What a voice-clone scam is

A voice-clone scam is a call or voice message where the speaker sounds like someone you know, but the voice was made with software. The aim is usually money or private details. You don't need to understand the technology to handle it. You need to know what it looks like, and what to do next.

What "cloned" means

Software can study recordings of a person speaking and then produce new speech that sounds like them. The result is a kind of deepfake. Wikipedia describes deepfakes as images, videos or audio that have been edited or generated using artificial intelligence.¹

How much recording does it take? Wikipedia says that by 2023 the technology could clone a voice from a recording of a few seconds to a minute.¹ That's a short clip. It's also why you shouldn't assume only famous people can be copied.

The FBI says criminals can use AI-generated audio to pass themselves off as well-known public figures or as people close to the victim, in order to get payments.²

What it can look like at work

Picture the call. Your manager's voice comes down the line, a little rushed, and asks you to pay a supplier today. It sounds right. The tone is right. That's the whole trick.

One widely reported case is from 2019. Wikipedia says a UK-based energy firm's chief executive was scammed over the phone and ordered to transfer €220,000 into a Hungarian bank account. The caller reportedly used audio deepfake technology to imitate the voice of the head of the firm's parent company.¹

Wikipedia words this as "reportedly", and so should you. Details of individual cases are often hard to confirm.

How it fits with older phone scams

Voice phishing, also called vishing, is the use of telephone calls to carry out phishing attacks.³ Phishing just means tricking someone into handing over money or information.

Vishing already had tricks of its own. One is caller ID spoofing, where the caller changes the name or number that shows on your phone so it differs from the real one.³ So a familiar name on the screen proves nothing either.

A cloned voice adds a second false signal to the first. Wikipedia notes that audio deepfakes have been used to fool people into thinking they're getting instructions from a trusted person.³

Why careful people still get caught

Most of us trust a voice we know. We don't stop to question it, and scammers count on that.

The FBI also points out that generative AI tools can correct human errors that might otherwise be warning signs of fraud.² Bad grammar or an odd accent used to give some scams away. That can no longer be relied on.

The same FBI notice says it can be difficult to tell when content is AI-generated.² So the safest habit is not to try to judge the voice at all. Check who's calling by another route instead.

That's the idea behind the rest of this site:

• What should make you pause is on the warning signs page.
• Confirming a caller is the job of the call-back routine.
• Official agencies that take reports are on the reporting page.

If you've already acted on a call like this, don't blame yourself. The point of the technology is to be convincing. Reporting early is the useful step.