Shufti – Fighting AI Fraud With AI

Tom Gadsden is Vice President of Product at Shufti, the London-based identity-verification and fraud-prevention platform built to let businesses onboard and screen users across borders in seconds rather than days. Gadsden has spent more than a decade building identity and financial-crime products, with earlier roles at credit bureau Experian and in card payments, giving him a close view of how fraud has scaled from a cottage problem into an industry.

Why you should listen Shufti sits at an awkward and revealing intersection of the crypto economy. Bitcoin was designed to strip out gatekeepers; Shufti, and the compliance layer it represents, exists to put a version of them back. Gadsden doesn't pretend that tension away. He accepts that know-your-customer checks are anathema to part of the crypto community, and that self-hosted wallets will always let people move funds outside the system. But where crypto collides with fiat on- and off-ramps, governments have decided they want to see where money comes from and where it goes, and firms like Shufti, alongside chain-analytics players such as Chainalysis and Elliptic, are how businesses meet that expectation. He reaches for a useful historical parallel: the card-payments crackdown that arrived with Europe's PSD2 rules, when regulators decided the fraud losses banks had quietly tolerated were grey money the wider economy shouldn't have to absorb. Stablecoin growth, he argues, makes that scrutiny more likely, not less.

The most gripping stretch of the conversation is on deepfakes, where Gadsden describes an arms race that has already run through several generations. Shufti is on the third iteration of its detection models, having moved past early metadata tells to techniques that examine how synthetic images blend, how their frequency signatures break down, and what shows up outside the visible spectrum where the generators stop trying. What has really changed, he says, is economics: AI dragged the cost of a convincing fake from tens of dollars to cents, turning fraud into something closer to a production line, with the same operator filing attempt after attempt from bed. The surge is real, with Shufti warning of a many-fold jump in deepfaked-document attacks year over year. The counterintuitive twist is that as detection has caught up, the pendulum has swung back toward human beings: real faces recruited down the pub, and money mules who sail through the initial check and only later start moving other people's