Deepfake
What is Deepfake?
A deepfake is a fake image, voice or video made by AI that looks like a real person or event, often hard to tell apart.
A cloned voice of a relative asking for money, a fake photo of damaged goods to claim a refund, a celebrity in an ad they never made: deepfakes are now cheap and fast. The best defence is to check through another channel rather than trust what you see or hear.
Deepfake in a shop
A customer sends a photo of broken goods for a refund: ask for an unbroken video or a photo from another angle first.
How does Deepfake work?
- 1An AI tool learns from photos, video or a few seconds of someone's voice.
- 2It produces new images, audio or video that look and sound like that person, or like a real event.
- 3Scammers use them for calls, messages or ads, usually with pressure to act fast.
- 4Deepfakes are now cheap and quick to make, so seeing and hearing is no longer enough to believe.
What a shop or café can use it for
- Knowing the scams aimed at shops: fake damage photos for refunds, fake calls from people you know asking for money.
- Setting a verification routine: ask for an uncut video, call back a saved number.
- For AI content the shop makes itself (AI models, AI voices): label it clearly.
What to watch out for
- Never send money because of a video call or voice message alone, however familiar it looks.
- Do not use a real person's face or voice without consent: Vietnam's AI Law requires labels and bans using AI to deceive.
- Deepfake detectors are unreliable; checking through another channel is still the dependable way.
Quick questions
A very short call, a face with little emotion, head and body out of step, sound out of sync, then pressure to transfer money to an unfamiliar account. With a few of these signs, hang up and call back a saved number.
Ask for an uncut unboxing video or a photo taken now from another angle, and match the order code. Real customers can usually send it at once.
Not a scam if you label them as AI-generated and they show the real product accurately. They become a problem when they use a real person's face without permission or mislead customers.