Deepfake Scams: Protect Your Business from AI Fraud
Criminals now clone voices and faces to trick staff into paying. How deepfake scams work and the simple checks that stop them.

A finance manager receives a video call from what looks and sounds exactly like the managing director. The face is right, the voice is right, the mannerisms are right. The director asks them to set up an urgent payment to a new supplier account. Everything about the call looks genuine. Everything about it is fake.
This is a deepfake scam, and it is no longer science fiction. AI tools can now clone a voice from a few seconds of audio and generate a convincing video of someone's face from a handful of photos. For a business, the risk is real and the stakes are high. This guide explains how these scams work, the warning signs, and the simple verification rules that stop them in their tracks.
How deepfake scams work
A deepfake is a piece of media, audio or video, that has been generated or altered by AI to impersonate a real person. The technology behind it has improved rapidly. Voice cloning only needs a short sample of someone speaking, which is easy to find from a public presentation, a podcast, or even a voicemail greeting. Video cloning uses photos and short clips, which most executives have on company websites and LinkedIn.
Criminals use this to impersonate people in authority. A cloned voice on a phone call asking a junior member of staff to buy gift vouchers. A fake video call with a supplier asking to confirm a change of bank details. A voice message from the chief executive asking the finance team to rush through a payment. The common thread is urgency and authority, the two ingredients that make people act before they check.
Warning signs to watch for
Deepfake scams share a pattern with traditional business email compromise, but they feel more convincing because the impersonation is so realistic. The warning signs are usually in the request, not the medium.
- Urgency. The request needs to happen now, today, before the end of the day. Pressure is the scammer's main tool.
- Secrecy. You are asked not to discuss it with anyone, or told the boss is in a meeting and cannot be disturbed.
- A change to payment details. A new bank account, a new supplier, a change to an existing arrangement.
- An unusual channel. The request comes by video call, voice message or personal phone rather than the normal email trail.
- Slightly off. The voice or face is almost right but something feels strange, a pause, a glitch, an odd phrasing.
Verification rules that stop deepfakes
The good news is that deepfake scams are stopped by the same controls that stop traditional fraud. The technology has changed, but the defence is simple, human, and effective.
Call back on a known number
If you receive an unusual request by voice or video, hang up and call the person back on a number you already have on file. Not the number they called from, not a number they gave you in the call. A number from your own contacts or directory. A cloned voice cannot answer a phone you dialled yourself.
Use code words
Agree a small set of code words or challenge phrases with senior staff and the finance team. If a request feels off, ask for the code word. A deepfake cannot know a phrase that was only ever shared in person or on a secure call.
Two-person approval for payments
No payment above an agreed threshold should be authorised by one person alone. A second pair of eyes, ideally someone who can verify the request independently, is one of the most effective fraud controls a business can have. Scammers rely on a single person acting in isolation; two-person approval removes that opportunity entirely.
Verify payment changes out of band
Any change to supplier bank details should be verified through a known channel, a phone call to a known contact, not a reply to the email or message that requested the change. This rule alone stops a large proportion of fraud attempts, deepfake or otherwise.
Train your team
Technology alone will not stop a convincing deepfake. Your team is the last and most important line of defence, and they need to know these scams exist. Short, regular training that covers what deepfakes are, how the requests typically sound, and the verification rules above will give people the confidence to pause and check rather than act on impulse.
The message is simple: if a request is unusual, urgent, and involves money, it is worth a thirty-second check. No legitimate boss or supplier will mind you verifying a payment. A scammer will mind very much.
FAQ
What is a deepfake scam? A scam where AI is used to clone someone's voice or face to impersonate them, usually to trick staff into making a payment or sharing sensitive information.
How do I protect my business from deepfakes? Use call-back verification on known numbers, code words, two-person approval for payments, and out-of-band checks for any change to bank details. Train your team so everyone knows the rules.
Can AI detect deepfakes? Detection tools exist and are improving, but they are not reliable enough to be your only defence. Human verification rules are still the most effective control.
Want to train your team to spot these scams? Explore our security awareness training, or read more about the wider risk of shadow AI in your business.


