The same AI that writes emails and clones voices is now a fraudster's best tool. Banks and regulators are warning that generative AI is driving a fast-rising wave of financial fraud, and the numbers are sobering: card network Mastercard projects that AI-fueled fraud losses could surge to nearly $60 billion by 2030, up sharply from around $23 billion in 2025, Yahoo Finance reported. Regional data tells the same story — reported fraud has jumped by double digits over the past year in markets from the UK to India.

How AI turbocharges the con

The danger isn't a new kind of crime so much as old scams made cheaper, faster and far more convincing:

  • Voice cloning. With only seconds of audio — scraped from a podcast, an earnings call or a social video — fraudsters can generate a convincing clone of someone's voice to fool relatives or to beat phone-based "voice ID" checks.
  • Deepfake video. AI can fake a live video call. In a now-infamous case, an employee at engineering firm Arup was tricked into wiring about $25 million after a video call populated by deepfaked colleagues, including a fake "CFO," Deloitte noted.
  • Synthetic identities. Generators can spit out a coherent fake identity — name, ID, utility bills, pay stubs — in minutes, used to open fraudulent accounts.
  • Flawless phishing. AI-written scam messages no longer have the clumsy grammar that used to give them away, making them harder for both people and filters to catch.

The common thread: AI removes the skill and cost barriers that once limited fraud, letting criminals run convincing scams at scale.

Banks are fighting AI with AI

Lenders aren't standing still. They increasingly deploy their own AI to detect fraud in real time — flagging odd transaction patterns, device and location mismatches, and other anomalies. A growing tool is behavioral biometrics: software that learns how you type, swipe and hold your phone, and raises a flag when an account is operated by someone whose behavior doesn't match — even if the password is correct. The result is an escalating "AI versus AI" arms race, with each side automating faster.

Regulators are also sounding alarms. The US Federal Trade Commission has reported billions of dollars in losses to imposter scams, and bodies in the UK and elsewhere have flagged the AI-fraud threat. (We've cited the headline projections and the regulators' own figures; some industry loss estimates circulate widely but are harder to verify, so treat very precise vendor numbers with caution.)

How to protect yourself

The practical defenses are unglamorous but effective:

  • Be suspicious of urgency — even from a familiar voice or face. Deepfakes thrive on pressure ("send it now").
  • Verify through a separate channel. If "your bank" or a relative calls asking for money or codes, hang up and call back on a number you look up yourself.
  • Turn on multi-factor authentication everywhere, and never share one-time codes or recovery keys.
  • Agree a "safe word" with family for emergencies, so a cloned voice can't pass.

Why it matters

For the financial system, AI-driven fraud is a rising cost and a trust problem: if customers can't believe a voice or a video, the friction of verifying everything goes up for everyone. For banks, it's both a threat and a spur to invest in defenses. And for individuals, it's a reminder that the weak link is increasingly human — the technology to fake a face or a voice is now cheap and good, so the old instinct to "trust but verify" has never mattered more. Boursel won't overstate the precise dollar tallies, which vary by source; the direction, confirmed by banks and regulators alike, is clear: the scams are getting better, and faster.