Hassan SalemNotebook · München
← Notebook
Communication27 September 20264 min read

Your Voice Is No Longer Proof of You

A familiar voice or face used to be proof. Now it's only a claim. Trust lives in context and simple habits, not in how someone sounds.


Last week I read that a bank lost €95 million because someone sounded like a lawyer. This week a startup showed an AI that almost half the people on a video call took for a real person.

Put these two stories together and one old assumption breaks: hearing a familiar voice, or seeing a familiar face, no longer tells you who is on the other side.

What happened

According to Reuters, in February 2026 Fideuram, the private banking arm of Intesa Sanpaolo, was tricked into sending about €95 million abroad. It started with a WhatsApp message that appeared to come from Carlo Messina, the CEO of Intesa Sanpaolo. Then came a phone call where AI was used to clone the voice of a senior lawyer at a well-known law firm, confirming the urgent transaction. About €53 million was recovered. Around €36 million is still missing, partly converted into crypto. Milan prosecutors are investigating.

Then on October 1, Tavus introduced Griffin, a real-time video model. In their study, 54 people had one-minute video calls with it after being told they were talking to another person. 26 of them, 48%, believed it was human. Their previous system fooled 1 out of 41. Griffin is still a research preview, open only to selected testers.

48% on a one-minute call is close to a coin flip. That's the point.

My observation

We have always used voice and face as a shortcut for trust. Think about how many things you approved because "the person asked me directly". A quick call from your manager. A message from a colleague you know well. You never asked yourself if it was really them. Why would you?

What strikes me in the Fideuram story is not the technology. It's the pattern: urgency, authority, and a second "confirmation" that felt independent. That's classic social engineering. AI just made the costume much better.

And I think about engineering teams. We get urgent pings all the time. "Can you rotate this key?" "Can you give me prod access for ten minutes?" "Approve this, the CEO is waiting." The pressure always comes with a familiar name attached.

A familiar voice used to be proof. Now it's only a claim, and claims need checking.

What to expect next

I expect more of these cases, not fewer, and in normal companies, not only banks. Voice cloning is already easy to get. Real-time video avatars will follow.

I also expect verification to become a normal part of how we communicate, the same way two-factor login did. A bit annoying at first, then nobody thinks about it anymore.

The learning

Trust was never really in the voice. It was in the context: shared history, the right channel, a request that makes sense. When the voice can be copied, context is what's left.

And the skill that matters most is calm. Being able to say "let me call you back" to someone who sounds exactly like your boss, without feeling rude about it.

What can we do next

  1. Verify on a channel you choose. If a request comes by call or chat, confirm it through a channel you start yourself, like the number in the company directory. Never through a link or number the requester gives you.
  2. Treat urgency as a signal. "Do it now and don't tell anyone" is the classic shape of a scam. Slow down exactly when someone pushes you to hurry.
  3. Agree on simple checks with your team and family. A code word, a question only the real person can answer, or a rule that money and access never move on a single call.
  4. Managers, make questioning safe. Tell your team clearly: if I ask for something unusual and you call me back to check, you did the right thing. People only verify when verifying doesn't feel like disrespect.
  5. Put guardrails in the process, not in people's courage. Two approvals for payments, no access grants over chat, time-limited permissions. A system should never depend on someone spotting a fake voice.

Your face and your voice are still yours. They are just not proof anymore. Your habits are.

Sources: Cybernews: AI voice scam hits Italian bank for €95 million, Gulf News: AI voice-cloning scam hits Italian bank, Tavus: Griffin, Crypto Briefing: Tavus unveils Griffin

Written by Hassan Salem in Munich — software engineer, slow German learner, compulsive note-taker. More about me →