The AI paradox: Why more AI models don't equal less fraud (www.techradar.com)

In the first half of 2025, fraud cases in the UK skyrocketed, with criminals stealing £629.3 million and over 2 million incidents reported. Despite heightened investments in AI for fraud prevention across banking, insurance, and government sectors, the reality is that simply deploying more models has not yielded a significant reduction in losses. This phenomenon, termed the "AI Paradox," highlights the disconnect between AI's theoretical abilities to detect fraud and the complexities of real-world application, where data relevance, model explainability, and the evolving tactics of fraudsters complicate the landscape. Criminals are increasingly leveraging generative AI tools to enhance their operations, automating sophisticated attacks that pose significant challenges for defenders. The dual-use nature of these technologies has intensified competition as both sides attempt to gain the upper hand. To combat this, organizations are shifting towards Decision Intelligence (DI), which emphasizes contextual understanding and data integration to create a more comprehensive view of threats. By prioritizing relevant data and improving the explainability of AI outputs, enterprises can bolster their fraud detection capabilities and develop strategies that adapt to the rapid evolution of criminal tactics, ultimately shifting from a reactive to a proactive stance in fraud management.