Can Scammers Use AI to Copy My Family's Voice?
Yes. And the uncomfortable part is not that it is possible — it is that you probably will not be able to tell by listening. Which means the usual advice, "be careful, listen closely," is the wrong advice.
The call arrives at a bad hour. It is your daughter, or your brother, or your mother. The voice is right. They are crying, or frightened, or speaking very fast. There has been an accident, or an arrest, or a hospital. They need money now, and they ask you not to tell anyone else in the family.
Everything about that call is engineered — including, now, the voice.
How much audio does AI need to clone a voice?
Less than most people assume. The companies selling voice-cloning tools advertise that around three seconds of clear speech is enough to produce a recognisable copy, with longer samples giving better results.
Three seconds is a voicemail greeting. It is the first sentence of a video. It is someone answering the phone and saying hello twice before hanging up.
This is worth sitting with, because it changes the shape of the problem: there is no realistic way to keep your voice private. The defence cannot be secrecy.
Where would a scammer get a recording of my voice?
Nowhere exotic. No hacking is involved.
- Any video you or a family member has posted publicly.
- Your outgoing voicemail greeting — which anyone can hear by calling and waiting.
- A podcast, webinar, recorded meeting, church service, or work presentation.
- A nuisance call where you spoke for a few seconds before hanging up.
Note that the person whose voice gets copied is often not the person who gets called. A short clip of an adult child speaks to their parent; a clip of a manager speaks to their staff. So "I do not post videos" is not protection if anyone connected to you does.
How can I tell if a voice on the phone is AI?
Increasingly, you cannot — and building your safety around detection is a mistake.
The old tells are fading. A few years ago cloned speech had flat emotion, slightly wrong pauses, a faint metallic edge. Current tools handle emotion, breathing and hesitation convincingly, and the scam is deliberately staged to hide whatever remains: a bad connection, background noise, someone crying, a rushed call that lasts ninety seconds.
Worse, the situation is designed to switch off exactly the judgement you would need. Fear narrows attention. You are not calmly assessing audio quality when you believe someone you love is in trouble — you are trying to help.
So the honest answer is: stop trying to hear the difference. Verify instead.
What should I do if I get a call like this?
Hang up. Then call back on the number you already have.
Not the number they give you. Not by pressing redial. The number already saved in your phone.
This is also the FTC's advice: call the person who supposedly contacted you using a phone number you know is theirs, and if you cannot reach them, try another family member or a friend.
Three more things worth knowing, because scammers rely on all three going unnoticed:
- Urgency is the tell, not the voice. Real emergencies survive a five-minute pause. A script that insists you cannot wait, cannot hang up, and cannot tell anyone is not describing an emergency — it is preventing verification.
- "Do not tell Mum" is part of the attack. Genuine trouble rarely requires isolating you from the rest of your family. This one instruction is doing an enormous amount of work for them.
- Gift cards, wire transfers, cryptocurrency and cash couriers are never legitimate. No court, hospital, lawyer or police force has ever asked to be paid this way. If the payment method is unusual, the story is false — regardless of how the voice sounded.
What is a family code word, and how do I set one up?
A code word is a short phrase your family agrees on in advance, which anyone calling in a genuine emergency can say. It costs nothing, takes one conversation, and it works precisely because it does not depend on your judgement in the moment.
- Choose a phrase that exists nowhere online. Not a pet's name, a street, a birthday, a school, or anything findable on social media. Something arbitrary and slightly odd works best — an unlikely pair of words is easier to remember than a random string.
- Share it in person, or by a channel you are sure of. Not on social media, obviously, but also not in a group chat with people outside the family.
- Do not store it in a note called "code word". If it needs writing down, disguise what it is.
- Agree what happens if it is missing. This is the step families skip. The rule should be simple and absolute: no code word means the call ends and you dial back. Anyone genuine will understand.
- Tell the youngest and oldest members too. A plan only two people know is not a plan.
If someone real forgets the word under stress, nothing is lost — you hang up and reach them on their own number, and the emergency is still there thirty seconds later.
Does saying "yes" on the phone let scammers steal from me?
This one deserves a direct answer, because it circulates constantly and it is not supported by evidence.
The warning says that answering "yes" to a caller asking "can you hear me?" lets them record your voice and use it to authorise charges. Snopes investigated and found no one who had actually lost money this way, and no plausible mechanism by which a recorded "yes" alone could authorise anything — a scammer able to charge your account would already need your personal and account details, which is the real theft.
Hanging up on unknown callers remains sensible. But the fear is pointed at the wrong thing, and that has a cost: people guard a single word while leaving the actual attack — a convincing call demanding urgent payment — completely unprepared for.
What if I already sent money?
Act quickly and without embarrassment. These scams work on careful, intelligent people; falling for one is evidence of a well-built trap, not of poor judgement.
- Contact your bank or the payment service immediately and say the payment was fraud. Speed matters more than anything else here — some transfers can still be stopped or recalled within hours.
- For gift cards, call the card issuer (the number on the back) and report it. Funds are occasionally frozen if the cards have not been drained yet.
- Report it to the FTC at ReportFraud.ftc.gov, and to local police if you want a record for your bank or insurer.
- Tell the family member whose voice was used. Their voice is likely to be used again, on someone else.
The short version
Yes, it is real. About three seconds of audio is enough, and it comes from ordinary public sources.
Do not rely on your ears. The differences are disappearing, and fear is exactly when your judgement is worst.
Hang up and call back on a number you already have. This single habit defeats the entire scam.
Agree a family code word this week — and agree what happens when it is missing.
Unusual payment methods mean it is false, whatever the voice sounded like.
Nothing here requires understanding how the technology works. It requires one habit, agreed in advance, while everyone is calm.
The same trick with images: why the usual advice no longer works →
The habit protects you once. Understanding protects you next time.
A code word stops this particular call. What stops the next one — the version nobody has written an article about yet — is having a rough picture in your head of what these tools can actually do. The people easiest to fool are the ones with no picture at all.
That is what the cards are for. Getting Comfortable with AI is twenty cards that go from "what is this thing" to "here is where it gets things wrong" — one clear idea and one useful thing to do on each. The whole set takes about 15 minutes to read — it is a set of cards, not a course. No jargon, no assumed background.
Free, and it stays free. 15 minutes — less time than it took to read this page.
Next: Is ChatGPT safe to use? · What is a prompt? · Explore AI for Beginners · Return to Yilume
Sources. Federal Trade Commission, Family Emergency Scams and Fighting back against harmful voice cloning, for the call-back guidance and reporting route. Snopes, "Can You Hear Me?" scam warning, for the absence of evidence behind the "say yes" warning. The three-second figure is drawn from the marketing claims of commercial voice-cloning services and should be read as their own advertised capability. Current as of August 2026.