Deepfake Attack: Cloned Voice Call Drains €95M From Bank

Picture a phone call from someone who sounds exactly like your spouse, your child, or your bank manager. They say it's urgent, and they need money moved now. That is a deepfake (a fake voice or video made by AI to pass as a real person), and it just cost one bank €95 million.

Short answer to the question you're really asking at 11pm: you're not doomed, and your savings aren't suddenly up for grabs. But after a deepfake attack like this one, the way banks decide "yes, that's really you" is about to change. You'll feel it as a little friction. Learning to like it is the smart move.

Deepfake Attack at a Bank: What Happened in the Fideuram Deepfake Loss

Here's the story as reported by FinanceX Magazine. Fideuram, a bank, lost about €95 million after criminals used a fake WhatsApp message together with a deepfake voice. A deepfake voice is a computer-made copy of a real person's voice, good enough to fool people who know it well. Trending Topics has a technical breakdown of how the deepfake scam worked.

Not a corner shop. Not a small business with one tired bookkeeper. A bank, which presumably has security rules, training, and people whose job is to be suspicious.

The ending was not entirely grim. The bank spotted the deepfake fraud fast and alerted authorities in several countries. Investigators in China, Portugal, and Italy worked together and got back about €53 million. But roughly €36 million is still missing, moved through overseas accounts and turned into cryptocurrency, where it gets very hard to trace.

Why this Fideuram deepfake loss feels different from earlier deepfake videos

Remember the Arup story? An employee at the British engineering firm moved about $25 million after a video call with colleagues who turned out to be deepfakes. Those deepfake videos made headlines, and then the story faded. The Fideuram deepfake loss is bigger, and it hit a bank instead of an engineering company. When it's a bank, regular people start to wonder about their own accounts. This article is part of a series, start with Deepfake Voice Scam News Today 77 Of Targets Lost Money.

Here's the uncomfortable part. Voices and faces were never built to be security. They were built to be shortcuts. You trust your mom's voice because it is your mom's voice, and for all of human history that was a fine rule. It isn't anymore.

Why Is Deepfake Voice Fraud Growing So Fast, and Is Voice Cloning Easy?

Yes, deepfake voice cloning is easy now, and that's the whole problem. According to the research behind this story, cheap off-the-shelf tools can copy a voice from just seconds of public audio, then read any script in it. Short videos you posted, a voicemail greeting, a clip from a family birthday on social media: any of it can be enough raw material.

The numbers back this up. A Pindrop security report found that deepfake fraud attempts rose by more than 1,300% in 2024. That's a jump from about one a month to about seven a day.

Those figures come from the research compiled for this story, and the Infosec Federation piece covers the banking side in more detail. The pattern is the same everywhere. Faking a person with deepfake technology got cheap, while checking a person stayed expensive.

Is a deepfake voice really good enough to fool a bank?

Sometimes, yes. Some banks use "voice verification" (software that decides it's really you because of how you sound). The research says weak versions of those systems can be passed by a deepfake voice. Even when detection software holds up, a human on the other end, under pressure, often doesn't. Urgency is the con artist's best tool. The deepfake just makes the urgency believable.

Why Deepfake Detection Alone Won't Protect Your Bank: The Prediction

Here's my bold call. Within a couple of years, "I recognize your voice" will stop counting as a reason to approve a big transfer at most banks, because a deepfake voice can now pass for yours. Your voice will be treated like a username: a claim, not a proof. The expert view in the research puts it plainly:

The solution is not perfect detection; it's radical friction. Previously in this series: Roblox Face Scan Predators Rarely Need To Fool The Camera.

Expert analysis in the research brief for this story

"Radical friction" is a fancy way to say speed bumps. Expect call-backs (the bank hangs up and phones you on a number it already has on file). Expect in-app confirmations, where you tap "yes, that was me" on your phone. Expect a waiting period before a very large transfer clears. For the biggest moves, expect more than one person at the bank to sign off.

Will the extra deepfake checks make banking annoying?

A little, yes. Security teams worry about frustrated customers, and that's fair. But look at what actually saved money at Fideuram. It wasn't a slow process. It was quick action once something looked wrong, plus rules that didn't hang everything on one person's gut feeling under pressure. The point of friction isn't to be slow. It's to make sure no single hurried decision can empty an account.

So when your bank asks you to confirm a transfer a second time, nobody's insulting you. They're making sure a deepfake caller wearing your voice can't spend your savings.

How Can You Detect Deepfakes Before You Send Money?

You usually can't spot a deepfake by ear anymore, so don't try. Instead, change the process. Hang up and call back using a number you already trust, such as the one on your card or in your contacts. Never use a number the caller gives you. Agree on a family code word now, before anything happens. Treat urgency plus secrecy plus money as a red flag all by itself.

One more gentle thought. If you've ever wondered whether a photo, a profile, or a face on a video call is really who it claims to be, that's exactly the identity question that facial comparison and deepfake detection tools exist to answer. The same instinct applies to your own life. A face or a voice is a clue, not a verdict.

Why does this scare us so much? It's the availability heuristic: a vivid story sticks in your mind, so it feels more likely. A €95 million loss at a real bank is a vivid story. Let it nudge you into one small habit, a pause before any money moves, rather than into panic.

Here's the question the €95 million call leaves hanging. If a bank full of trained professionals couldn't tell a deepfake voice from the real thing, what exactly is your plan when that voice says "Mom, it's me"? Pick your code word tonight, because the next deepfake caller won't give you time to think.

Deepfake Detection: Frequently Asked Questions