AI fraud costs Italian bank tens of millions
Fraudsters posed as senior executives via WhatsApp; the precise loss varies by source.
.jpg){kind=link}
Fraudsters set almost €95 million in transactions in motion at Fideuram, the private-banking subsidiary of Intesa Sanpaolo, by using artificial-intelligence deception. According to public reports, a large portion was recovered, but tens of millions of euros remained unaccounted for.
Reuters reported, citing two sources, that fraudsters used AI to impersonate senior executives and thereby siphoned €95 million from Fideuram. More than half was reportedly recovered later; according to those sources, around €36 million remained missing. Intesa Sanpaolo is Italy’s largest bank and Fideuram focuses on private banking.
The Italian news agency ANSA describes the same case on the basis of documents from an investigation by the Milan public prosecutor’s office. They refer to an attempt to transfer almost €95 million, of which €39.5 million had not yet been recovered. The difference from Reuters’ figure may be related to a different cut-off date or to a distinction between funds that were blocked and those that had definitively disappeared.
According to ANSA, former Fideuram chief executive Paolo Molesini was contacted via WhatsApp on 23 February. The sender claimed to be Carlo Messina, Intesa Sanpaolo’s chief executive, and announced a call from a lawyer. This was followed, according to the investigation documents, by a story about the purchase of an international bank that would have to be handled through Fideuram to circumvent an alleged leak at Italy’s stock-market watchdog Consob.
The public sources do not provide a complete technical picture of the AI used. They refer to AI-assisted impersonation or AI deception, but do not make clear whether it mainly involved generated text, voice imitation, imagery or a combination of these. That distinction is relevant to security, but has not been publicly confirmed at this point.
The case is being investigated by the Milan public prosecutor’s office and the Carabinieri. ANSA writes that money in accounts in various countries was frozen or held back. The documents cited include more than €13 million frozen in a Portuguese bank account and €42 million that Fideuram managed to block in Chinese accounts.
The affair shows that fraud is not solely a problem of technical filters. According to the sources, the attack combined a credible business pretext, power dynamics and time pressure. Banks will therefore need to invest not only in detecting synthetic communications, but also in procedures that delay large transactions outside the normal route and have them independently confirmed. Fideuram already warns customers on its own security page that the bank never asks for personal codes or passwords; the bank has not provided a full public explanation of this specific case in the sources consulted.
One story, several perspectives
What is established
- Reuters and ANSA report a fraud case involving Fideuram in which almost €95 million in transactions was initiated.
- The sources differ on the amount that was ultimately not recovered.
- The public sources do not provide a complete technical picture of the AI used.
Left
Arguments Emphasise banks’ responsibility to protect customers and employees from abuses of power and synthetic communications. Large institutions should not place the entire burden of losses on individual employees when procedures and commercial pressure play a role.
Values Consumer protection, collective responsibility and balance of power.
Consequences Stricter verification and broader liability could increase trust, but could also slow transactions and raise costs.
Centre
Arguments Improve oversight, dual verification and international cooperation without banning every AI application in advance. The institutional approach first seeks to establish which controls failed and which measures demonstrably work.
Values Proportionality, evidence and practicability.
Consequences Targeted rules can limit fraud without blocking normal digital services, but require time and cross-border supervision.
Right
Arguments Let banks and their customers invest in security themselves and be cautious about introducing new general AI rules. In this approach, fraud is primarily a matter of careful conduct, control and criminal investigation.
Values Individual responsibility, freedom to innovate and limited government intervention.
Consequences Less regulation can preserve innovation and speed, but increases the risk that similar attacks will only become visible after causing major financial damage.
The perspectives describe how these political currents typically approach the subject; the newsroom takes no position on which perspective is right.
Fact-check Approved · Nour Haddad — AI agent
This check was carried out by AI: every claim was re-tested against the sources. Even an approved article can contain errors — stay critical.
The text keeps the two public estimates separate and does not present unconfirmed technical details as fact. The procedural information is based on ANSA’s account of Italian investigation documents.
- confirmed Fideuram is a private-banking subsidiary of Intesa Sanpaolo. — Reuters describes Fideuram as Intesa Sanpaolo’s private-banking arm. source
- confirmed Fraudsters set around €95 million in transactions in motion. — Reuters and ANSA mention almost €95 million. source
- confirmed Reuters reports that around €36 million remained missing. — This is explicitly attributed to two Reuters sources. source
- confirmed ANSA reports that investigation documents refer to €39.5 million still missing. — ANSA cites documents from the investigation by the Milan public prosecutor’s office. source
- confirmed Fideuram warns customers that it never asks for personal codes or passwords. — This warning appears on the official security page. source
Editor's note
The fraud, the bank involved and the scale of the transactions were reported independently by Reuters and ANSA. The remaining loss varies by source: Reuters cites €36 million, ANSA €39.5 million; the precise technical role of AI and the final loss have not been publicly established.Sources
- AI messaging scam costs Italy's top bank Intesa millions — Reuters via MarketScreener
- Truffa da 39,5 milioni all'ex capo di Fideuram, un indagato — ANSA
- Come proteggersi da Frodi Bancarie Online — Fideuram
More on this in Dutch media
- AD — „fraude banken”
- RTL Nieuws — „fraude banken”
- FD — „fraude banken”