10 deepfake examples: real fraud cases and how to recognise them (2026)

In brief: A deepfake is audio, video or an image made with AI that shows a real person saying or doing something that never happened. The ten deepfake examples below are documented cases from 2019 to 2026, from the HK$200 million video conference at Arup to a fake Dutch prime minister selling investments on Facebook. Six of the ten were aimed at a company. The pattern is constant: a familiar face or voice, an urgent and confidential request, and a channel that is not the usual one. The attempts that failed were stopped by an employee who verified, not by someone who spotted a glitch.

What a deepfake is, and why it is an employer’s problem

The raw material is public: the attackers who targeted WPP used YouTube footage of the CEO. For a company, the deepfake is rarely the whole attack. It is a convincing layer on top of familiar social engineering, such as CEO fraud or vishing. The procedure that stops it is set out step by step in our guide to deepfake awareness training.

• The call comes second. Most business cases start in WhatsApp or email. The deepfake follows to remove the doubt.
• The request is the weak spot, not the quality. Urgent, secret, and about money or access.
• Procedure beats perception. Calling back and asking a control question stopped the attempts below.

The 10 deepfake examples at a glance

The 10 deepfake examples in detail

1. UK energy firm (2019): a cloned voice and €220,000

What happened: In March 2019 the CEO of a UK energy firm took a call from what sounded like the chief executive of its German parent company, asking for €220,000 for a Hungarian supplier within the hour. He paid, according to insurer Euler Hermes.

The lesson: Only a third call, from an Austrian number, made him suspicious. One call back to the known number would have ended it.

2. bunq (2023): the Dutch employee who checked

What happened: An employee of Dutch bank bunq got a WhatsApp message from “CEO Ali Niknam” about an acquisition that needed a fast down payment, followed by a video call with a fake Niknam, NOS reported.

The lesson: The blocky image was blamed on a bad connection. The employee did not pay and checked with the real Niknam.

3. Arup (2024): HK$200 million after one video conference

What happened: In January 2024 an employee at Arup’s Hong Kong office joined a video conference with the CFO and several colleagues, all digital clones. He made 15 transfers: HK$200 million, about £20 million (US$25.6 million). Arup confirmed that “fake voices and images were used”.

The lesson: It surfaced only when he asked headquarters. Several familiar faces in one call are still one channel.

4. Biden robocall (2024): a deepfake for voters

What happened: Two days before the New Hampshire primary of January 2024, thousands of voters got a call in an AI voice resembling President Biden, telling them to save their vote for November. The consultant behind it was fined $6 million by the FCC; a jury acquitted him of criminal charges in 2025.

The lesson: Employees meet this type as citizens. The habit is the same: a familiar voice is not a source.

5. LastPass and WPP (2024): two CEOs on WhatsApp

What happened: A LastPass employee received calls, texts and a voicemail on WhatsApp with an audio deepfake of the CEO, the company wrote. At WPP, attackers used a WhatsApp account with CEO Mark Read’s photo to set up a Teams meeting with a voice clone and YouTube footage.

The lesson: Both contacts came outside the normal business channels. The LastPass employee ignored them and reported it. Neither company lost anything.

6. Ferrari (2024): the question the caller could not answer

What happened: In July 2024 a Ferrari executive got WhatsApp messages and then a call from “CEO Benedetto Vigna”, southern Italian accent included, about a big acquisition and a currency transaction, The Drive wrote, citing Bloomberg.

The lesson: He asked which book Vigna had recently recommended. The caller hung up. A personal question costs nothing.

7. KnowBe4 (2024): the deepfake that applied for a job

What happened: Security awareness vendor KnowBe4 hired a software engineer who turned out to be a fake IT worker from North Korea. He used a stolen US identity and an AI-altered stock photo and passed four video interviews, CyberScoop reported.

The lesson: His new laptop loaded malware as soon as it arrived, and security contained it in about 25 minutes. Verify identity beyond the video call.

