Most advice about artificial intelligence is about what it can create. This article is about something darker and more immediate: a phone call that sounds exactly like your son, your sister, or your boss, and isn’t.
AI voice cloning scams are the newest version of an old trick. The FTC has warned that scammers now clone voices to make phone fraud more convincing, particularly “family emergency” calls in which a relative supposedly needs urgent money. What has changed is how little effort it takes to sound real.
Below, you’ll find how these scams work, what the published numbers actually say (and why they disagree so wildly), why families split across countries are especially exposed, and the practical steps that hold up even if the voice is perfect.
How the scam works
The structure is simple. A call arrives from an unknown or spoofed number. The caller, in a voice you recognize, says something terrible has happened: an arrest, an accident, a lost passport in another country. They need money immediately, and usually secretly. As one report describes it, scammers typically request quick money transfers or gift cards.
The technology adds realism, not a new plot. According to the Federal Trade Commission, a scammer needs only a short audio clip of a person’s voice, which can come from content posted online, plus a voice-cloning program. One security guide notes that early versions of these scams used pre-recorded clips, while voice cloning can now run in real time during a live call, letting a scammer hold an interactive conversation in someone else’s voice.
That last point is worth pausing on. A stilted, repetitive script used to be a giveaway. A live, responsive conversation removes it.
Why families split across borders are exposed
This part is my own analysis rather than a published finding, but it follows from how the scam works. The scam depends on two things: a believable emergency, and a reason the victim can’t easily check. Families where someone works abroad have both built in.
A parent in one country receives a call from a “son” who says he has been arrested or injured in another. The distance explains the unfamiliar number. The time difference explains the odd hour. Money transfers across borders are normal, so a request to wire cash doesn’t stand out. Reversed, the same trick works on a worker abroad who gets a panicked call about a sick parent.
If you or your relatives are migrant workers, students abroad, or family members of either, treat the scenario as likely, not remote.
What the numbers say, and why they disagree
Search for statistics and you’ll find figures that don’t remotely match. Here’s how to read them.
| Claim | Where it appears | How much to trust it |
|---|---|---|
| The FBI’s 2025 Internet Crime Report, released April 2026, tracked AI-related fraud as a formal crime category for the first time | Security blog summarizing the report | Plausible and consistent; verify at ic3.gov |
| About $893 million in losses across more than 22,000 complaints, with over $352 million from adults over 60 | A scam-awareness site citing FBI data | Secondary source; check the FBI report itself |
| One in four respondents in a McAfee study of 7,000 people experienced an AI voice-cloning scam or knew someone who had | News report on the study | A vendor survey, so it measures awareness, not confirmed losses |
| Figures like billions in annual global losses or hundreds of thousands of complaints in a single quarter | Statistics blogs and commentary | Often no traceable source, and they conflict with one another |
The honest summary: the trend is real and authorities are tracking it, but precise totals vary from source to source. Where an article states a dramatic global figure without naming a primary source, treat it with caution. For any figure you plan to repeat, go to the FBI’s Internet Crime Complaint Center or the FTC directly.
The warning signs that matter
One security guide makes a point that is easy to miss: the strongest warning signs are not in the voice itself but in the situation. People are told to listen for robotic tones or odd pauses, but a good clone may have none of that.
Look instead for:
- Extreme urgency, with a reason you can’t wait or call anyone else
- A request for secrecy, such as “don’t tell Mum”
- Unusual payment methods, such as gift cards, wire transfers, or cryptocurrency
- An unfamiliar number, with a story about a lost or broken phone
- A second caller, such as a “lawyer” or “officer,” who takes over to collect money
The habit that beats it: a family code word
Security guidance consistently recommends the same simple defence: agree on a shared code word or phrase with close family members, to be used to verify identity in an emergency.
The reasoning is sound. A cloning tool can copy how someone sounds, but it can’t know a phrase that was never spoken or posted in public. Pick something that isn’t on social media and isn’t guessable from your profile, such as a pet’s name you’ve never mentioned online. Tell your family in person or on a trusted channel, and agree that anyone asking for urgent money must say it first.
Pair it with a second rule: hang up and call the person back on a number you already have. If your relative’s phone is genuinely lost, contact them through another family member, a messaging app, or their workplace before sending anything.
Limit the raw material, but don’t rely on it
Guides point out that publicly posted videos and voice recordings are the raw material for cloning. Making social media accounts private and being thoughtful about posting long, clear voice clips reduces what is available. It is worth doing, particularly for children’s accounts.
But it can’t be your main defence. Voice samples also come from calls, voicemails, and videos other people post of you. That is why verification habits, not hiding your voice, are the reliable fix.
The workplace version
The same technique targets workplaces. Authorities warn that fraudsters can impersonate family members or colleagues, and the cloned voice of an executive asking an employee for an urgent payment is a natural extension. Businesses should have a rule that payment requests made by phone, however convincing, need confirmation through a second, independent channel and, for larger sums, a second approver.
If you’ve already sent money
Act quickly:
- Contact your bank or transfer service immediately and ask whether the payment can be stopped or recalled.
- Report it. In the US, the guidance points to ReportFraud.ftc.gov and the FBI’s IC3 at ic3.gov. Elsewhere, use your country’s national cybercrime reporting channel or local police.
- Keep the evidence: phone numbers, payment receipts, and messages.
- Warn your family, because scammers often try again.
- Beware of “recovery” callers. Someone offering to retrieve your money for an upfront fee is typically running a second scam.
Who should worry most
Older relatives are frequent targets, and the numbers cited above single out adults over 60. But anyone with family abroad, anyone with a public voice online, and anyone who handles payments at work should take the risk seriously. If you’re none of those, the family code word is still a five-minute fix worth making.
Frequently Asked Questions
Can scammers really clone a voice from a short clip?
Yes. The FTC says scammers need only a short audio clip, which they can find in content posted online. Quality varies, but a convincing result no longer requires much material.
Can AI voice cloning work during a live call?
According to security guidance, yes. Real-time cloning allows a scammer to hold an interactive conversation in someone else’s voice, which is why a natural-sounding call isn’t proof of identity.
How do I tell if a call is cloned?
You usually can’t tell from the voice alone. The better test is the situation: urgency, secrecy, and unusual payment demands are the strongest signs. Hang up and call the person back on a number you know.
What should my family code word be?
Choose a phrase that isn’t posted online and can’t be guessed from your social media. Share it in person, and agree that anyone asking for urgent money must say it first.
Are the big loss figures I see online accurate?
Treat them carefully. Some come from the FBI’s Internet Crime Report, which began tracking AI-related fraud as its own category, while others appear on statistics blogs with no traceable source. Check the primary report before repeating a number.
Where do I report an AI voice scam?
In the US, report to the FTC at ReportFraud.ftc.gov and the FBI at ic3.gov. In other countries, contact your bank first, then your national cybercrime reporting body or local police.