The Scam No Longer Looks Like a Scam
Deepfake fraud is no longer a strange internet problem. It is becoming a billion-dollar crime machine built around something much older than artificial intelligence: trust.
Surfshark’s April 2026 research found that reported deepfake fraud losses reached $2.19 billion globally between 2019 and March 2026, with $1.65 billion stolen in 2025 alone. The United States was the hardest-hit country, with $712 million in reported losses.
That number should make people pause. Not because every scam now uses deepfakes, but because the technology is moving fraud into a more dangerous place. The old scam looked suspicious. Bad grammar. Strange links. A message from someone you did not know.
The new scam can arrive as someone you already trust.
AI Fraud Is Learning How Trust Sounds
Deepfakes are not only fake celebrity clips anymore. They can be fake executives, fake investment endorsements, fake family emergencies, fake bank calls, and fake video meetings.
Surfshark’s dataset found that 52% of reported deepfake fraud losses came from fake government-official or celebrity investment endorsements. Another 25% came from corporate attacks, including executive impersonation. Romance scams, family-member impersonation, and financial crimes made up smaller but deeply personal parts of the damage.
That is why this story matters beyond cybersecurity. AI fraud attacks the basic signals people use to decide whether something is safe: a familiar face, a known voice, a confident tone, a trusted title.
Once those signals can be copied, the victim is not only fighting a scammer. They are fighting their own instinct to trust.
The $25.6 Million Video Call
One of the clearest warnings came from Arup, the British engineering firm behind landmarks including the Sydney Opera House.
In January 2024, a finance employee at Arup’s Hong Kong office received an email that appeared to come from the company’s CFO, requesting a series of confidential transactions. The employee suspected it was a phishing attempt. Then he joined a video call.
On that call, the CFO was present. So were several colleagues. They looked right. They sounded right. They gave instructions with authority and familiarity. The employee put aside his doubts and made 15 wire transfers totalling $25.6 million to accounts controlled by the fraudsters.
Every person on that call was an AI-generated deepfake. CNN Business later confirmed Arup as the victim. None of the stolen funds have been recovered.
That case matters because it breaks the comforting myth that only careless people fall for scams. This was not a suspicious email from a stranger. It was a workplace setting. A meeting. Senior people. Instructions that appeared to come from authority. The employee had even been suspicious, and the deepfake still convinced him.
AI Scams Are Accelerating Faster Than Traditional Fraud
The speed of growth is what makes this different. Pindrop’s 2026 analysis found that AI fraud surged 1,210% in 2025, compared with a 195% rise in non-AI fraud. That means AI-powered scams are growing at six times the rate of traditional fraud.
The FBI’s 2025 Internet Crime Report places the broader cybercrime environment at record levels, with cyber-enabled crimes defrauding Americans of nearly $21 billion in 2025. The FBI logged more than 22,000 AI-related complaints, with adjusted losses exceeding $893 million.
AI does not need to invent fraud. It only needs to make old fraud faster, cheaper, more convincing, and easier to scale.
The Real Danger Is Not the Fake Video
The most dangerous part of deepfake fraud is not the technology itself. It is what the technology does to everyday life.
A child’s voice asking for emergency money. A boss telling an employee to move funds. A bank manager confirming a transfer. A public figure endorsing an investment. A video call that looks too normal to doubt.
That is why many victims may not recognise the AI layer until after the money is gone. The scam does not announce itself as artificial intelligence. It arrives as urgency, authority, fear, or love.
What You Can Actually Do
Security researchers say the most effective protection is a pre-agreed safe word or verification code between colleagues and family members for any request involving money or sensitive information. If a call, message, or video request asks you to transfer funds or share credentials, hang up and call back on a number you already know. No legitimate employer, bank, or family member will object to that pause.
For organisations, multi-person authorisation for any financial transfer above a set threshold is now considered essential. The Arup case succeeded because one person had the authority to approve $25.6 million alone.
The old advice was simple: do not click strange links. The new advice is harder: do not trust a face, a voice, or a video call just because it feels familiar.
Deepfake fraud is not just stealing money. It is stealing the confidence people once had in what they could see and hear.
By Shizza Farooqui
SOURCES
Surfshark Deepfake Fraud Study, April 2026 | Pindrop AI Fraud Spike Analysis, 2026 | FBI IC3 Internet Crime Report 2025 | CNN Business reporting on Arup deepfake fraud case | World Economic Forum interview with Arup CIO | Vectra AI Analysis 2026 | Experian Fraud Forecast 2026
