---
title: "A victim in Singapore lost nearly $4m to a deepfake scam"
description: "Criminals are using AI to impersonate officials and public figures, and researchers say the tools are getting easier to use. This is a short report because the source is a short report, and the useful part of it is the mechanism rather than the number."
category: "Technology"
category_url: https://newsparlor.com/category/technology
author: "Priya Sharma"
published: 2026-08-23T13:40:46.000Z
updated: 2026-08-23T13:40:46.000Z
canonical: https://newsparlor.com/article/a-victim-in-singapore-lost-nearly-4m-to-a-deepfake-scam
tags: ["deepfakes", "fraud", "artificial-intelligence", "singapore"]
---
# A victim in Singapore lost nearly $4m to a deepfake scam

Criminals are using AI to impersonate officials and public figures, and researchers say the tools are getting easier to use. This is a short report because the source is a short report, and the useful part of it is the mechanism rather than the number.

One victim in Singapore lost close to $4 million to a deepfake scam, [Al Jazeera reports](https://www.aljazeera.com/video/newsfeed/2026/8/23/ai-fraud-costing-victims-millions-of-dollars) in a short video piece by Patrick Fok. Its caption reads: "Deepfake scams are becoming harder to spot as criminals use AI to impersonate officials and public figures."

Researchers quoted warn that the technology is becoming easier to make.

That is the whole of the source. There is no name for the victim, no date for the fraud, no account of how it was carried out, no police or regulator quoted, and no national or global figure for losses of this kind. We are not filling any of that in.

## What is worth saying anyway

The mechanism is the part that generalises, and it does not require more reporting to explain.

Fraud of this type does not defeat a security system. It defeats a person's judgement about who they are talking to. For most of the history of telephones and video calls, hearing a familiar voice or seeing a familiar face has functioned as authentication, and it worked because faking either convincingly was expensive and slow. Cheap synthesis removes that, and it removes it from a channel that was never designed to be secure in the first place.

Which is why the losses in these cases tend to be large rather than small. The scam is not worth building for a modest sum, so the targets are people or finance staff who can authorise substantial transfers, and the impersonated party is usually a senior figure or an official whose instruction would not normally be questioned.

The countermeasure is unglamorous and does not involve technology: an agreed way of confirming an instruction through a separate channel that the caller did not choose. If the request arrives by video call, the confirmation happens by a phone number you already had.

## What we would want before writing at length

The identity of the impersonated party, the sequence of the fraud, whether the money was recovered, and what Singaporean authorities say about the frequency of these cases. If we can obtain any of that, we will return to it.

## Sources

- [AI fraud costing victims millions of dollars](https://www.aljazeera.com/video/newsfeed/2026/8/23/ai-fraud-costing-victims-millions-of-dollars)