- Criminals used fake WhatsApp messages and an AI cloned voice to impersonate senior officials and a lawyer.
- The scam resulted in international transfers worth more than $100 million.
- About $60 million was recovered, while roughly $40 million reportedly remained missing.
- The case shows how AI is making traditional business fraud more convincing through voice cloning and deepfake technology.
Italy’s banking sector has become the latest target of a sophisticated fraud scheme involving artificial intelligence, deepfake technology and impersonation. Criminals reportedly persuaded a senior executive at Fideuram, the private banking arm of Intesa Sanpaolo, to approve international transfers worth more than $100 million.
The incident shows how traditional business fraud is changing as criminals gain access to increasingly convincing AI tools. What once relied largely on fake emails and compromised accounts can now involve cloned voices, convincing messages and carefully staged conversations designed to make an instruction appear legitimate.
According to Reuters, the incident took place in February 2026, when Paolo Molesini was chairman of Fideuram and Intesa Sanpaolo Private Banking.
The fraud began with a WhatsApp message that appeared to have been sent by Carlo Messina, the chief executive of Intesa Sanpaolo. The message reportedly asked Molesini to assist with an urgent overseas transaction.
The request was followed by contact from an individual claiming to be a senior partner at a major law firm. That person reportedly used an AI generated clone of the lawyer’s voice, making the conversation sound authentic and reinforcing the impression that the transaction had been properly authorized.
Molesini subsequently contacted the bank’s finance department and arranged a series of transfers to accounts overseas.
Deepfake technology added another layer of deception
The most striking part of the incident was not simply the use of a fake identity. The criminals reportedly combined several methods to create a believable chain of communication.
The initial message created the impression that the request had come from the bank’s top leadership. The supposed lawyer then provided another layer of confirmation through a voice call.
That combination is particularly concerning for financial institutions because employees are accustomed to verifying unusual transactions through conversations with senior executives, lawyers and other trusted contacts.
AI generated voices can make that process much harder to rely on. A voice that sounds exactly like a familiar colleague or senior executive is no longer necessarily evidence that the person is actually on the other end of the call.
The fraud also reportedly used the familiar characteristics of business compromise scams. The transaction was presented as urgent and sensitive, creating pressure to act quickly while limiting the opportunity for routine verification.
The transfers were sent mainly to accounts in mainland China and Hong Kong.
The fraud was discovered relatively quickly, allowing Fideuram and authorities in several countries to recover a substantial portion of the money. Around $60 million was reportedly recovered with assistance from authorities in China, Portugal and Italy.
However, approximately $40 million remained missing. According to the report, some of the unrecovered funds were converted into cryptocurrency, making recovery more difficult.
A new version of an old business scam
The incident is closely related to what security professionals have traditionally called Business Email Compromise, or BEC.
In a typical BEC operation, criminals impersonate a company executive or compromise an email account before contacting employees responsible for payments. The victim is then instructed to send money to an account controlled by the criminals.
AI is expanding the number of ways these attacks can be carried out.
Instead of relying solely on email, criminals can now use messaging applications, phone calls and potentially video conversations. Voice cloning can make an impersonated executive sound genuine, while AI generated images and video can add another layer of credibility.
The Italian bank case demonstrates why simply recognizing suspicious email addresses may no longer be enough.
A message can arrive through a legitimate messaging platform. A follow up call can feature a convincing voice. The person on the other end may appear to know internal details about the organization.
Each individual piece of evidence may look credible while the entire interaction has been carefully constructed by criminals.
Senior executives are also becoming targets
The case also highlights an important change in the threat landscape. These scams are not necessarily aimed at junior employees who may be unfamiliar with company procedures.
Senior executives can also become targets because they have the authority to approve or initiate large financial transactions.
Molesini later announced that he was stepping down from his positions at Fideuram and Intesa Sanpaolo Private Banking in March 2026, citing personal reasons. Reuters reported that neither Molesini nor other Fideuram executives were under investigation in connection with the fraud.
Authorities were reportedly investigating a foreign national living outside Europe over potential involvement in the scam.
The incident follows another reported AI voice fraud in Italy. In that case, criminals allegedly cloned the voice of an Italian minister and used it to persuade businessman Massimo Moratti to transfer approximately $1.13 million to an overseas account. That money was subsequently recovered.
The growing use of AI in financial crime means companies may need to rethink how they verify high value transactions. A familiar voice, recognizable writing style or apparently authoritative message can no longer be treated as sufficient proof of identity.
For banks and other businesses handling large transfers, independent verification through a separate trusted channel can provide an additional layer of protection. The key issue is making sure that the verification process does not rely on the same communication channel used to issue the original instruction.
The Fideuram incident is therefore significant beyond the amount of money involved. It illustrates how AI is making established fraud techniques more convincing and harder to detect, particularly when criminals combine social engineering with deepfake technology.
Follow TechBSB For More Updates
