Skip

Friday, 20 November 2026

How to Spot AI Voice Clone and Deepfake Scams: 10 Ways to Keep Your Family Safe

Priyo News 24 Desk

Published: · 7 min read

How to Spot AI Voice Clone and Deepfake Scams: 10 Ways to Keep Your Family Safe
Illustration of AI voice clone and deepfake scams — Priyo News 24

At a glance

  • AI can now copy a voice from a few seconds of recording, and Bangladesh has already seen arrests in such scams.
  • A crisis story, pressure to act fast and a demand for secrecy together are the main warning signs.
  • Hang up and call back on a number you know, and agree a secret family code word in advance.

Your son's voice is on the line, close to tears: "Dad, the police have arrested me, I need money right now." Then a "police officer" takes the phone and says your son will be released if you send money by bKash. In a panic, many parents pay, only to learn later that their son was safe all along and the voice was a fake made with artificial intelligence (AI).

This is no longer a made-up story. Several Bangladeshi news outlets have recently reported that 2 people were detained or arrested over allegations of cloning a child's voice with AI and posing as police to extract money from the father. In another case, 2 accused were remanded over a deepfake video that copied the face and voice of a well-known religious speaker to defraud people. Knowing how to spot AI voice clone and deepfake scams is now something every family needs.

What AI voice cloning and deepfakes actually are

Voice cloning means using AI to copy someone's tone, pitch and accent and make the copy say new things. With just a short recording from Facebook or YouTube videos, TikTok or WhatsApp voice messages, today's tools can produce a convincing voice.

Deepfakes are fake videos or pictures made with AI in which one person's face and voice are placed onto another. Fraudsters use celebrity deepfakes to spread fake investment, charity or "double your money" offers, and they also pose as relatives on video calls to ask for money.

A cloned voice is hard to catch by ear. In a 2023 experiment by researchers at University College London, participants identified fake speech about 73 percent of the time, meaning roughly one in four went undetected. The US Federal Trade Commission (FTC) also warned in 2023 that scammers were using AI voices in family emergency schemes.

How fraudsters usually set the trap

  1. Collecting a voice: They take the voice of the target's child or relative from public videos or audio on social media.
  2. A crisis story: They talk about an arrest, accident, hospital stay or kidnapping.
  3. Posing as authority: Someone else plays a "police officer", "doctor" or "lawyer" to make it believable.
  4. Pressure and secrecy: "Pay now or a case will be filed", "Don't tell anyone".
  5. Payments that cannot be reversed: They ask for money to a personal bKash or Nagad number, recharge cards or crypto.

AI voice clone and deepfake scams: key facts at a glance

Warning signs of an AI scam

Signs of a cloned voice on a call

  • A call from an unknown number from someone claiming to be family.
  • Very short speech, the same sentence or crying repeated, and questions dodged.
  • A slightly robotic voice with no natural breathing or pauses, and odd emotion.
  • The line keeps dropping because of a "bad network".
  • You are given no time to think and told not to hang up.

Signs of a deepfake video

  • Lip movements do not match the words.
  • Unnatural blinking, blurry edges around the face or hair.
  • Light on the face comes from one side and the background from another.
  • A celebrity offers incredible profits or asks you to click an unknown link.
  • On a video call, the picture breaks up when you ask the person to turn their head or wave a hand in front of their face.

Signs of fake screenshots and images

AI-made pictures often show distorted fingers, meaningless text or uneven jewellery and glasses. Fraudsters also show fake "money sent" screenshots from banks or mobile wallets, so always open your own app and check the balance or transaction list.

Remember that the technology improves quickly; the absence of warning signs does not prove something is real. Verification is your strongest weapon.

10 ways to keep your family safe

  1. Hang up and call back yourself. Phone the person's own saved number, or someone you know who is with them.
  2. Agree a family code word. A word or question known only to family members and written down nowhere.
  3. Ask a personal question. "Where did we go last Eid?" is something a fraudster cannot answer.
  4. Verify police claims. Call your local police station or 999. Police do not ask for money to a personal bKash number.
  5. Never share an OTP or PIN. No bank, mobile financial service or government office asks for these over the phone.
  6. Limit public videos and voice clips. Review your Facebook privacy settings; do not leave your children's videos open to everyone.
  7. Talk to elderly parents. Fraudsters often target older people. Go through the points in this article with them.
  8. Turn on two-factor authentication. Asking relatives for money from a hacked account is another common trick.
  9. Check celebrity offers at the source. See whether the offer appears on their verified page or in established news media.
  10. Speak up without shame. If you stay silent after being cheated, the fraudster moves on to the next victim.

