EditorialThreats

A call from a family member in trouble with an AI-cloned voice: how to check

You hear your son, your granddaughter or your partner: they've had a car accident or been arrested and need money now. It could be a voice cloned with AI. Here's how to check in a couple of minutes and what to do.

Summarize this article with

A call from a family member in trouble with an AI-cloned voice: how to check

The phone rings and it's your son, your granddaughter or your partner. They're crying and talking fast: they've had a car accident, they've been arrested, they've been robbed overseas. They need money now and they beg you not to tell anyone. The voice is spot on, and yet it may not be theirs. Artificial intelligence tools can now imitate a voice from just a few seconds of audio, and scammers are using them to give an old trick a new coat of paint: the family member in trouble. The good news is that the defence is simple: hang up and check through a channel you control.

How does the cloned-voice scam work?

The script usually runs like this:

  1. They get a voice sample. A video posted on Instagram, TikTok or Facebook is enough, or a forwarded voice message, or even an earlier call where someone answered "Hello, who's this?" and kept talking for a few seconds.
  2. They gather context. Names, relationships, where the person studies or works, whether they're travelling. A lot of it sits on open social media profiles.
  3. They call with an emergency. The cloned voice says a few emotional lines ("Mum, I've crashed the car", "Pop, I'm in trouble") and quickly hands over to a supposed police officer, lawyer, doctor or "the other driver", who does the negotiating.
  4. They ask for a fast payment that's hard to get back. An instant bank transfer or PayID payment, cryptocurrency, gift cards, or even a courier who turns up at your door to collect cash.

The cloned lines tend to be short, and the call is kept brief on purpose: the longer you talk to the voice, the easier it is to notice something off. That's why the "middleman" usually does the heavy lifting. Sometimes there's no AI at all: a crackly line and some sobbing is enough for your brain to fill in the gaps. The effect is the same.

What gives the scam away?

  • Extreme urgency: you have to pay "right now" or something terrible will happen.
  • Secrecy: "don't tell Dad", "don't call anyone, you'll make it worse".
  • An unknown or private number, or even your relative's own number spoofed: caller ID can be faked.
  • A middleman who takes over as soon as the voice has done its job.
  • Unusual payment methods: crypto, gift cards, a courier collecting cash or jewellery.
  • Dodging personal questions: the voice doesn't answer, the call "drops out" or the crying starts again.
  • Details that don't add up: your relative wasn't travelling, doesn't drive, or is calling at a strange hour.

None of these proves it on its own, but urgency + secrecy + an unusual payment is the signature of this scam, with or without AI.

What do I do during the call?

  1. Take a breath and don't decide anything on the spot. Pressure is the scammer's tool. A real emergency will survive five minutes of checking.
  2. Ask something only your relative would know, ideally not on social media: the first family pet's name, what you ate on Sunday, a childhood nickname.
  3. Ask for the family code word, if you have one (see below).
  4. Hang up and call them yourself on the number already saved in your contacts. Don't use a number the caller gives you and don't ring back the incoming number. If they don't answer, call their partner, housemates, workplace or another relative.
  5. If the caller says they're police, a hospital or a court, check it yourself. Look up the official number independently, or call the Police Assistance Line on 131 444 (in most states). Police in Australia don't ask you to pay bail over the phone or send a courier to pick up money. If someone is in immediate danger, call 000.
  6. Don't send money, codes or bank details until you've confirmed through an independent channel that the emergency is real.

Hanging up will feel heartless if it really is your child. It isn't: if the emergency is real, you'll know within a couple of minutes and can help safely.

How do I set up a family code word?

It's the most effective measure, and it's free:

  1. Choose a word or phrase that's easy to remember and hard to guess. Nothing that appears online: no pet names you've posted, no birthdays.
  2. Share it in person, never by text or in a family group chat.
  3. Agree on the rule: any urgent request for money by phone, voice message or video gets confirmed with the code word. No word, no money.
  4. Include the people who need it most: grandparents, teenagers and people who live alone are favourite targets.
  5. If it's ever used or leaked, change it.

It also helps to give scammers less raw material: private profiles, fewer public videos with children's voices, and not staying on the line with strangers who just want you to keep talking.

What if I've already paid?

Don't blame yourself: this scam is built to fool anyone who loves someone. Act fast:

  1. Call your bank straight away on the number on your card or in its app and ask them to try to stop or recall the payment. Banks such as CommBank have dedicated scam and fraud pages explaining how to reach them. The sooner you call, the more options there are.
  2. If you handed cash or jewellery to a courier, call police as soon as possible with a description of the person and the vehicle: 000 if it's happening now, otherwise 131 444.
  3. Keep the evidence: the incoming number, the time and length of the call, payment receipts, any message or audio.
  4. Report it to police through ReportCyber, via the AFP's cybercrime page, and to Scamwatch. Mention that they used a voice imitating your relative.
  5. Get support. IDCARE (1800 595 160) offers free help working out what else to secure.
  6. Tell your family the scam is going around so nobody else falls for it.

More on what to do after a scam is on our blog.

Frequently asked questions

Can AI really copy a voice from so little audio? Yes. Current tools can produce a convincing imitation from short samples, especially for short, emotional lines over the phone, where call quality hides the flaws. It doesn't need to be perfect: just believable for a few seconds of panic.

The call came from my son's number. So it was him? Not necessarily. The number on your screen can be faked. That's why the rule is to hang up and call the number in your contacts yourself.

Can this happen on WhatsApp too, with voice messages or video calls? Yes. Cloned voice messages and doctored videos follow the same logic. The protocol doesn't change: a check question, the code word and verification through another channel.

What if it really is an emergency? Then checking takes a couple of minutes and costs you nothing. Hospitals and police don't demand immediate payment over the phone to help anyone.

In short

A familiar voice is no longer proof of who's calling. When a relative in trouble rings asking for money: don't decide on the spot, ask a check question, ask for the code word, hang up and call the number you already have. If you've already paid, call your bank, report it through ReportCyber and Scamwatch, and contact IDCARE. And if you get one of these calls, report the number in the NoCall spam numbers directory: your warning could help another family hang up in time.

N
By NoCall team
NoCall Editorial
6 min read
#AI voice scam#cloned voice#voice deepfake#family emergency scam#family code word#vishing

Sources and review

Reviewed:

Numbering, carrier tools, and reporting procedures can change. Verify sensitive steps with the linked primary authority.

Article details

Editorial content reviewed by NoCall with practical context for spotting suspicious calls and messages.

Author: NoCall team6 min read

Received a suspicious call?

Look up the number in NoCall before sharing data, calling back, or clicking any link.

Search a phone number or a company name (Telstra, Telstra and Optus...) to check if it has been reported as spam.