AI Deepfake Scam News Today: Fake Video Call Costs Bank €95M

Picture this: your boss calls you on video, sounds exactly like himself, and says he needs you to move money right now because a deal is about to fall apart. You'd probably do it too. That's basically what happened at Fideuram, a wealth management arm of one of Italy's biggest banks, and it reportedly cost €95 million. This is the kind of ai deepfake scam news today that should make anyone who's ever wired money at work stop and think twice.

The target keyword ai deepfake scam news today matters right now because a fake video call, built with cloned voice tech, reportedly moved €95 million out of an Italian bank before anyone caught on, and the case shows a familiar face and voice are no longer proof of anything.

Here's how it reportedly went down, according to Trending Topics. Back in February, Fideuram's chairman, Paolo Molesini, got a WhatsApp message that looked like it came from Carlo Messina, the CEO of Intesa Sanpaolo (that's the parent bank, basically the boss's boss). The message asked for urgent help with an overseas transaction. Then came a phone call from someone claiming to be a senior partner at a law firm. Except it wasn't him. It was an AI clone of his voice, good enough that nobody on the other end questioned it. That's the whole trick right there: a familiar name, a familiar voice, and a request that feels too urgent to slow down and check.

What This AI Deepfake Scam News Today Actually Tells Us About Trust

So why does this particular story matter more than the dozen other deepfake headlines floating around this week? Scale, mostly. Ninety-five million euros isn't a rounding error, it's one of the largest deepfake-related frauds reported anywhere so far. But the bigger deal is what it proves: the tools we've used for decades to confirm "yes, this is really my colleague" (a familiar voice, a familiar face on a video call) don't work anymore. Not against synthetic audio built from a minute of sample recording. Not against an AI scam script running in the background while a scammer reads urgent instructions in someone else's voice.

Think about how casually we all still say "I'd know their voice anywhere." That confidence is exactly what makes this AI scam work. Fraudsters don't need to fool a machine. They need to fool a tired, busy person who trusts their own ears. And honestly? Most of us would fail that test at 5pm on a Friday when someone we report to is yelling about a deadline.

AI Deepfake Scam Numbers Behind the Fideuram Case

Roughly €53 million of the €95 million was recovered, according to reporting cited by Trending Topics, thanks to fast cooperation between authorities in Italy, China, and Portugal. The remaining €36 million was moved through overseas accounts and converted into cryptocurrency, and it hasn't been traced. That's not a small detail. It means the money that got saved wasn't saved because someone spotted the fake sooner. It was saved because the bank had a plan for what to do the moment something looked wrong, and moved on it immediately. This article is part of a series, start with Deepfake Ai One Public Photo Is All Blackmailers Need.

This isn't a one-off, either. In 2025, Italian businessman Massimo Moratti reportedly transferred nearly €1 million after fraudsters used an AI voice clone to pose as Italy's Defense Minister, Guido Crosetto. Same country, same playbook, smaller number. That's the part that should worry you more than the headline figure: this isn't a fluke, it's a pattern, and it's repeating.

---

How Video Call Deepfakes and Synthetic Voice Scams Actually Work

Voice cloning used to need hours of recordings and a decent budget. Not anymore. Researchers say cloning software can now replicate someone's voice from as little as 60 seconds of audio, the kind of clip anyone could grab from a podcast interview, a conference talk, or an old video call. A scammer feeds that clip into an AI tool, types out a script, and the software reads it back in that person's tone, accent, and rhythm. No actor needed, no impersonation skill required. Just software doing the heavy lifting while a human reads a script off a screen.

According to data cited by security researchers at Keepnet Labs and Gartner, deepfake attacks that get past biometric checks (that's identity verification using your face, voice, or fingerprint, the stuff that's supposed to be uniquely you) jumped 704% in 2023. Gartner has also predicted that by 2026, 30% of companies will stop trusting standalone identity verification tools on their own, without a second layer of checking. That's a huge shift for an industry that spent the last decade selling "look at the camera and we'll know it's you" as the answer.

Roughly 62 percent of the €95 million was frozen or recovered because Fideuram's systems flagged the transfers and the bank moved fast through international channels, turning a €95 million event into a €36 million one. Previously in this series: Fingerprint Scanner Gel Fingers Fool It 70 Of The Time.

reported analysis via AML Intelligence

Why Detection Alone Won't Save You From a Deepfake Call

Here's the uncomfortable truth: there's no app you can download that reliably spots every fake voice in real time, mid-call, while someone's pressuring you to act fast. Detection tools help, but they're not magic. What actually worked at Fideuram wasn't catching the fake sooner, it was having a rehearsed plan for what to do once something felt off: named contacts, a second approver, escalation paths written down in advance, not improvised while the wire transfer window is closing.

If you take one thing from this whole mess, take this: the availability heuristic (our brain's habit of trusting the most vivid, familiar thing in front of us, in this case a voice we recognize) is exactly what these scams exploit. A real person, asked to pause and verify, will understand. A scammer needs you to act before you think. That single distinction is the whole game.

---

Can a Video Call Really Be a Deepfake Today?

Yes, and this case is proof. A video or voice call can now be built with synthetic audio and, increasingly, synthetic video too, convincing enough that a chairman at a major financial firm didn't catch it in the moment. The old assumption, that seeing and hearing someone live on a call is basically the same as verifying their identity, no longer holds when real money or account access is on the line.

This is where the CaraComp perspective comes in, and we'll say it once, gently: if you've ever wondered whether a call, a video, or a profile is really who it claims to be, that's exactly the question this whole corner of technology exists to answer. One thing you can actually watch for in real footage versus a fake: pay attention to the mouth and jaw during fast speech. In many deepfakes, the skin appears too smooth around the cheeks and jawline compared to the rest of the face, and lighting on the skin doesn't shift naturally when the person turns their head. It's not a perfect tell, and it won't catch a great fake, but it's a free, useful thing to glance at before you act on anything urgent.

Old assumption	New reality after this case
A familiar voice on a call is proof of identity	Voice cloning from 60 seconds of audio can fake that voice convincingly
Detection software catches fakes in real time	Detection helps but is not reliable alone during a live call
Trust the request, act fast to help	Verification through a second approver or callback stops the scam
One executive's approval moves money	Layered bank controls require more than a single call to authorize a transfer

What a Real Verification Call Looks Like Versus a Scam Call

A legitimate request survives a pause. If a real executive, client, or family member asks you to send money urgently, they won't mind you hanging up and calling them back on a number you already have saved, or looping in a second person to confirm. A scam call falls apart under that same pause, because the fraudster's entire plan depends on speed and pressure, not patience.

Some will argue that adding extra verification steps slows everything down and annoys people trying to do legitimate business fast. Fair point, but the Fideuram numbers argue the opposite. It wasn't a slow, cautious approval process that saved €53 million. It was speed of response once red flags appeared, paired with rules that didn't rely on one person's gut feeling under pressure. A second approver, an enforced callback, a transaction limit that trips an alarm, none of these need a human to stay calm and clear-headed in the moment a scammer is counting on panic.

So here's the question worth sitting with: if someone you trust called you on video right now, sounding exactly like themselves, and asked you to send money in the next ten minutes, what's the one thing you'd check before you hit send? If you don't have an answer ready, that's the actual gap this €95 million case just exposed, not in some distant bank's systems, but in the split second any of us might face on an ordinary Tuesday.

AI Deepfake Scam News Today: Frequently Asked Questions