Italian Bank Loses €95 Million in AI Voice Cloning Fraud Targeting Top Executives

Intesa Sanpaolo's private banking subsidiary Fideuram was defrauded of €95 million through a scheme using AI-cloned voices and fake WhatsApp messages impersonating the CEO. Fraudsters requested urgent international transfers and used artificially generated voice to confirm the request, resulting in transfers to accounts in China and Hong Kong.

How the Fraud Worked

The scheme began in February when Fideuram's President Paolo Molesini received a WhatsApp message allegedly from Carlo Messina, CEO of Intesa Sanpaolo, requesting urgent help with a foreign transaction. Shortly after, a person claiming to be a senior partner at a prestigious law firm called to confirm the request. The voice in this call was artificially created using AI technology. Believing the messages to be legitimate, Molesini instructed the finance department to transfer funds to accounts in China and Hong Kong.

Extent of the Loss

A total of €95 million was stolen. When Fideuram discovered the irregularities and alerted authorities in multiple countries, investigators from China, Portugal, and Italy collaborated to recover approximately €53 million. However, about €36 million remains missing after being routed through foreign accounts and converted into cryptocurrency.

Warning Signs and How to Identify This Scam

Characteristics of AI-Generated Voices:

Slight delays or unnatural intonation patterns
Absence of background noise or unusually clear silence
Overly direct answers to complex questions
Minor audio artifacts or processing sounds

Situational Red Flags:

Requests that emphasize urgency or time pressure
Instructions coming through unusual channels
Large transfers requested through a single authorization point
Requests that bypass normal approval procedures

Prevention and Protection Measures

Organizations and banks should implement the following safeguards:

Multi-Channel Verification: Large transactions must be confirmed through multiple independent channels. When verbal confirmation is required, call back the registered phone number to verify the request.
Emergency Code Words: Establish pre-agreed authentication phrases or code words to be used during high-value transactions.
Voice Verification Technology: Deploy tools such as Pindrop or Reality Defender to detect synthetic voices in real-time.
Employee Training: Conduct regular training for all staff, including executives, on voice fraud tactics and proper verification procedures.
Transaction Limits and Controls: Implement threshold controls requiring multiple approvals for unusually large transfers.

Related Incidents

Italy has experienced a rising trend of such scams. In 2025, Massimo Moratti, former president of Inter Milan, lost approximately €1 million to a similar AI voice fraud impersonating Defense Minister Guido Crosetto (funds were later recovered). Internationally, the 2024 case involving British engineering firm Arup resulted in a loss of approximately US$25 million when an employee in Hong Kong transferred funds after a video conference with deepfake colleagues. The Fideuram incident, at €95 million, ranks among the largest known deepfake-enabled frauds globally.

Where to Report

If you suspect you have been targeted by voice fraud, immediately contact your national banking regulator or fraud investigation authority. In Italy, reports can be filed with local Carabinieri or the Guardia di Finanze (Financial Police).