AI Voice Scams Spread: 1 in 4 Americans Has Received a Deepfake Call, and an Italian Bank Lost €95 Million
In our last post, we looked at how AI is changing the way people search. This time, AI reaches somewhere much closer to home: your phone. AI voice cloning (deepfake voice) is now cheap and convincing enough for scammers to sound like a relative, a manager or a partner in a single call.
In October 2026, police in several countries issued fresh warnings, after a private bank in Italy lost €95 million to a cloned-voice call. The shared lesson is simple: a familiar voice is no longer proof.
1. One in four Americans has received an AI deepfake voice call
Hiya's State of the Call 2026 report, which surveyed more than 12,000 consumers in 6 countries, shows that deepfake calls are no longer rare.
Widespread: 25% of Americans say they received at least one AI-generated voice call in the past 12 months.
Hard to tell apart: another 24% admit they cannot tell a cloned voice from a real one.
Constant pressure: the average American receives 9.9 unwanted calls per week, more than 500 a year.
The overall cost: according to the US Federal Trade Commission (FTC), American consumers reported losing $15.9 billion to fraud in 2025, with impostor scams alone above $3.5 billion.
Note: the Hiya figures come from a self-reported consumer survey published by a call-blocking company, so they reflect perception more than verified incidents.
2. From "your child is in trouble" to "the boss needs a transfer now"
Recently reported cases follow two familiar scripts, and both rely on urgency.
Impersonating family: on October 19, 2026, police in Cyprus warned that scammers are using AI to clone the voices of relatives and friends "in trouble" to ask for money. Vietnamese police have repeatedly warned about the same tactic through video calls and voice messages on Zalo and Messenger.
Impersonating leaders: in Italy, the chairman of Fideuram (part of Intesa Sanpaolo) received a WhatsApp message that appeared to come from the group CEO, followed by a call using the cloned voice of a senior lawyer. The result was €95 million sent abroad, as Reuters reported on September 25, 2026. Sources say part of it has been recovered, but the figures differ.
Raw material is already public: scammers usually build fake voices from videos, talks and clips posted publicly on social media. The more someone appears online, the easier their voice is to clone.
The Agenorix view: verify with a process, not with your ears
We build chatbots and AI agents every day, so we know how real AI-generated voice and text have become. That is why the effective defense is not "listening more carefully" but having verification steps that do not depend on gut feeling.
Agree on a family password: a word or question only your household knows, used whenever someone calls asking for money urgently.
Call back on a saved number: hang up and call the familiar number yourself, never a number or link the caller gives you.
Two channels for every transfer: within your team, any payment request made by phone or message needs confirmation through a second channel and a second person.
Slow down when rushed: pressure to "do it now" is the clearest sign of a scam. No legitimate payment is lost because you took 5 minutes to check.
Transparent AI is trustworthy AI: the AI assistants Agenorix builds always introduce themselves as AI and never ask customers to send money or share an OTP code.
Where to start?
Tonight, take 10 minutes to agree on a family password with your parents and relatives, and remind each other to call back on a saved number before sending any money. If you want to review your verification process, or build safe and transparent chatbots and AI agents for your team, the Agenorix team is ready to help, and your first consultation is free: get in touch.
Sources: Hiya State of the Call 2026 (Mar 2, 2026), FTC (Mar 25, 2026), Reuters (Sep 25, 2026), Cyprus Police via Cyprus Mail (Oct 19, 2026), Vietnam Ministry of Public Security.
Share