More than four million cases of fraud were reported in the UK last year – equivalent to nearly eight every minute – as criminals used artificial intelligence to steal almost £1.3bn, according to an annual report by UK Finance. The total stolen has risen by more than £1bn in two years, with investment scams alone costing victims £221.5m, a 40% increase from the previous year.
Banks described fraud as a “national security threat”, citing the impact on victims and the huge sums funnelled to organised criminals. Criminals have turned to AI to mimic the voices of celebrities and even victims’ own family and friends, enabling them to carry out larger-scale fraud. Ruth Ray, UK Finance’s managing director for economic crime, said AI allows scammers “to spin up websites quickly and easily to make your business look legitimate” and “to mimic voices of celebrities or even people’s friends and family to fool people”.
“Nearly £1.3bn lost to fraud in UK last year, with 4m cases reported as AI fuels romance and investment scams.”
Investment scams, which promised high returns on gold, cryptocurrencies, property, wine and carbon credits, accounted for almost 15,000 reported cases. Romance fraud also surged, with UK Finance noting examples of fraudsters marrying victims to continue stealing money. Kirsty Guest, a florist from North Yorkshire, lost £80,000 after meeting a man on a dating app who used stolen photos. “They are professional and they are making massive volumes of money,” she said. Julie Osgood, 60, said the first four men she matched with on a dating site were all potential fraudsters.
Paul Davis, Barclays’ head of economic crime, said “the impact goes beyond financial loss; it can cause huge emotional harm, leaving victims burdened by guilt and shame”. The mandatory fraud reimbursement scheme for authorised push payment fraud reimbursed 88% of losses, but UK Finance said the scale of the problem can only be tackled if tech companies step up monitoring and security on their platforms. Experts believe the majority of scams go unreported, meaning the true figure is likely far higher.