From the years 2007 to 2015, one of Denmark’s largest banks became embroiled in a massive money laundering scandal. Billions of dollars were laundered through the bank, highlighting significant deficiencies in the bank’s anti-money laundering (AML) controls. By the time regulators caught up, some $ 200 million had been laundered through the bank, and it remains one of the most staggering financial crimes in European history.
This case rattled the banking sector, but the harder truth is that financial crimes don't make headlines every time. They happen quietly, constantly, and at enormous scale. Though banks are getting smarter and putting rigorous checks in place to combat financial crimes, fraudsters are also constantly evolving and upping their game.
The state of fraud in banking
Fraud in banking is like an epidemic. Scam losses worldwide crossed US $1 trillion in 2024, according to the Global Anti-Scam Alliance (GASA). Almost half of consumers are targeted by scams at least once a week. Yet for those who fall for them, recovering losses is the exception, not the rule – just 4% get their money back in full. As scams escalate in volume and complexity, the message to the banking industry is clear: stronger safeguards are needed at the earliest.
Artificial Intelligence (AI) can help make a difference and has been found to be a strong counter to fraud in banking.
Gone are the days when traditional fraud detection methods worked. In simpler times, rule-based systems were set up and worked well. If transactions exceeded a particular value, alerts would be triggered so that those transactions could be closely studied. Fraudsters now know how to get past these kinds of checks. Another issue with rule-based systems is that they would trigger several false positives. This would require compliance teams to spend time looking into such transactions that were often legitimate.
AI-based fraud detection uses machine learning (ML) to analyze large volumes of data and spot anomalies in this data. It does not depend on static rules. It learns from past behaviour and adapts over time, thus becoming better and better at spotting fraud.
Let’s look at how AI is transforming the landscape of fraud detection.
Machine and deep learning
Machine learning (ML) is central to AI-based fraud detection. The ML algorithms are trained on humongous volumes of datasets of both fraudulent and legitimate transactions. These algorithms learn to distinguish between genuine transactions and fraudulent ones. Any kind of behaviour that deviates from what is normal for a particular account is instantly flagged.
JPMorgan Chase upleveled its fraud detection capabilities by integrating large language models (LLMs) to evaluate transaction patterns in real time. With AI at the helm, fraud-related losses dropped by 40%, and detection became significantly faster, giving the bank a decisive edge.
Behavioural biometrics
Behavioural biometrics focuses on how a user physically interacts with a device. How fast are the mouse movements and touchscreen gestures? Even if fraudsters steal the usernames and passwords of accounts, these kinds of micro behaviours by customers are difficult for fraudsters to replicate and mimic.
NatWest's BioCatch (one of the global leaders in behavioural biometrics) deployment has helped the bank successfully prevent online fraud attempts, helping to protect customers. By tracking more than 500 behavioural signals, BioCatch builds a unique profile for each user, making it significantly harder for fraudsters to impersonate legitimate customers, even with stolen credentials.
Graph analytics
Behind most large-scale frauds, there are shell companies and intermediaries who are present to obscure the trail. Graph analytics uses AI to map relationships between companies, visualising hidden trails and connections so that suspicious clusters and fraud networks can be identified before they escalate. Some of the world’s largest banks use graph analytics to fight fraud. A leading US investment bank is using graph analytics layered onto its existing ML system to detect hidden links between known fraud cases and new card applications. This resulted in more fraud rings surfacing, and compromised cards shutting down faster.
Natural language processing
Many of the fraud signals are present in emails, chats, and call transcripts. Natural Language Processing, which is a branch of AI, has the ability to scan vast quantities of text and detect anomalies. This capability is particularly powerful for detecting trade finance fraud and procurement fraud, where the tell-tale signs are buried in paperwork that no human team could realistically review in full.
Conclusion
Banks have to invest in AI-powered fraud detection to stay on top of the game and protect themselves. Banks may be leveraging AI to fight fraud, but fraudsters are not standing still. They too are adopting the same technologies, making the battle smarter, faster, and harder to win. They are using AI tools to find loopholes to exploit, such as crafting deepfake voices, deploying bots to test stolen credentials at scale, and generating synthetic identities to open accounts. Banks that act decisively today will retain customer trust tomorrow. In fraud prevention, the threats you don't see coming are the ones that hurt the most.
How Infosys BPM can help
Fraud is one of the most complex and costly challenges facing banks, financial institutions, retailers, and telcos today.Infosys BPM helps organisations fight back with real-time fraud detection, advanced analytics, and tailored solutions that cut false positives, surface anomalies fast, and protect revenue at scale.
