Battling fraud with AI

As fraud tactics shift, how can retailers tackle ecommerce threats with AI? Philip Plambeck, MD, Computop UK explains

Earlier this month, UK Finance published its Annual Fraud Report for 2025, painting a complex picture of fraud trends across the UK financial landscape. One of the most notable findings was a 2% decrease in Authorised Push Payment (APP) fraud losses, paired with a 20% drop in APP case volumes—the lowest since 2021. While this decline is encouraging, it comes with a warning: criminals are rapidly changing tactics. Remote Purchase Fraud, particularly using stolen card data in online transactions, surged by 22% in volume and 11% in value—the first increase since 2018.

This shift reflects the dynamic nature of fraud. As banks strengthen APP fraud prevention—through investment in behavioural analytics, two-factor authentication, and collaborative intelligence-sharing—fraudsters are refocusing their efforts. The new battleground is the online retail space, where remote purchase fraud now dominates the card fraud landscape, making up more than 80% of unauthorised card fraud and representing over £399 million in losses, according to the report.

Wake-up call for eCommerce merchants

For retailers, this marks a critical moment. Online transactions remain the prime target due to their sheer volume, reliance on card-not-present purchases, and the growing consumer shift to digital wallets. Fraudsters are using increasingly sophisticated social engineering tactics to bypass one-time passcodes, exploit digital wallet registration processes, and leverage stolen credentials obtained from data breaches. The threat isn’t just financial, retailers face reputational risk, chargebacks, and disrupted customer experiences.

Reactive fraud detection is no longer enough. Retailers need a proactive, layered defence strategy. And at the heart of that strategy should be intelligent, data-driven decision-making—powered by AI.

AI fraud scoring is a defensive game-changer

To combat evolving fraud patterns, many forward-thinking eCommerce retailers are turning to AI-driven fraud scoring systems. These evaluate the risk of each transaction in real time using machine learning models trained on historical fraud data, behavioural analytics, and contextual inputs such as device ID, IP address, geolocation, transaction patterns, and even biometric signals.

An AI Fraud Score provides a dynamic probability that a transaction is fraudulent. Transactions that exceed a predefined risk threshold can be flagged for further review, delayed, or outright declined—without impacting the user experience for legitimate customers. Here’s how AI makes the difference:

Integrating AI into retail fraud strategy

Retailers implementing AI fraud scoring should follow a multi-pronged approach:

Agentic AI threat could be next

The UK Finance report also flagged the emergence of Agentic AI as a rising concern. This new class of AI can autonomously plan and execute fraudulent schemes—streamlining phishing, identity theft, and synthetic fraud campaigns at scale. Retailers must be prepared for a future where the speed, scale, and sophistication of fraud outpace human detection capabilities.

The only viable countermeasure? Matching AI with AI. Retailers need systems capable of defending against machine-driven fraud with machine-speed responses—instantly identifying anomalies, blocking suspicious actions, and learning from each attempt.

Retailers are now frontline defenders in the UK’s evolving fraud landscape, and it has become imperative that they understand the innovations happening in retail fraud prevention. By adopting AI fraud scoring, collaborating across sectors and adopting a preventative, not reactive strategy, retailers can protect themselves—and their customers—from the next wave of financial crime.