Field note · Everyday AI

AI Voice Cloning Scams in 2026: How to Protect Yourself

Published June 28, 2026 · Empire Publishing

Short answer: AI can now copy a voice from a few seconds of audio, so a caller who sounds exactly like someone you love may not be them. You don't beat this by listening harder — you can't hear the difference. You beat it with one habit: hang up and verify through a channel you control.

What changed

For most of history, a familiar voice was proof. If your daughter called and it was her voice, it was her. That assumption is now broken. Voice-cloning tools can build a convincing imitation from the audio in a voicemail greeting, a social-media clip, or a few seconds of a podcast — material that's public for almost everyone. The technology isn't exotic anymore, and neither is the crime. The raw ingredient for impersonating the people you trust is already out there.

The patterns to know

• The family emergency. A cloned voice — a grandchild, a child — calls in distress: an accident, an arrest, a hospital. They need money right now and beg you not to tell anyone. The voice plus the panic is the whole attack.
• The executive wire. A "boss" or "CFO" calls or video-calls an employee and orders an urgent transfer to a new account, framed as confidential and time-critical.
• The verification spoof. A cloned voice is used to defeat phone-based identity checks, or to add credibility to a follow-up text or email.

Every version runs on the same two levers: a voice you recognize, and urgency engineered to stop you from pausing to check.

The one defense that beats almost all of it

Don't act inside the call. Hang up and call the person back on the number you already have for them. A cloned voice cannot pick up a phone you dial independently. That single move — verify through a separate, trusted channel — defeats nearly every voice-cloning scam, because the entire attack depends on you staying in the conversation it controls.

Two reinforcements make it bulletproof:

• A family safe word. Agree on a private word or question, in advance, that a real emergency call must include. Scammers can clone the voice; they can't know the word.
• A flat rule about money and urgency. Any unexpected, urgent request to move money is treated as suspicious by default — no matter whose voice is asking. Urgency is the scammer's tool, so urgency itself becomes your trigger to slow down.

When it's not just a voice

The same playbook now runs on video. In 2024, an employee at the engineering firm Arup was convinced to transfer roughly $25 million after joining a video call where deepfaked versions of the CFO and colleagues appeared and gave the instruction. Everyone on the call but the victim was synthetic. The lesson scales down to your phone: seeing and hearing are no longer proof on their own. Verification through a channel you trust is. A live face and a familiar voice are now things that can be manufactured — so they stop being the thing you rely on.

If you think you're being targeted

Stop. Don't move money, don't read out codes, don't confirm details. Hang up and call back on a known number. Tell one other person before you act — saying it out loud breaks the spell urgency casts. If money already moved, contact your bank immediately and report it to your local fraud authority. Being targeted isn't a failure; acting without verifying is the only mistake that costs you.

Frequently asked

How much audio does AI need to clone a voice?

Often just seconds to a minute of clear speech — the kind in a voicemail greeting or social clip. The samples to impersonate most people are already public.

What's the single best defense?

Hang up and call back on the number you already have, or use a pre-agreed family safe word. Verify through a channel you control; never act inside the suspicious call.

Can deepfake video really fool people out of money?

Yes — a 2024 deepfake video call cost the firm Arup about $25 million. Treat any urgent money request as suspect regardless of who appears to be asking.

← More field notes