AI-led detection, human-led decisions

At the core of this layered defence are AI models trained on large and complex data sets in real time. By analysing actual fraud cases, the models learn to identify patterns, assess risk and flag suspicious activity at an early stage.

The need for such advanced capabilities has never been greater. Financial crime is evolving faster than ever, and fraudsters are increasingly using AI to their own advantage. They imitate genuine customer behaviour, automate attacks and conceal digital traces that would previously have been detected. 

Since 2018, FCP has actively used artificial intelligence to combat fraud, combining advanced technology with deep domain expertise and close collaboration across the banking sector. 

“In fraud prevention, we use AI strategically – where it has the greatest effect. At the same time, people are always in control,” says Nikolay Martyushenko, Head of Data & AI at Financial Crime Prevention (FCP). 

When fraudsters use AI, we respond with the same precision 

Traditional rule‑based systems have long formed – and still form – the backbone of fraud prevention. They still capture many clear deviations, but struggle when fraud becomes more subtle and sophisticated. 

“Today, fraudsters use AI to adapt language, timing and behaviour so that they appear almost indistinguishable from legitimate customers. In such cases, fixed rules alone cannot keep up,” explains Martyushenko. 

This is where AI‑based analysis proves its value. By learning what normal customer behaviour actually looks like, the models can detect small irregularities – patterns that are difficult for humans to spot. 

“AI is particularly valuable when signals are weak and deviations are small. This is often the case in identity fraud, where criminals impersonate someone else,” he adds. 

Knowledge as a line of defence 

Stopping a suspicious transaction is only the first step. Equally important is the follow‑up work: determining whether the decision was correct. This often involves direct contact with the end customer, via phone or SMS. 

“This investigative work takes time, but it is absolutely essential. If cases are not classified correctly, the models will learn the wrong things,” says Martyushenko. 

The more precisely FCP can determine what was actually fraud – and what was not – the stronger the foundation for continued learning, testing and improvement. 

“For us, fraud prevention is not just about algorithms, but about building, understanding and maintaining the right knowledge at all times.” 

This knowledge is not developed in isolation. FCP collects and analyses fraud patterns across all Norwegian banks and large parts of the Swedish market. 

“When a pattern emerges at one bank, it can offer vital protection to others. By sharing insights across institutions, we create a powerful defence against increasingly coordinated and technologically advanced crime.” 

Artificial intelligence strengthening payment fraud detection 

With Fraud Explore, FCP is taking a new and significant step in AI‑driven fraud prevention. Built on the company’s own Large Financial AI Model (LFM), the solution is designed to strengthen real‑time fraud detection across the Nordics – and eventually internationally. 

“At the core is a shift away from predefined rules and known fraud patterns. Instead, the model learns what normal financial behaviour looks like, based on billions of transactions,” Martyushenko explains. 

This makes it possible to detect unusual activity at an early stage – even when the fraud does not match any previously documented cases. 

AI agents in case handling 

On top of the financial model sits a set of LLM‑based AI agents, together forming the next generation of case handling within FCP. 

“This is where the future lies: AI assists – people decide,” says Martyushenko. 

The agents can structure complex information, suggest assessments, automate routine tasks and provide clear explanations based on vast amounts of data. They closely support case handlers in their daily work, delivering assistance in real time. 

“The goal is to remove the most time‑consuming and repetitive tasks, allowing case handlers to focus on decisions where human judgement truly matters,” he explains. 

This agent‑based approach will be tightly integrated with FCP’s real‑time AI service layer. As payments and fraud attempts happen ever faster, prevention must occur at the moment of transaction. 

“Fraud must be stopped as it happens – not afterwards.” 

Layered defence against card fraud 

In card fraud prevention, FCP applies a layered defence model. The first line of defence is Token Enrolment Monitoring (TEM), which monitors and controls attempts to add payment cards to digital wallets. 

“TEM stops between 70 and 80 per cent of fraud attempts already at enrolment, before the cards are even used,” says Martyushenko. 

The next layer is Card Transaction Monitoring (CTM), which analyses card transactions in real time and learns what constitutes normal card usage for each individual customer. 

“Here, AI analyses patterns in purchase behaviour, amounts, timing and location. By detecting deviations early, transactions can be risk‑assessed and stopped – while reducing the number of false positives.” 

Read more about this approach here: How phishing and fraud in digital wallets are stopped 

AI with a clear direction 

Looking ahead, Martyushenko points to three strategic priorities for FCP: expanded use of AI agents, access to broader and richer data sources, and a clearer shift from being a system provider to becoming a knowledge‑driven organisation. 

“FCP doesn’t just build systems. We build, preserve, make accessible and continuously develop knowledge – with the help of AI. Technology is a means, not an end in itself.” 

For FCP, the AI strategy is therefore not about replacing people, but about empowering them. Technology is applied where it delivers the greatest impact – in pattern recognition, real‑time assessment and decision support – while people remain responsible for judgement, accountability and context. 

True value emerges at the intersection of advanced artificial intelligence, human expertise and shared insight across banks. 

By combining these elements, FCP is not only building solutions, but one of the strongest defences against financial crime in the Nordics – ready to meet both today’s threats and those of tomorrow.