIFT method.
17:40It's an acronym that stands for STOP.
17:42Investigate the source, find better coverage, and trace the claim to its original context.
17:47It's not just about digital hygiene, it's a cognitive circuit breaker.
17:51I like that, a cognitive circuit breaker.
17:53Yeah.
17:53If you see a crazy video on social media, don't just hit share.
17:56Stop.
17:56Look to see if established, rigorous news outlets are reporting on it.
18:00Try to find the uncut, original version.
18:02Slowing down our processing speed is our absolute best defense against the generative speed of AI.
18:08If we connect this to the bigger picture, that intentional friction forcing yourself to manually slow down your consumption
18:14is the absolute antidote to the viral nature of deepfakes.
18:18Okay.
18:19We've covered the multi-million dollar heists, the algorithmic slop, the erosion of objective truth.
18:25It's incredibly heavy stuff.
18:26It really is.
18:27But here's where it gets really interesting.
18:30Because both of our sources acknowledge that this technology is just a tool.
18:35It isn't inherently evil.
18:37There is a silver lining here.
18:39There are constructive applications that are actually quite beautiful.
18:42Yes.
18:43We have to look at the full spectrum of the technology.
18:46Joom's analysis highlights how, with proper ethical frameworks and transparency,
18:50deepfakes are going to revolutionize several fields.
18:53Take accessibility, for instance.
18:55This was the part of the deep dive that actually gave me some hope.
18:57It is remarkable.
18:58The exact same mechanism that can clone a CEO's voice to steal money can be used to give a voice
19:03back to someone who is actively losing theirs.
19:06Individuals suffering from degenerative diseases, like ALS, can utilize something called voice banking.
19:12They record a specific phonetic script while they still have full vocal control.
19:17The AI maps the unique timbre, pitch, and cadence.
19:21Then, as the disease progresses and they lose the physical ability to speak,
19:25the AI can generate real-time speech from whatever they type.
19:29That is amazing.
19:30And the output isn't a robotic, generic, text-to-speech voice like Siri.
19:34It is their authentic, acoustic identity.
19:37It restores a profound piece of their humanity.
19:40That is incredible.
19:41And the sources also mention real-time translation into sign language,
19:44which obviously opens up massive communication avenues.
19:47What about the educational and entertainment applications?
19:50Well, imagine bringing history to life.
19:52Instead of reading a dry textbook,
19:53students could interact with an AI-generated, historically accurate avatar of Abraham Lincoln or Marie Curie,
19:59answering questions in the classroom in real time.
20:02Kids would love that.
20:03And in the entertainment and business sectors,
20:05we are already seeing actors being seamlessly de-aged for flashback scenes,
20:09which reduces massive physical reshoot costs.
20:13Furthermore, instead of poorly dubbed movies where the audio clearly doesn't match the mouth,
20:18AI can sync an actor's lip movements perfectly to a foreign language track.
20:22Oh, creating seamless global localization for films and marketing.
20:27Exactly.
20:27So Tom Cruise can literally look like he is speaking perfectly fluent Mandarin in the next Mission Impossible,
20:34expanding that global connection.
20:36The potential is vast.
20:38Doom quotes the entrepreneur and futurist Peter Diamandis, who states,
20:42The future of this technology remains unwritten.
20:45But the absolute non-negotiable requirement to unlocking these benefits, however,
20:49is ethical consent and complete transparency.
20:52Right.
20:52We have to know without a shadow of a doubt when we are talking to a machine.
20:55Exactly.
20:56The technology itself is completely agnostic.
20:58It's the human intent behind it that determines whether it operates as a miracle or a menace.
21:03So what does this all mean?
21:05We have journeyed from the goofy low stakes days of AI spaghetti videos right into the heart
21:11of a technological revolution that's actively targeting our wallets,
21:15our political systems, and our very perception of reality.
21:19It's been a rapid shift.
21:20The core takeaway for you today is this.
21:23We have crossed a fundamental threshold.
21:25You can no longer take what you see or hear on a screen at face value.
21:29The era of implicit trust in digital media is over.
21:33Sadly, yes.
21:34But you don't have to exist in a state of constant paranoia.
21:37You just have to be prepared.
21:39By adopting a mindset of pause, question, verify, using the SIFT method as a cognitive circuit breaker,
21:46recognizing when your emotions are being hijacked,
21:48and setting up simple analog defenses like code words with your family and your coworkers,
21:53you do not have to be a victim of the next $25 million scam
21:56or fall for the latest piece of viral political slop.
21:59You can navigate this new reality.
22:01It's really about adapting our human instincts to survive in a synthetic world.
22:05But, you know, this conversation raises an important question,
22:08one that builds on what we just discussed about preserving voices for ALS patients.
22:12Okay, what's that?
22:13If artificial intelligence can perfectly mimic a person's voice, their face,
22:17their exact mannerisms, and their conversational style,
22:20what happens to human grieving in the future?
22:23Oh, wow.
22:23When we pass away, will it become a standard cultural practice
22:27to leave behind interactive, deepfake versions of ourselves
22:30for our grandchildren to talk to, to seek advice from?
22:33And if we do integrate that into our life,
22:36are we beautifully preserving a legacy,
22:38or are we just preventing our loved ones from ever truly letting go?
22:43Oh, man, that is a deeply profound place to leave it.
22:46Next time you get a strange voice note from a friend
22:49or an odd video call from your boss and you do that double take,
22:52remember the mechanics of everything we unpacked today.
22:55Thank you so much for joining us on this deep dive.
22:57Stay curious, stay skeptical, and we'll catch you next time.
Comments

Recommended