Beware of AI fraud at work and in business

Families are not the only targets. Fraudsters clone the voice of an owner, managing director or major buyer, call someone in accounts and ask for an "urgent" transfer to a number or account. A few rules greatly reduce this risk in any organisation:

  • Never pay on instructions received by phone or voice message alone. Require written approval and a call-back to a number agreed in advance.
  • Require two approvals for large transfers, so that one person under pressure cannot make the mistake alone.
  • Verify sudden requests to change account numbers through an old, known channel.
  • Train your staff. Discuss these warning signs at least once a year.

Teaching children and older relatives

Teenagers post the most videos on social media, and older people tend to send money fastest when a frightening call comes. Talk to both generations separately. Ask children to share less of their voice and family details in public videos. Teach parents to call another family member first whenever an emergency call comes in. Keep a sheet with the family's emergency numbers next to the phone or in the living room, so nobody has to search in a moment of panic. Doing a "practice run" together once makes the real response easier.

What to do if you have been cheated

  • Call the mobile wallet helpline immediately: bKash 16247, Nagad 16167. Give the transaction ID, time and recipient number so action can be taken quickly.
  • Report to the police station or 999: Keep call records, screenshots, the recipient's number and proof of the transaction.
  • Women can seek help through the official Facebook page or hotline of the Police Cyber Support Center for Women.
  • If you see a deepfake video, report it on the platform and warn others, but do not share the video yourself.
Situation Your first step
Call says a child or relative is in trouble Hang up and call a known number yourself
"Police" demand money Verify with the police station or 999
Video of a celebrity investment offer Check the verified source
You have already sent money Wallet helpline and police at once

Final word

AI is making many tasks easier, but it has also handed fraudsters a new weapon. We can no longer trust a voice or a face blindly. When a crisis story, pressure to hurry and a demand for secrecy arrive together, stop, hang up and verify for yourself. Sit down with your family today and agree a code word; those few minutes could prevent a loss of thousands of taka.

FAQ

Can AI really copy someone's voice exactly?

Yes. Voice cloning tools can create a close copy from a short recording taken from social media videos or voice messages. In one experiment by University College London researchers, people correctly identified fake speech only about 73 percent of the time.

What should I do first if a caller says my child is in trouble?

Hang up and call your child's own number, or someone you know who is with them. If the caller mentions the police, verify with the nearest police station or 999. Never send money before checking.

How can I tell a deepfake video?

Be suspicious of lip movements that do not match the words, unnatural blinking, blurry edges around the face, mismatched lighting and unbelievable offers, such as a celebrity promising to double your money. As the technology improves, you cannot always tell by eye, so checking the source matters most.

Where do I complain if I am cheated by an AI scam?

Report it without delay at the nearest police station or to the national emergency service 999. If you sent money by bKash or Nagad, call that provider's helpline immediately with the transaction ID. Women can also seek help from the Police Cyber Support Center for Women.

What is a family code word and how does it help?

It is a secret word or question agreed among family members that only you know. If someone calls claiming to be in trouble, ask for it. A fraudster may copy a voice but will not know the code word.

Sources

This report was written in our own words after checking facts across several reliable sources.

  1. US Federal Trade Commission — Scammers use AI to enhance their family emergency schemes
  2. Mai et al. (arXiv) — Warning: Humans Cannot Reliably Detect Speech Deepfakes
  3. Bangladesh Pratidin — Son's voice cloned with AI to call father posing as police (Bangla)
  4. NTV — Fraud by cloning son's voice with AI, 2 arrested (Bangla)
  5. Risingbd — Money extorted from father by cloning child's voice with AI (Bangla)
  6. Prothom Alo — Fraud using deepfake video, two accused remanded (Bangla)
  7. McAfee — A Guide to Deepfake Scams and AI Voice Spoofing
  8. Digital Literacy Center (Bangladesh) — How to report cyber crime

Comments

Text only — no links allowed · /1000

No comments yet — be the first to comment.

Get the latest news by email

New stories and articles land in your inbox as soon as they are published. Stop any time with one click.