
BlogCorporate Wire Fraud (Voice Clone)
AI Voice Clone and Fake WhatsApp Cost Italian Bank €95M
A spoofed WhatsApp message and an AI-cloned voice were all it took to convince the chairman of Fideuram, one of Italy's largest private banks, to authorize roughly €95 million in wire transfers to accounts in China and Hong Kong. International investigators recovered about €53 million. Around €36 million has not been found.
Paolo Molesini received two things in quick succession: a written message appearing to come from Intesa Sanpaolo CEO Carlo Messina via WhatsApp, and a phone call from someone who sounded exactly like a senior partner at a trusted law firm. Both were fabricated. Together, they were enough.
The fraud unfolded in late September 2026 and stripped Fideuram of approximately €95 million. International cooperation recovered about €53 million. Around €36 million passed through overseas accounts, was converted to cryptocurrency, and has not been traced.
The design was deliberate. A single spoofed message is easier to question. A single voice call, however convincing, invites a second opinion. Running both simultaneously manufactured the appearance of independent corroboration: a written instruction from the CEO, a verbal endorsement from legal counsel. Two sources, two channels, one fraudulent outcome.
How the Channels Reinforced Each Other
Voice cloning now requires as little as 20 to 30 seconds of audio to produce a convincing replica. Earnings calls, investor presentations, and media interviews give fraudsters ample raw material to clone a named executive or professional. WhatsApp spoofing adds a written layer that looks familiar, and each channel made the other more believable.
This exploits a specific cognitive shortcut: corroboration implies legitimacy. When two trusted contacts appear to confirm the same instruction at the same time, the instinct to pause and verify weakens. According to the FBI's Internet Crime Complaint Center, business email and voice compromise schemes produced losses exceeding $2.9 billion in 2023 alone. Attacks on senior executives follow a different logic than standard phishing. Fraudsters do not need to compromise systems. They need to compromise judgment. A chairman authorizing a large international transfer is doing something within his normal authority. The action itself is not unusual. Only the instruction behind it is fraudulent.
The Gap Standard Controls Cannot Close
Most organizations build fraud controls around transaction size and destination. Those controls address the payment, not the authorization behind it, which is where this fraud actually occurred. The attacker did not defeat a technical system. The attacker replaced a person.
A common response to cases like this is to call back on a known number before approving any large transfer. That step is worth taking, but a successful callback only confirms you reached the real person on a second attempt. It does not prove the original caller was legitimate. Verification must happen before approval, not after the funds have moved. If someone appears to be under duress during a high-stakes approval, alert authorities directly.
Voice and written communication are treated as inherently trusted channels. Both can be fabricated. Both were fabricated here.
Trust Onion addresses this with rotating three-word codewords, shared only within a verified chain of personnel and calculated locally on each device with no server dependency. Before any wire authorization, the finance team member asks one question: "What are the words?" The real executive knows them. An impersonator does not. A cloned voice cannot answer.
For a higher-stakes visual check, a Proofie™ is a selfie signed using your passkey in the Trust Onion app. Verification checks the passkey signature against the full public key, the original photo's SHA-256 hash, and the signed record, binding that key to the exact image and the date, time, and location at signing. It does not, by itself, guarantee the intended person's physical presence.
In the Fideuram scenario, a codeword challenge inserted before the transfer authorization would have stopped the payment chain before the first euro left the account.
€36 million is still missing.
A controlled voice-cloning safety demonstration
Try our tool to see just how easy voice cloning is. This controlled safety demonstration exists only to show how easy voice cloning is and to make the risk tangible. It is not a real incident or proof of anyone's identity. Your submitted sample is handled transiently to generate the fixed demonstration audio. Trust Onion does not identify you or the person from the sample. Trust Onion does not store the submitted recording, clone, or generated audio.
Public by default, under your control. Proofie™ photos and signed records are stored securely on Amazon Web Services (AWS). A Proofie™ is public by default: anyone with its link or QR code can view the photo and signed record. You can add a password to make a Proofie™ private, or delete it at any time. Deleting a Proofie™ removes its photo and signed record from our AWS storage and takes its link offline, but cannot recall copies someone has already downloaded or saved.
Frequently asked questions
How did fraudsters steal €95 million from an Italian bank using AI?
Attackers used a spoofed WhatsApp message impersonating Intesa Sanpaolo CEO Carlo Messina and an AI-cloned voice impersonating a law firm partner to convince Fideuram chairman Paolo Molesini to authorize large international wire transfers. The two channels made the instruction appear to come from two independent trusted sources simultaneously.
What is a two-channel voice and text fraud attack?
A two-channel attack combines a written message, such as a WhatsApp or email, with a phone or audio call to create the appearance of independent corroboration. When two seemingly separate trusted contacts confirm the same instruction, targets are far less likely to challenge it before acting.
How can finance teams verify executive wire transfer requests?
Finance teams should require a codeword challenge before approving any wire transfer or payment instruction change. Rotating codewords shared only with verified personnel cannot be obtained from public audio, spoofed messages, or AI-cloned voices. The real executive knows the current words; an impersonator does not.
Does calling back on a known number confirm a caller's identity?
A successful callback confirms you reached the real person on a second attempt, but it does not prove the original caller was legitimate. For high-stakes approvals, verification must happen before the transfer is authorized, not after funds have already moved.
How much of the Fideuram fraud money was recovered?
International cooperation recovered approximately €53 million of the roughly €95 million transferred. Around €36 million remains untraced after passing through overseas accounts and being converted to cryptocurrency.
Add Trust Onion's rotating codeword check to executive and wire-transfer approvals: before any wire, vendor change, or urgent payment instruction, your team asks "What are the words?" and only the real executive can answer. Protect Your Business Free.
Protect Your Business FreeCite this page
AI Voice Clone and Fake WhatsApp Cost Italian Bank €95M — Trust Onion, October 11, 2026. https://trustonion.io/blog/ai-voice-clone-fake-whatsapp-italian-bank-wire-fraud



