A smarter future for fraud prevention: why detection models need to learn people, not just fraudsters

Fraud is no longer just a game of stolen credit cards and fake identities – it has become an adaptive, AI-powered ecosystem of deception.

Fraudsters are now leveraging artificial intelligence (AI) and machine learning (ML) to scale their operations, automate attacks, and exploit vulnerabilities across multiple payment rails. From scams in real-time payments to application and merchant-level fraud, financial institutions are facing an unprecedented level of complexity in detection across every rail.

The implications are staggering. In 2022, payment fraud in Europe resulted in €4.3 billion in losses¹, while in the UK alone, account takeovers have risen 36% year-over-year².

Beyond financial losses, these trends signal a deeper crisis: an erosion of trust. When fraudsters win, businesses lose customer confidence, and consumers face growing uncertainty about the security of their payments.

Fighting AI with AI: the next frontier in fraud prevention

Visa has long been at the forefront of AI-driven fraud prevention, investing $10 billion in security technology over the past five years. Visa’s models have become highly effective at identifying anomalies, tracking risky activity, and helping block fraudulent transactions.

Visa’s investment has paid off – strengthening fraud detection across the entire payments ecosystem, blocking an estimated $40 billion in fraud in a single year³.

But as fraud tactics evolve, learning fraud alone is no longer enough. Today, fraudsters don’t just imitate customers – they are using AI to become them. Their tactics are constantly evolving: mimicking normal transaction behaviour, bypassing traditional red flags and blending seamlessly into legitimate financial activity. A fraud model that only learns fraud is reactive; it’s always responding to yesterday’s threats rather than anticipating tomorrow’s.

The next evolution in fraud prevention is AI that anticipates people’s behaviours, not just fraud. Real customers have behaviours, habits, and interaction patterns that tend to be more stable than fraud itself. When fraud models learn about individuals – how they spend, where they log in from, when they typically transact – they can detect subtle changes that signal fraud without adding unnecessary friction.

So now, Visa is taking fraud prevention even further with the acquisition of Featurespace, a leader in adaptive fraud detection. Featurespace’s Adaptive Real-time Individual Change-identification (ARIC™) Risk Hub applies advanced machine learning to recognise not just fraud patterns but understand real customer behaviour.

Featurespace’s Adaptive Behavioral Analytics is designed to analyse all transactions and recognise patterns of ‘normal’ or ‘good’ payment behaviour, as well as ‘bad’ behaviour. This allows them to detect changes in behaviour and address new and evolving fraud tactics.

By integrating Featurespace’s AI-driven approach with Visa’s global network intelligence, financial institutions gain a powerful, adaptive fraud prevention framework – one that helps stop fraud before it happens while maintaining a smooth experience for legitimate customers.

What’s next for fraud prevention

The battle against fraud is far from over. With Visa and Featurespace leading the way, fraud prevention is evolving to be more adaptive, more precise, and more customer-centric. A key step forward is integrating Visa’s risk capabilities with Featurespace’s AI-driven fraud detection into a single, unified platform, making advanced fraud protection more accessible and effective across all payment types.

This next-generation approach isn’t just about identifying bad actors – it’s about protecting the people behind every legitimate transaction.