From Novelty to Criminal Industry
A few years ago, deepfakes were mostly a curiosity - face-swapped videos and glitchy voice clones shared for laughs. That era is over. Deepfake-enabled fraud caused an estimated $1.1 billion in losses in the United States alone in 2025, roughly tripling the previous year's figure, according to research compiled from incident-tracking databases. Reported incidents nearly quadrupled in the first half of 2025 compared to the whole of 2024. This is no longer an edge case - it is one of the fastest-growing categories of financial crime.
How Little It Takes to Clone Someone
The technical barrier has collapsed. Security researchers have demonstrated that roughly three seconds of clear audio - a voicemail greeting, a short social media clip, a snippet from a video call - is enough to generate a voice clone convincing enough to fool a close family member on a phone call. Video deepfakes still take more source material to look flawless, but low-effort versions, good enough to fool someone for a few seconds on a shaky video call, can now be produced with free or cheap consumer tools.
Where the Money Actually Goes
The single biggest category of deepfake fraud is not the "grandparent scam" that gets the most media attention - it's investment fraud, where scammers use AI-generated video or audio of celebrities, business leaders, or even government officials to promote fake trading platforms and investment schemes. This category alone has been responsible for more than half of all reported deepfake-related financial losses. Corporate impersonation - a cloned "CEO" instructing an employee to urgently wire money - is the second-largest category, and has already produced multi-million-dollar single incidents at major companies.
Social Media Is the Delivery Mechanism
The vast majority of deepfake fraud in 2025 - well over three-quarters, by some tracking estimates - originated on social media platforms, a sharp jump from a year earlier. Facebook, WhatsApp, and Telegram have been the most commonly exploited platforms, largely because they combine wide reach, weak content verification, and direct messaging in one place. A convincing fake video of a well-known figure "endorsing" an investment app can spread to thousands of people before anyone flags it.
How to Actually Protect Yourself
The advice that matters most is simple but easy to forget in the moment. If you receive an urgent request for money or sensitive information from someone you know, over a call or video that feels slightly off, hang up and call them back on a number you already have saved - not one they just gave you. Agree on a family "safe word" for genuine emergencies, something no AI system could reasonably guess from your public social media. Be sceptical of any investment opportunity promoted by a celebrity video, no matter how convincing - genuine investment advice is essentially never delivered this way. And treat urgency itself as a warning sign: scammers, human or AI-assisted, rely on you not having time to think.
Where This Goes Next
Detection tools are improving, but they are locked in a genuine arms race with generation tools, and right now the generators are winning on cost and accessibility. Platforms and regulators are beginning to require labelling of AI-generated content, but enforcement is patchy and inconsistent across countries. Until verification catches up with generation, the most reliable defence is still a healthy, habitual scepticism toward anything that asks you to act fast, send money, or share personal details based on a video or voice call alone.
