Deepfake Scams: Only 1 in 5 in Hong Kong Trust Banks to Stop Them

Picture your phone lighting up at 10:55 at night. The caller says they're from your bank. Your account is under attack, and they need you to act now. The voice is calm, professional, and exactly what you'd expect. It could also be a deepfake, a fake voice or video made by AI to sound and look like a real person.

Deepfake scams work because a convincing voice feels like proof, and only about 1 in 5 people in Hong Kong trust their banks to stop them, so the safest move is to hang up and call your bank back on a number you already know is real.

Deepfake Scams and Hong Kong's Trust Problem: What the Survey Found

Let's start with the number that should make everyone sit up. According to Asian Banking & Finance, only about 23% of Hong Kong consumers trust their banks to stop a deepfake scam before money leaves their accounts. The headline rounds that to "1 in 5." The same survey data shows that 41% of people reported being targeted by AI impersonation.

Read that again. About four in ten say someone has already tried to fool them with an AI fake. Most people don't believe their bank will catch the next attempt. Honestly? They aren't being paranoid. They're doing the math. This article is part of a series, start with Uk Shop Face Recognition Innocent Shoppers Named As Thieves.

Why a deepfake voice feels so believable

Here's the quiet trick. Your brain trusts a voice it recognizes, or one that simply sounds official. That shortcut served humans well for thousands of years. Now an AI voice clone (software built on neural networks that copies how a real person talks from any audio sample, even a short one) can fake it in minutes.

Consumer tools make this easy: with a deepfake app or a deepfake generator, a scammer can just upload a clip and get a face swap back. That is deepfake technology at its most accessible.

There's a second trick, and it's about memory. When a scary story is fresh in your mind, you assume it's common. When it's not, you assume it won't happen to you. Psychologists call this the availability heuristic (your mind judges how likely something is by how easily an example pops into your head). Most people have never heard a deepfake scam call in person, so it doesn't spring to mind when the phone rings. Scammers count on that blind spot.

---

Can a Deepfake Really Fool a Bank? Real Cases From Hong Kong and Italy

Yes. A deepfake has already fooled professionals who were trained to be careful. In 2024, an employee at the engineering firm Arup in Hong Kong sent about $25 million to fraudsters after a video call. The "executives" on the call were deepfake copies, according to Yahoo Finance. Every face on screen looked familiar. None of it was real.

The problem isn't limited to Asia. In February 2026, scammers cloned the voice of a trusted lawyer and used a fake WhatsApp message and phone call to deceive a bank president in Italy. The reported transfers totaled €95 million, and €36 million still hasn't been recovered, according to OECD.AI. If a bank president can be fooled, an ordinary customer at 11pm doesn't stand a chance on instinct alone.

And the volume is climbing. Infosec Federation reports that synthetic voice attacks on banks (calls using audio deepfakes, meaning computer-made voices) rose 20-fold in one year. That's not a trend line. That's a wall.

What banks are doing about the deepfake video problem

Banks aren't sitting still. In 2025, Hong Kong's banking regulator, the Hong Kong Monetary Authority, added a "D" for deepfake detection to its e-banking security framework. Banks that adopted the related controls reportedly improved their ability to spot live deepfake attacks by more than 95%. Meanwhile, Biometric Update reports that Hong Kong banks are testing AI detection technology in a regulator-run test space.

Much of that defense relies on liveness detection (a check that the camera is looking at a live person, not a photo, a recording, or an artificial image made by AI). It's smart work. But notice the gap. Those tools protect the bank's front door. They can't stop you from picking up your phone and being talked into opening it yourself.

---

Hong Kong Deepfake Scams: Why Voice and Video Are Context, Not Proof

This is the part to screenshot. Fraud specialists have landed on a simple rule. Treat a voice or a video as a clue, never as the final answer. A caller who sounds like your banker doesn't prove your banker is calling. Hong Kong deepfake scams succeed precisely because people treat a face or a voice as an ID card. It isn't one anymore. Previously in this series: Face Recognition Iphone Beats Masks Still Opens For Twins.

Voice and video should be treated as "context, not proof" when you're verifying a payment request.

Verification guidance summarized by Netsurit

So what actually proves a call is real? Not how it sounds. What proves it is a channel you control. Compare the two approaches side by side.

What a deepfake caller gives you	What a real check gives you
A familiar voice that deepfakes can clone	A phone number you already trust, from your card or your bank's official site
Urgency: "act now or lose your money"	Time to hang up and verify, with no penalty for pausing
A deepfake video that looks like a boss or banker	A separate confirmation through a second, known channel
A request to move money immediately	A call back to the bank, which can see your real account

What to do when a deepfake call about your account arrives

Here's the whole playbook, and it fits on a sticky note.

First, don't argue and don't explain. Second, say you'll call back, then end the call. Third, dial the number on the back of your card or on your bank's official website. Never use a number the caller gave you, and never press "call back" on a message they sent. If the call was real, your bank will happily confirm. If it was a deepfake, you just broke the spell.

A fake caller needs you to stay on the line. A real bank doesn't mind if you hang up.

Is It a Deepfake? How to Tell If a Familiar Voice Is Real

Honest answer: often you can't tell by listening. Today's fake voices can be smooth, warm, and free of the robotic glitches people expect. So the better question isn't "does this sound fake?" but "did I start this contact, on a channel I trust?" If you didn't, treat the call as unverified, no matter how real it sounds.

If you've ever wondered whether a photo, a video, or a profile is really who it claims to be, that's the exact question facial comparison technology exists to answer. Investigators use it after a fraud to line up an authentic face against suspect footage, side by side, and see whether they match. Notice what that tells you. Even the professionals don't trust their eyes alone. They compare against a known, genuine sample. Your version of that is the number on your card. Up next: Deepfake Scams 1 In 5 In Hong Kong Trust Banks To Stop Them.

Here's what sticks with me. The Arup employee sat on a video call full of faces they recognized, and every one of them was fake. If a screen full of familiar colleagues isn't proof, what exactly were you planning to use your ears for?

Deepfake: Frequently Asked Questions