Deepfake video calls: when the "executive" on screen asking you to wire money isn't real
Scammers can now fake a live video call — an executive's face and voice in real time — to order an urgent payment or a change of bank details. The meeting looks real; the person on it isn't. Here is how AI-driven payment fraud works and the free move that stops it.
This scam takes business payment fraud and gives it a face. For years, scammers tricked finance staff into wiring money or changing a supplier’s bank details using a spoofed email from “the CEO”. Now they can add a live video call — an executive’s face and voice, generated by AI in real time — so the request comes with what looks like ironclad proof. The meeting is convincing. The person on it doesn’t exist.
Note: these scams impersonate real, innocent executives and companies. We don’t name anyone — this is about the pattern, not any particular person or business.
How the scam works
The con bolts a deepfake onto an old, expensive crime:
- The setup. Scammers research a target organization and the people in it — names, roles, who approves payments — often from public posts and leaked data.
- The request. An urgent message appears to come from a senior executive or a trusted supplier: wire a payment now, or update a vendor’s bank details for the next invoice.
- The “proof”. When you hesitate, they escalate to a live video or voice call. AI clones the executive’s face and voice for a “real time video chat” that seems to confirm the request is genuine.
- The push. Urgency and secrecy do the rest — “this is time-sensitive”, “keep it between us for now” — so the money moves before anyone verifies through normal channels.
The fact to internalise: a live video or voice call is no longer proof of who you’re talking to. AI has removed “I saw and heard them” from the list of safety checks — for relatives, for celebrities, and now for the boss.
How to spot it
The tells are in the box below, but the cluster is: an urgent money move or bank-detail change + pressure and secrecy + a request that skips your normal approval process. A real payment instruction can wait for one verification call; a scam can’t afford to.
What to do instead
You don’t need to buy anything to beat this — the free move comes first (see the defense box). The habit is Two-Channel Verification: before releasing any payment or changing any bank details, confirm the request through a second, independent channel — call the person on a number you already have, not one from the message or the meeting. Agree that rule in advance so it’s routine. The “relative in a crisis” version of voice cloning is covered in the AI family-emergency call, and the “famous face endorsing an investment” version in deepfake celebrity investment ads. For more, see our defense moves and the 60-second quiz.
- An urgent request to wire money or change a supplier's bank details appears to come from a senior executive or a known contact.
- The "proof" is a live video or voice call — and you're pushed to act fast and keep it confidential.
- The request skips normal process: a brand-new account, a rush, "don't mention this to anyone yet".
- Confirm any payment or bank-detail change through a second, independent channel — call the person back on a number you already have.
- Treat live video and voice as not-proof. AI can fake a real-time face and voice convincingly enough to pass a meeting.
- Slow down urgent, confidential money requests. A genuine one survives a five-minute verification call.
- Agree a verification step for payment requests in advance — a callback rule or a code word — so it's routine, not awkward.
For most of history, a face on a screen and a familiar voice were as close to proof as you could get. That era quietly ended — and the people exploiting it got there first. I don't think the lesson is to distrust every call; it's to move the trust off the call. The thing that can't be deepfaked is a callback to a number you already had and a process you agreed on before the pressure started. Build that habit now, while it's boring, not in the ninety seconds an 'urgent' wire request gives you.
Sources
Frequently asked
Can a live video call really be faked? I saw and heard the person.
Yes. The FBI warns that criminals use generative AI to create video for "real time video chats with alleged company executives, law enforcement, or other authority figures" and to "prove" an online contact is "a real person". A convincing live face and voice is now exactly what this fraud manufactures — seeing and hearing someone is no longer proof of who they are.
This sounds like a corporate problem — is it really common?
It's one of the costliest scams there is. The FBI's IC3 puts business email compromise — fraud that tricks someone into sending a payment to the wrong account — at over $55 billion in reported losses from 2013 to 2023. AI deepfakes are a powerful new way to make the "boss" or "supplier" behind that request look and sound real.
How do I verify without offending my boss or the client?
You don't have to. A callback to a known number before releasing a payment is a standard financial control, not an insult — the FBI's own advice is to "verify the identity of the person calling you by hanging up... and call the phone number directly." A real executive wants that check in place. If a request can't survive a verification call, that's your answer.