Italy’s Fideuram Bank Loses €95 Million in AI Voice Scam

Italy’s Fideuram Bank, the private banking arm of Intesa Sanpaolo, was targeted in a sophisticated AI-powered scam that allegedly tricked officials into transferring about €95 million ($108 million) to overseas accounts using a cloned voice and fake messages, authorities said.

MILAN, Italy — An elaborate financial scam involving artificial intelligence and a cloned voice has resulted in the theft of about €95 million ($108 million) from an Italian bank, exposing how rapidly criminals are using AI to make fraudulent instructions appear genuine.

The bank targeted in the sophisticated scheme was Fideuram, the private banking arm of Intesa Sanpaolo, one of Italy’s largest banking groups.

The fraud began with what appeared to be an ordinary WhatsApp message. Paolo Molesini, Fideuram’s former chairman, received a message from someone pretending to be Carlo Messina, the chief executive of Intesa Sanpaolo.

The fake executive reportedly requested urgent assistance involving a major overseas transaction. The message was followed by a phone call from a person claiming to be a prominent lawyer who was supposedly involved in the transaction.

What made the deception particularly convincing was the use of AI voice-cloning technology. The scammers allegedly reproduced the lawyer’s voice and used it during the call to make the transaction appear legitimate.

Believing that the request was genuine, Fideuram officials proceeded with the transfer of tens of millions of euros to overseas bank accounts, including accounts connected to China and Hong Kong.

The deception was eventually discovered after the bank detected irregularities surrounding the transactions. Authorities were then alerted, triggering an international effort involving investigators from several countries.

READ OTHER ARTICLES:

A significant portion of the stolen money was recovered or frozen before it could disappear completely. Reports indicate that authorities managed to recover or secure more than €53 million, with some estimates putting the amount at around €59 million.

Roughly €36 million remains unaccounted for the investigators believe the remaining funds were moved rapidly through multiple overseas accounts before being converted into cryptocurrency, making the money increasingly difficult to trace.

The use of cryptocurrency has added another layer of complexity to the investigation because funds can be transferred across borders through a series of digital wallets and transactions, potentially obscuring the identities of those controlling the money.

The case is being investigated by prosecutors in Milan, with international cooperation playing an important role in efforts to trace the missing funds and identify those behind the operation.

At least one foreign national living outside Europe has reportedly been placed under investigation in connection with suspected computer fraud. However, the person being investigated has not necessarily been established as the mastermind of the entire operation.

No final determination has yet been made about who organized the broader scheme. The case highlights a growing challenge for banks and other financial institutions: traditional security procedures can become vulnerable when criminals combine social engineering with increasingly convincing artificial intelligence tools.

A fraudulent message can establish the initial contact, while a cloned voice can provide what appears to be human confirmation. Together, the techniques can create a convincing chain of communication designed to pressure officials into making decisions quickly.

For financial institutions, the incident demonstrates why a familiar voice or seemingly authoritative message may no longer be sufficient proof that an instruction is genuine.

The recovery of more than half of the stolen money represents a major development in the investigation, but the millions that remain missing show how difficult international financial fraud can become once money crosses multiple jurisdictions and enters cryptocurrency networks.

The Milan investigation remains ongoing, and authorities are continuing to trace the missing funds while working to determine the full network of individuals involved.

For now, the case stands as a striking example of how AI voice cloning has moved beyond simple impersonation and become a powerful tool in sophisticated financial fraud, forcing banks to rethink how they verify high-value transactions even when the person giving the instructions appears to be someone they know.

AI Voice Clone Used in €95 Million Fideuram Bank Scam, But Software Remains Unknown

Authorities have not publicly identified the specific software, artificial intelligence platform, or website allegedly used by scammers to clone a lawyer’s voice in the sophisticated €95 million fraud targeting Fideuram, the private banking arm of Intesa Sanpaolo.

The case has drawn attention to the growing use of artificial intelligence in financial fraud after criminals allegedly combined a fake WhatsApp message, fabricated communications and an AI-generated voice to convince Fideuram’s former chairman, Paolo Molesini, that he was dealing with people connected to the bank and a major international financial transaction.

According to Italian media reports, the fraud began in February 2026 when Molesini received a WhatsApp message that appeared to have been sent by Carlo Messina, chief executive of Intesa Sanpaolo. The message allegedly requested urgent assistance with an overseas financial operation.

Molesini received a telephone call from someone who appeared to be Paolo Nastasi, managing partner of A&O Shearman Italy. Nastasi was actually unrelated to the scheme, and investigators believe his voice had been reproduced using artificial intelligence.

The use of a familiar voice allegedly made the fraudulent request appear more credible. Following the communications, approximately €95 million was transferred, with the money sent primarily toward accounts in China and Hong Kong.

A substantial portion of the money was subsequently recovered through international cooperation. Reports indicate that more than €50 million was either recovered or frozen, while approximately €36 million remains to be traced. Some of the missing funds were reportedly converted into cryptocurrency, making the recovery effort more difficult.

Despite the growing attention surrounding the use of AI in the case, authorities and Fideuram have not publicly named the particular AI software or platform used to create the cloned voice.

That distinction is important. While numerous commercial and open-source AI systems are capable of generating or cloning voices, there is currently no reliable public evidence establishing that the scammers used a particular service such as ElevenLabs, Resemble AI, Murf, Bark or VALL-E in the Fideuram operation. Therefore, naming any of those platforms as the software used in this crime would go beyond what has been publicly established. Instead, the available reporting describes the technology more generally as AI-based voice cloning or voice replication.

Investigators believe the criminals were able to reproduce the voice of Paolo Nastasi, a lawyer whom Molesini knew. That familiarity may have made the telephone conversation more convincing.

Publicly available recordings can potentially provide material for voice-cloning systems, including interviews, conferences, webinars, podcasts and other recordings. However, there is no confirmed public evidence that the scammers specifically obtained Nastasi’s voice from YouTube, podcasts, webinars or news interviews in this case.

That means those sources can be discussed as possible ways a voice could be obtained in similar crimes, but they should not be presented as the confirmed source of Nastasi’s voice.

The Fideuram case therefore illustrates a broader cybersecurity problem: criminals do not necessarily need to break directly into a bank’s computer systems to create a convincing fraud. By combining publicly available information, impersonation, messaging platforms and AI-generated audio, they can construct a communication chain designed to appear legitimate.

The investigation remains ongoing under the Milan Prosecutor’s Office. Italian authorities are working with foreign counterparts to trace the remaining funds and identify the individuals behind the operation.

At least one foreign national has been reported as being under investigation in connection with the suspected fraud, although the investigation has not publicly established that person as the mastermind behind the entire operation. For now, the exact AI tool used to reproduce the lawyer’s voice remains undisclosed.

What is already clear, however, is that the scammers used AI-generated voice technology as one component of a broader impersonation scheme that helped persuade a senior banking official to authorize transactions worth tens of millions of euros.