8. Guido Crosetto (2025): a minister’s voice, a Dutch bank account

What happened: In February 2025 Italian business leaders were called by “defence minister Guido Crosetto”, who asked for around €1 million to free kidnapped journalists. Former Inter Milan owner Massimo Moratti paid almost €1 million, Reuters reported.

The lesson: A minister does not ask citizens for money. Moratti filed a complaint, and police froze the money on a Dutch bank account.

9. Dick Schoof (2025): the prime minister as investment salesman

What happened: In June 2025 the Dutch consumer organisation Consumentenbond warned about a deepfake video in which Prime Minister Dick Schoof appears to present an investment platform. Fraudsters spread it through Facebook ads.

The lesson: This is the deepfake employees see first, on their own phone. No politician sells investments in an ad.

10. Singapore (2026): a prime minister on Zoom

What happened: In May 2026 Singapore police warned about a victim who lost at least S$4.9 million after a Zoom call with deepfakes of Prime Minister Lawrence Wong and other officials. The request: urgent funding linked to the Strait of Hormuz, to be reimbursed by the government.

The lesson: The follow-up email came from a proton.me address. Governments do not ask individuals to transfer money.

The four types of deepfakes

• Voice clone: a call or voicemail in a familiar voice, as in examples 1, 5, 6 and 8.
• Live video call: fake participants in Teams or Zoom, as at bunq, Arup and in Singapore.
• Pre-recorded video: a clip spread through ads or social media, like the Schoof video.
• Synthetic identity: a stolen or invented identity with an AI-altered photo, used to get hired, as at KnowBe4.

How to recognise a deepfake, and why “spot the glitch” is getting weaker

The classic tips still apply: lips that lag behind the sound, unnatural blinking, a face that blurs when a hand passes in front of it, a flat or slightly mechanical voice.

As a control, though, the trained eye is failing. In a 2025 test by iProov among 2,000 consumers in the UK and the US, 0.1% correctly identified all real and fake content. At Ferrari the accent was nearly perfect. At bunq the poor image was simply explained as a bad connection.

What stays recognisable is the request. In all ten examples the content gave the fraud away before the pixels did: urgency, secrecy, an unusual channel, and money or access at the end.

What works instead: checks that do not depend on your eyes

1. Call back on a number you already have. Never the number in the message.
2. Four-eyes principle on payments. A second approver for large transfers and for every change of bank details.
3. A code word or a personal question. Agreed between management, finance and assistants, never written down in chat.
4. Switch channel. Confirm a request by WhatsApp or video through another route.
5. Report every attempt. The next target is usually a colleague.

How to train employees for deepfakes

Start with the roles that can move money or grant access: finance, executive assistants, HR and recruiters, the IT help desk. Give them cases, not theory. The ten above work because they are real, and they fit next to the attacks in our overview of types of phishing attacks.

Then practise the procedure itself. Who do you call, on which number, and what happens to the payment in the meantime? Management has to say out loud that being called back is welcome. Short and frequent works better than one annual session: Guardey gives every employee a gamified challenge of about three minutes each week, with phishing simulations included.

Frequently asked questions

What is an example of a deepfake?

A video call in which the CFO and colleagues are AI-generated, as at Arup in 2024. Other examples are a call in the cloned voice of a CEO (Ferrari) and a fake video of a politician promoting investments (Dick Schoof).

Are there Dutch deepfake examples?

Yes. In 2023 a bunq employee was targeted with a fake video call from the CEO, and in 2025 a deepfake of Prime Minister Dick Schoof appeared in investment ads on Facebook.

Can you still recognise a deepfake by looking closely?

Sometimes, but do not rely on it. In the iProov test only 0.1% of participants got everything right. Verification through a second channel works however good the fake is.

What should you do if money has already been transferred?

Call your bank immediately and report it to the police. In Singapore in March 2025 a finance director sent US$499,000 after a deepfake Zoom call. He alerted his bank a day later, and police traced and held the money the day after that.