How AI voice-cloning scams work, and how to protect your family
Scammers can now copy a relative's voice from a few seconds of audio and use it to fake an emergency. Recent cases range from family scams in India to a $129 million fraud against US companies.
A businesswoman in Kolkata lost 1.2 lakh rupees this month after answering what she believed was an emergency call. The voice on the line had been produced with artificial intelligence.
Her case is small next to some of the losses now being reported, but it shows how an old trick has been rebuilt. The fake emergency, a call from a relative who has been in an accident or arrested and needs money at once, has been around for decades. What AI changes is the voice. A caller no longer has to rely on a bad line and a panicked listener to cover the gaps. They can sound exactly like the person they claim to be.
How the voices are made
Voice-cloning software can produce a convincing imitation of a person from a short recording. That audio is often easy to find: videos posted to social media, podcasts, voicemail greetings, or a brief phone call in which the target says little more than hello. The cloned voice can then read any script the scammer writes, either live or as a recording.
It is usually paired with other techniques. Caller ID can be spoofed so the call appears to come from a familiar number, and the script is built to create pressure, with an urgent problem, a demand for secrecy and an unusual way to pay, such as gift cards, a wire transfer or cryptocurrency. The voice gets past the listener’s first instinct. The pressure keeps them from stopping to check.
Families and companies are both targets
Police and local officials have begun warning people directly. In early November, officials in Los Angeles County warned residents about AI voice-cloning scams, and newspapers have been running advice for families ahead of Thanksgiving, when relatives are often travelling and harder to reach.
Businesses face a version aimed at their payment processes. This month, US prosecutors charged three men over a scheme in which they are accused of using AI to impersonate company executives and lawyers, taking about $129 million from businesses. The Record reported that the scammers used AI to help steal more than $100 million by posing as chief executives.
Officials who track financial crime expect the problem to grow. The head of the Financial Action Task Force, the international body that sets standards against money laundering, said this month that AI and faster payment systems mean the worst of fraud and online scams may be “still to come”. Telecom companies are responding with their own filters. Virgin Media O2 says its AI-based system in the UK now blocks about three million scam calls a week.
What to watch for
Because a cloned voice can sound flawless, listening for glitches is not a reliable defense. The circumstances of the call are a better guide. Be suspicious of any call that combines an emergency with a demand to act immediately, a request to keep it from other family members, or a payment method that is hard to reverse. A caller who explains that they are using a friend’s phone, or who sounds right but turns vague when asked about details, is another warning sign.
Simple steps that work
Most of these scams fail when the person receiving the call takes a minute to check. Consumer and law enforcement agencies commonly recommend a few habits:
- Agree on a family code word that a caller has to say in an emergency, and make sure older relatives and teenagers know it.
- Hang up and call the person back on a number you already have saved, not the number that called you.
- Contact another family member before sending money.
- Keep voice recordings of yourself private where you can, and use a generic voicemail greeting.
- Talk to older relatives about this scam specifically, since they are frequent targets.
Anyone who has been targeted in the US can report it to the Federal Trade Commission at reportfraud.ftc.gov and to the FBI’s Internet Crime Complaint Center at ic3.gov. If money has already been sent, the bank should be called immediately, as transfers can sometimes be stopped if they are flagged quickly.
For businesses, the $129 million case is a reminder that a familiar voice, or even a familiar face on a video call, is no longer proof of identity. Companies that have avoided losses typically refuse to change payment details or approve transfers on the strength of a call alone, verify requests through a separate channel, and require a second person to sign off on large payments.
The technology behind these scams is improving faster than the tools designed to detect it, so the most dependable protection is likely to stay a human one: a pause, a second phone call and a word only the family knows. New cases are logged on our AI scams and fraud page.
Sources
Show all 7 sources
Introduction
- News Mobile: AI Voice Cloning Scam: Kolkata Businesswoman Duped Of Rs 1.2 Lakh By Fake Emergency Call
Families and companies are both targets
- KTLA: Los Angeles County warns residents about AI voice-cloning scams
- Omaha World-Herald: What families can do about AI voice-cloning scams this Thanksgiving - Omaha World-Herald
- PYMNTS.com: FBI Charges Three in $129 Million AI Impersonation Scam
- The Record: CEO-spoofing scammers used AI to help steal more than $100 million, feds say
- AML Intelligence: NEWS: AI, digitalisation and better connected payments mean ‘worst still to come’ from fraud and online scams – FATF chief
- Virgin Media O2: O2 upgrades AI scam defences to block 3 million scam calls a week