Social Engineering & FraudMANIPULATIVE

Voice-Cloning Scam

What it is

A fraud that uses AI-synthesized audio of a real person's voice, built from a short sample, to impersonate a relative, executive, or official and trigger urgent payments or disclosures.

How it works

The human brain treats a familiar voice as strong identity proof, a shortcut that held for a lifetime until generative audio broke it. Modern tools can approximate a specific person's voice from a few seconds of audio scraped from social media, a voicemail greeting, or a video, and vendors and researchers report clones convincing enough that most listeners cannot reliably tell them apart. The clone supplies the missing ingredient that made older impersonation scams stumble: it sounds exactly like the person it claims to be. Layered onto existing scripts (the grandchild in jail, the CEO ordering a wire, the executive on a bad line), the cloned voice removes the doubt a stranger's voice would raise and compresses the victim's decision under manufactured urgency. McAfee's 2023 survey found a quarter of adults had encountered some form of AI voice scam and that most targeted victims who engaged lost money, and the FTC has warned specifically that voice cloning is turbocharging family-emergency scams.

Real-world examples

•A parent receives a call in their child's exact voice, sobbing about an accident and a lawyer who needs bail money immediately; the voice was cloned from the child's public videos.
•In 2019 a UK energy firm reportedly wired about 240,000 dollars after an executive believed he recognized his German parent-company CEO's voice on the phone, an early audio-deepfake fraud.
•In 2024 the engineering firm Arup lost about 25 million dollars in Hong Kong when an employee was convinced by a video call populated with deepfaked colleagues to authorize transfers.
•McAfee's 2023 research reported that a large share of people could not distinguish a cloned voice from the real one, and that many who received such calls lost money.

Ethical guidelines

●Cloning a real person's voice to deceive and defraud is fraud and impersonation; there is no honest deployment of a nonconsensual voice clone for this purpose.
●Synthetic-media tools have legitimate uses (accessibility, dubbing, consented performance), but using someone's voice without consent to extract money or data is abuse of the technology.
●Providers of voice-cloning tools bear responsibility for consent checks, provenance watermarking, and abuse prevention.

How to defend against it

►Do not treat a familiar voice as proof of identity; hang up and call the person back on their known number before acting on any urgent request.
►Agree a family or team "safe word" or a shared question a clone could not answer, and use it whenever a call demands money or secrecy.
►Be most skeptical exactly when the voice is perfect and the situation is urgent; realism plus pressure is now the signature of the attack, not reassurance.
►Limit the voice you make public where practical, and set social accounts so strangers cannot harvest long clips of you or your children.
►Verify any executive payment request through an out-of-band channel and a second approver, regardless of how convincing the voice or video appears.

From the Defense Playbook

Out-of-Band Verificationminutes

Confirm any request that arrives through one channel (email, text, chat, a phone call) by reaching the supposed sender through a different channel you already trusted before the request existed.

The Callback Ruleminutes

When anyone contacts you claiming to be your bank, a government agency, a utility, or a company you deal with, hang up and call back on the number printed on your card, statement, or the organization's official website, never the number they gave you.

Gift Cards Mean Scamseconds

Treat any demand to pay a debt, fine, fee, bail, or "security deposit" with gift cards, cryptocurrency, a wire to a stranger, a payment app, or cash handed to a courier as proof of fraud, because no legitimate institution collects money that way.

Tell Someone Before You Send Moneyminutes

Before sending money or sharing account details in response to any unexpected request, describe the situation out loud to one person who is not involved, because scams depend on the target deciding alone.

Every playbook entry states how strong its evidence is and when not to use it. Browse the full playbook.

References

McAfee (2023). Beware the Artificial Impostor. McAfee Labs (survey of 7,054 people across seven countries) · link
Survey finding widespread encounters with AI voice scams and that most engaged victims lost money.
Puig, A. (Federal Trade Commission) (2023). Scammers use AI to enhance their family emergency schemes. FTC Consumer Advice · link
Official warning that voice cloning from short clips makes family-emergency impersonation more convincing.

Last reviewed 2026-09-07

Suggest a correction

Detect Voice-Cloning Scam in any text

Paste any message, email, or article into our free Manipulation Detector to see if Voice-Cloning Scam or other techniques are being used on you.

Try the Detector Free Create Free Account

Social Engineering: Why Smart People Fall for It

Social engineering does not target stupidity. It targets helpfulness, deference, and time pressure, which are the same habits that make people good at their jobs. Here is how the attacks are built and which procedures actually stop them.

7 min read

The Disinformation Playbook: How Campaigns Are Built and Where They Break

Disinformation campaigns follow a recognizable sequence: seed, launder, amplify, and let real people carry it the rest of the way. A field guide to the stages, drawn from the research, with the defences that work at each one.

8 min read