FrançaisEnglishEspañolالعربيةहिन्दीবাংলা中文

← Back to the blog

Published on 2026-07-25

Could the voice on the phone be an AI clone, not your relative?


A mother heard her daughter screaming for help. It took less than ten seconds of old video to fake that voice. Here is the one question that still catches it.

Rachel, a mother in Missouri, picked up the phone and heard her college age daughter crying that she had been in an accident. Seconds later a man's voice took over, saying the daughter had been kidnapped and would be killed unless Rachel wired money to Mexico immediately. She believed the voice. Of course she did: it sounded exactly like her daughter, because in a way, it was.

Ten seconds is the whole budget

Cybersecurity expert Ann Dickherber, interviewed after Rachel's ordeal, put a number on how little material a scammer needs: 'Less than 10 seconds of your voice is all I need to be able to make a clone of your voice that would fool anyone.' A birthday video posted online, a voicemail greeting, a clip from a school project: any of it is enough raw material. The software does not need your daughter to have said the sentence it makes her say. It only needs to have heard her voice say something, anything, once.

Rachel wired thousands of dollars before the scam unraveled in the strangest way possible: her actual daughter started calling her, alive, confused, and very much not in Mexico. The fake had a flaw no amount of audio training fixes, it could not stop the real person from also being reachable.

The demand always arrives fast

A similar case reported by local news in Ohio followed the same script almost line for line. Deborah Del Mastro received a call demanding 20,000 dollars for her daughter's release, the voice on the line sounding enough like her child that she sent 5,000 dollars before the picture fell apart and she realized nobody had actually been taken anywhere. In both cases the money moved through channels built for speed and built to be nearly impossible to claw back: wire transfers, the kind with no undo button.

The FBI has been tracking a rise in these calls, sometimes paired with a fabricated photo to make the panic feel even more concrete. Its advice does not try to compete with the audio quality, which keeps improving. It shifts the fight somewhere the software cannot follow: agree on a private word with the people closest to you, a family password nobody outside that circle would ever know or guess, and treat any panicked call as a prompt to verify, not to act. A password works precisely because it lives nowhere online for a scammer to scrape: no voice sample teaches the software a word that was never spoken into a microphone in the first place.

None of this makes Rachel or Deborah careless. Quite the opposite: they reacted the way almost anyone would to their own child sounding terrified on the line. That is precisely what makes the scam effective and why it keeps spreading. It does not rely on a victim being gullible. It relies on a nervous system doing exactly what it evolved to do when a loved one seems to be in danger, which is to act first and question later. Knowing that in advance is what lets you build the pause in ahead of time, instead of trying to invent it in the middle of a panic.

A voice is not an identity

This is really the same old question dressed up in new technology: who is actually speaking to you? A familiar voice used to be proof enough, because faking one took a skilled impersonator and a lot of luck. That shortcut is gone. The sound alone can no longer answer the question, so the question has to be answered another way: hang up, and call your relative back on the number you already have saved, not one given to you by the caller. Ask something only they would know. Wait the ten seconds it takes for a real answer to arrive.

The scam needs your panic to move faster than your phone can dial a callback.

None of this requires you to become suspicious of every call from someone you love. It requires one habit: when a request for money arrives wrapped in fear, urgency and a voice you trust, you put the phone down for thirty seconds and dial the number you already had. If the emergency is real, thirty seconds changes nothing. If it is not, those thirty seconds are the whole difference.

Topics : scams voice cloning family

Sources

Keyboard shortcuts

?Open this help
EscClose the panel or this help
1-3Pick an answer during the quiz
EnterNext case, once you've read the answer
TabMove from link to link: the whole site works by keyboard