Invest1 publisherNot yet confirmed elsewhere2 min readPublished
Singapore's deepfaked cabinet cost one victim S$4.9m, and the NDA did the real work
The fake Zoom call had mismatched lips and audio that did not line up. It did not matter: the victim had already signed a confidentiality agreement that ruled out asking anyone.
The Investor · Invest desk
What happened
- A business professional in Singapore lost at least S$4.9 million, about US$3.8 million, to a fabricated Zoom call impersonating Prime Minister Lawrence Wong and other senior officials.
- The victim was asked to sign a non-disclosure agreement before the video call, which cast the whole interaction as classified.
- The money went out in several transfers to a corporate bank account the scammers had opened.
- Singapore Police released the call footage on 16 May 2026 and walked the public through the deepfake elements.
Compiled by The InvestorSomething wrong?How this is made
Why it matters
- constraint Any approval process whose identity step ends at "I saw them on the call" now approves whatever the caller can render.
- decision Confidentiality demanded ahead of verification has to become grounds for stopping a payment rather than grounds for keeping quiet about it.
- exposure Anyone with a large published video record has already supplied the material for their own impersonation, and the older archive works fine.
- precedent Training staff to spot lip-sync and audio artefacts buys a control with a short shelf life, since the next generation of tooling removes the tells the police just taught.
Detectability was never the control that failed here. The police release of the call annotates exactly the artefacts a careful viewer would catch, lips out of sync and audio that does not match [8]. Someone briefed to look for those things finds them. Someone who has signed a non-disclosure agreement and believes he is in a classified session about a funding emergency in the Strait of Hormuz is not a careful viewer, and was never intended to be [4][3].
The NDA is the load-bearing part of this fraud, and it was signed before the video call rather than after it [4]. There was no secret to keep. Its function was to void in advance the check that actually defeats impersonation, which is a second person hearing the story and finding it odd. The rest of the approach is dressing: a cabinet secretary's photograph on a WhatsApp profile, a forged letter of guarantee attached to an email sent through Proton [2][3].
The likeness supply is where the economics sit. CNA traced the fabricated officials back to genuine footage from 2022, apparently cut and recombined from separate legitimate sources [9]. That is about four years between raw material and police disclosure [12], and none of it required access to anything that was not already public. Fidelity scales with archive volume, so the people cheapest to counterfeit are the ones with the largest published video record.
Then the payment leg. Money left in multiple transactions to a corporate account the scammers had set up [6], which means at least two separate authorisations of a large sum to a brand-new payee [14]. Each was a chance to compare the instruction against a channel the instructing party did not control, and none of them was. The report frames the case as a stress test for know-your-customer processes, which were not designed for impersonation at this fidelity [15]. KYC is the wrong frame. It screens the party opening an account, and by the report's own account the scammers' shell entity cleared that screen without difficulty [16]. What failed was payment authorisation.
Singapore has been logging deepfake attempts against its own leaders since 2023 [10], roughly three years of notice before this loss became public [13], and the advisories that followed tell citizens to verify unusual official requests through independent channels [11]. That instruction is correct, and it carries a second half that is rarely stated: the verifying channel has to be one the requester never touched, and an instruction that forbids consultation is itself the finding.
What to watch
- Whether the S$4.9m was personal money or funds the victim controlled on an employer's behalf, which decides who absorbs the loss.
- Whether any of the transferred funds are recovered from the corporate account, or had already been moved on.
- Whether the bank that onboarded the scammers' corporate entity is named, and whether onboarding records are made public.
Clarity's read
What the record supports and how the coverage leans. The claims behind it follow.
Reality
- Evidence54
- Adoption61
- Hype gap+14
- Incentives38
- Confidence57
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
A business professional in Singapore lost at least S$4.9 million, roughly US$3.8 million, after scammers used AI-generated deepfake video to impersonate Prime Minister Lawrence Wong and other senior government officials on a fake Zoom call.
- [2]
The operation began with a WhatsApp message from a sender using a profile photo of Cabinet Secretary Wong Hong Kuan.
- [3]
The victim then received a fake email sent via Proton claiming to request urgent funding assistance tied to a purported critical situation in the Strait of Hormuz, accompanied by a forged government letter of guarantee.
- [4]
Before the video call, the victim was asked to sign a non-disclosure agreement, which framed the interaction as classified and discouraged the victim from seeking outside advice or verification.
- [5]
The Zoom meeting was populated by AI-generated likenesses of PM Lawrence Wong, President Tharman Shanmugaratnam, Minister Indranee Rajah and other recognisable Singaporean officials.
- [6]
Multiple transactions were subsequently sent to a corporate bank account set up by the scammers.
- [7]
The Singapore Police Force released footage of the compromised call on May 16, 2026, breaking down the deepfake elements for the public.
- [8]
The released footage revealed noticeable technical flaws, including mismatched lip movements and audio inconsistencies.
- [9]
A subsequent investigation by CNA traced the deepfake material back to real footage from 2022, with clips apparently manipulated or stitched together from different legitimate sources.
- [10]
Deepfake threats targeting Singaporean leaders have been documented since 2023.
- [11]
Singapore authorities have issued heightened public advisories urging citizens to verify any unusual requests from official sources through independent channels.
- [12]
About four years separate the 2022 footage used to build the deepfakes from the police disclosure of the call in May 2026.
- [13]
Roughly three years passed between the first documented deepfake threats against Singaporean leaders in 2023 and the police disclosure of this case in May 2026.
- [14]
Because the funds left in multiple transactions, the payment to the scammers' newly created corporate account was authorised at least twice.
- [15]
The report states that traditional know-your-customer and identity verification methods were not designed to handle AI-generated impersonation at this fidelity.
ReportedInsufficientSource: cryptobriefing.com2 sources— create a free account to open themView cited source - [16]
The report says the corporate account used to receive the stolen funds points to the ease with which shell entities can be established to funnel illicit proceeds.
Sources
1 independent publisher whose own reporting we read for this story.
- cryptobriefing.comAI-generated video of Singapore PM Lawrence Wong used in $3.8M fraud
1 article · August 23, 2026
Topics and entities
Follow any of these and your For You feed starts watching them — no settings page required.