Fraud is no longer just a game of trickery, it’s now a tech arms race. The latest report from Feedzai shows the bad guys are winning more often than not, thanks to generative AI.
According to Feedzai’s 2025 AI Trends in Fraud and Financial Crime Prevention report, more than half of all fraud now involves artificial intelligence, with criminals increasingly using tools like deepfakes, voice clones, and AI-generated phishing to con their victims. Ninety-two percent of financial institutions say they’ve seen fraudsters misusing AI.
Hyper-realistic impersonations are becoming fraud’s favourite weapon. Forty-four percent of fraud professionals report seeing deepfakes used in scams, while 60% say voice cloning is a rising threat. In addition to AI-driven social engineering and phishing attacks (cited by 56% and 59% of professionals, respectively), account takeovers and scams are becoming increasingly hard to detect but also more likely to be reimbursed, making them an expensive headache for banks.
Fighting AI-driven fraud with AI
Banks aren’t sitting still. Nearly 90% are now using AI to detect and prevent fraud, and two-thirds of those adopted the tech within just the past two years. AI is helping identify suspicious activity in real time, catch sophisticated scam tactics, and boost efficiency within fraud teams. Unlike criminals, banks can’t just plug in a model and let it rip.
While criminals can exploit AI with zero oversight, banks must operate under strict ethical, regulatory, and technical constraints. Data management, in particular, is a sticking point: 87% of banks report issues with fragmented data and regulatory barriers slowing AI adoption.
There’s also a growing emphasis on explainability. With rising pressure from regulators and the public, 89% of banks say they’re prioritising transparency and fairness in AI systems.
Despite the challenges, financial institutions are relying on AI as the backbone of their fraud defence. AI is currently being used to tackle scams (50%), transaction fraud (39%), and money laundering (30%). Rather than replacing human investigators, it’s making them more effective — 43% of institutions report greater efficiency across fraud teams, as AI helps surface the most complex and high-risk cases for expert review.