Banks are fighting AI fraud with AI. Data sharing may matter more

Machine learning helps spot scams in real time, but criminals move money across many institutions. No bank sees the full picture.

Fraud Watcher NewsroomEditorial team

26 Oct 2026, 09:00·1 min read·No comments

Banks have used machine learning to score transactions for years. What’s new is the shift from catching stolen-card fraud to spotting scams, where the customer authorises the payment themselves.

Behaviour is the signal. Hesitant typing, an active phone call during a banking session, or a first-ever large payment to a new crypto exchange can all indicate a customer is being coached.

One bank sees one piece. Mule accounts receive stolen funds and move them on within minutes, often across several institutions. Spotting the pattern requires shared intelligence.

The policy push

International bodies including the IMF have argued that AI can tip the balance against fraud if banks share data more effectively. Privacy rules and competition concerns remain obstacles.

Source: International Monetary Fund

Understand it · Fraud LibraryAPP fraud: when you’re tricked into sending the money yourself

Image | Photo: Taylor Vick / Unsplash

On Fraud Watcher | “Move your money to a safe account”: why this old scam still works

Fraud Watcher Newsroom

Editorial team

The Fraud Watcher editorial team covers fraud, scams and financial-crime regulation. Every story links to its primary source.