Artificial intelligence is changing more than productivity. It is changing the way criminals manufacture trust.
A striking example emerged from Italy in September 2026, when details became public about a fraud involving Fideuram, the private banking arm of Intesa Sanpaolo. According to Reuters and Trending Topics, fraudsters impersonated Intesa Sanpaolo CEO Carlo Messina through WhatsApp and followed up with a phone call using an AI-cloned voice of a senior lawyer known to the target. Believing the request was genuine, Fideuram’s then-chairman Paolo Molesini instructed a series of overseas transfers totaling about €95 million.
Around €53 million was later recovered through international cooperation, while roughly €36 million was reported still missing. The incident is a powerful reminder that in the age of deepfakes, recognizing a familiar voice is no longer proof of identity.
Deepfakes Have Moved From Curiosity to Financial Risk
For years, deepfakes were discussed mainly in the context of manipulated videos, celebrity impersonations and misinformation. Voice cloning has changed the risk profile.
Executives regularly appear in interviews, webinars, conference recordings, podcasts and earnings calls. Those recordings can provide criminals with material that may be used to imitate speech patterns and voices. When that synthetic voice is combined with urgency, authority and a believable business story, traditional human instincts can be manipulated.
The Fideuram case reportedly used more than one channel. A WhatsApp message appeared to come from the chief executive, a telephone call appeared to confirm the request, and payment instructions were then supplied for overseas transfers.
That matters because people often interpret agreement across several channels as independent confirmation. If every channel is controlled by the same attacker, however, multiple messages can create only the illusion of verification.
Executive Authority Can Become an Attack Surface
Cybersecurity programmes often focus heavily on systems: endpoints, networks, applications, identities and cloud environments.
They also need to consider authority.
Senior executives can approve payments, override normal workflows, request secrecy and create urgency simply because of their position. Criminals understand this. Generative AI makes it easier to reproduce the signals people associate with authority, including voice, writing style and appearance.
This means “I recognised the person” can no longer be an adequate control for sensitive actions.
Organizations should instead design processes that remain secure even when a message, email, voice call or video meeting looks authentic.
How Businesses Can Reduce Deepfake Financial Fraud
The strongest defence is procedural rather than psychological.
Large or unusual payments should require independent approval by more than one authorized person. New beneficiary details should be validated through a trusted channel already held by the organization, not a phone number or email address supplied in the request itself.
Organizations should also define transaction thresholds that automatically trigger additional checks. Urgency, confidentiality and executive seniority should never remove those safeguards.
Finance and treasury teams need deepfake-aware training as well. Employees should understand that a convincing voice or video may be synthetic and that challenging an unusual request is part of their job, not an act of insubordination.
Security teams can reinforce these controls with anomaly detection, logging, payment monitoring and clear escalation procedures.
Trust Must Become Verifiable
The lesson from this case is not that organizations should stop trusting their leaders.
It is that trust needs a technical and procedural foundation.
In the past, hearing a familiar voice during a confirmation call might have been reassuring. In 2026, that same voice can potentially be generated by software.
This changes the meaning of verification.
A secure organization does not ask employees to become perfect deepfake detectors. It builds processes in which one convincing interaction cannot move millions of euros.
Deepfake financial fraud is likely to become more sophisticated as generative AI improves. The organizations that respond effectively will be those that redesign approval processes before the next urgent message arrives.
The question for every business leader is simple:If someone cloned your CEO’s voice today, what could they authorize tomorrow?
Have knowledge, experience, or a powerful idea you want to turn into a book? Get your book published with DevOM Publishing:
https://www.devompublishing.com/
If your business needs cybersecurity services, AI security guidance, cloud security review, or protection against modern cyber threats, visit DevOM Consulting:
https://devomconsulting.com/
