Deepfake: Cloned Voice Call Drains €95 Million From Bank

A bank lost ninety-five million euros because someone cloned a voice and sent a fake message on WhatsApp. Investigators say the money moved through accounts in several countries and some of it turned into cryptocurrency. About thirty-six million euros is still missing.

If you've ever answered a call from a number you

If you've ever answered a call from a number you didn't know, this story is about you. Most of us assume a voice we recognize is proof of who's talking. That assumption is what this theft exploited. The bank is Fideuram, a major Italian private bank. According to reports, the attackers pretended to be someone with authority, and they used a cloned voice to sell the lie. Fideuram spotted the irregularities quickly and alerted authorities in several countries. Investigators in China, Portugal, and Italy worked together and got about fifty-three million euros back. So if a voice can no longer prove who you are, what does?

Start with how little the attackers needed. According to the analysis from FinanceX Magazine, off-the-shelf tools can clone a voice from just seconds of public audio. Then they can read any script in that voice. Think about your own voice. A birthday video, a voicemail greeting, a clip on social media. That's already raw material.

For decades, banks built their checks on a simple assumption. Copying a whole person, voice and face and manner, would be too expensive to bother with. That assumption has flipped. Now it's cheap to fake an identity, and expensive for the bank to prove one is real. The article puts it plainly. Voice and video were never security layers. They were trust shortcuts. For a bank, that means old call-center scripts don't hold up. For you, it means a familiar voice on the phone is no longer a guarantee of anything.

The numbers show how fast this moved

The numbers show how fast this moved. According to the voice security company Pindrop, deepfake fraud attempts rose by more than thirteen hundred percent in 2024. That's from about one attempt a month to about seven every day. Pindrop also reports that synthetic voice attacks on insurance companies jumped nearly five-fold. Banks saw a rise of roughly one and a half times. Contact centers now flag about one in every seven hundred fifty calls as possibly fraudulent. That's a call you might have been on the other end of.

Fideuram isn't the first case, either. In 2024, an employee at the British engineering firm Arup sent about twenty-five million dollars after a video call. Every colleague on that call was fake. Voice and face both lied, and the employee did what any of us would do with a boss on screen. Deloitte's Center for Financial Services projects that losses from this kind of A.I. fraud could reach forty billion dollars in the U.S. by 2027. That's up from about twelve billion in 2023. For banks, that's a budget problem. For families, it's the call that sounds exactly like your son asking for help.

So what's the fix? The article doesn't point to better detection software. It points to deliberate friction. That means call-backs on a number the bank already has on file. It means two people approving any large transfer. And it means re-verifying identity before big money moves. Security teams push back that extra steps slow business and annoy customers.