The Arup Deepfake Scam — HK$200 Million Lost After a Video Call With Fake Executives

An employee in the Hong Kong office of a multinational company paid out about HK$200 million — roughly US$25 million — after a video conference in which every other participant was a digital fake. Hong Kong police described the case at a briefing on Sunday 4 February 2024 without naming the company. In May 2024 the British engineering firm Arup confirmed to CNN that it was the victim, and that "fake voices and images were used".

The case became a reference point because it defeated the control that companies had been told to rely on. Email requests for urgent transfers were a known fraud pattern; seeing and hearing your chief financial officer and colleagues on a call was supposed to be the check. Here the call was the attack.

What happened

Acting Senior Superintendent Baron Chan of the Hong Kong police set out the sequence, as reported by AFP via Hong Kong Free Press. The employee, who worked in the finance department, joined a video conference with several people who appeared to be senior officers of the company; according to Hong Kong media reports, the fraudsters posed as the firm's UK-based chief financial officer. The police said all the participants except the victim were impersonated, using publicly available video and audio of the real executives found on YouTube.

The money went out in 15 transfers. The police received the report on 29 January 2024, by which time the funds were gone; at the February briefing they said no arrest had been made.

How the deception worked

The government's account to the Legislative Council on 26 June 2024 adds a detail that changes how the attack should be understood. The Secretary for Security, Tang Ping-keung, wrote that the fraudster, posing as the UK CFO, had first sent a phishing email and then invited the employee to a group video conference, and that "Since the meeting was pre-recorded, there was actually no interaction between the informant and the fraudster." After the employee authorised transfers to five local bank accounts, the fraudster kept giving instructions by instant messaging.

In other words, the attackers did not need real-time deepfake conversation. Hong Kong's computer emergency response team, HKCERT, wrote in its alert that the fraudsters ended the meetings quickly on various pretexts so the victim could not ask questions. A short, pre-recorded performance by convincing faces was enough to authorise a transfer that ordinary process would have queried.

HKCERT's advice reads like a new etiquette for remote work: verify invitations and links, ask questions only the real person would know, ask the other party to nod, wave, cover their face or move the camera, and watch for mismatches between image and voice. The police added a test of their own — ask the person on screen to turn their head — and urged the public to confirm any request by calling back.

Arup's statement

Arup, a British engineering and design firm, confirmed its identity to CNN in May. A spokesperson said the company had notified Hong Kong police in January, could not go into details because of the ongoing investigation, and that "Our financial stability and business operations were not affected and none of our internal systems were compromised." Rob Greig, Arup's global chief information officer, said the firm was subject to regular attacks including invoice fraud, phishing scams, WhatsApp voice spoofing and deepfakes, and that "the number and sophistication of these attacks has been rising sharply in recent months".

The point about systems matters. Nothing was hacked in the conventional sense: no malware, no breached server. The attack targeted a person and a payment procedure. That is why it is filed under social engineering rather than intrusion, and why technical defences alone would not have stopped it.

Not an isolated case

The same Legislative Council reply listed three deepfake-related fraud cases received by Hong Kong police since 2023:

Police had also tracked 21 deepfake video clips impersonating officials or celebrities since November 2023, all removed at their request.

Regulators elsewhere drew the same conclusion. On 13 November 2024 the US Financial Crimes Enforcement Network issued an alert, FIN-2024-Alert004, on fraud schemes involving deepfake media targeting financial institutions. FinCEN said it had seen an increase in suspicious-activity reports describing suspected deepfakes, especially fraudulent identity documents used to get past identity verification.

What changed

The direct loser was Arup, which by its own account absorbed the loss without damage to its financial stability. The wider cost falls on trust in remote communication. The defences that Hong Kong's police and HKCERT recommended afterwards are revealing: almost none are technical. Verify the invitation, call back on a known number, ask a question only the real person could answer, ask them to turn their head. In effect, a face and a voice on a screen had stopped counting as proof of identity, and the burden of checking shifted to the individual employee taking the call.

It also widened the circle of people who can mount such fraud. In the police account, the raw material was ordinary public footage of executives on YouTube. The same generative tools that made voice cloning cheap were used in January 2024 in the US for a fake Biden robocall aimed at voters; in Hong Kong they were aimed at a payments desk.

See also the cybersecurity timeline. For a different kind of AI-era corporate risk — data flowing out through employees' use of chatbots — see Samsung's ChatGPT leak.