How Does an AI Voice Cloning Scam Actually Work?

Short answer: a scammer only needs about three seconds of your voice, pulled from something you've already posted online, to build a convincing fake.

Cloning a voice used to take a recording studio and hours of samples. Not anymore. Researchers at McAfee tested modern voice-cloning tools directly and found that just three seconds of audio was enough to produce a clone with 85% accuracy to the original speaker. Give the tool a little more to work with, and accuracy climbs to 95%.

Where the audio comes from

Scammers don't need to trick you into recording anything special. They just need a few seconds of your voice from somewhere public: a TikTok, an Instagram story, a YouTube comment, a voicemail greeting, or a voice note shared in a family group chat. If your voice exists online anywhere, it's raw material.

How the call actually plays out

Once the clone is ready, the scammer places a call, often to a parent or grandparent, using the cloned voice to claim a sudden emergency: an accident, an arrest, being stranded while traveling. The fake voice asks for money to be sent urgently, usually by wire transfer, gift cards, or crypto, and often asks you not to tell other family members "so we don't worry them."

The scam works because it skips straight past logic and goes for panic. Nobody stops to fact-check a call when they think their child is in trouble.

The one thing that actually stops it

You usually can't tell a clone apart from the real voice anymore, that's the whole point of the technology. But you don't need to. Agree on a family safe word in advance, something ordinary that would never come up naturally in conversation, and never post it online. Anyone calling about money or an emergency has to say it before you act on anything they ask. If they can't, hang up and call that person back on a number you already have saved, not one they give you on the call.

Source: McAfee — Scammers use AI voice cloning tools to fuel new scams