How banks are using Gen AI to detect fraud

Banks are using generative AI (Gen AI) in the ongoing battle against payment fraud. The technology is effective at reducing false positives – a long-standing issue that’s been a source of frustration for financial institutions and their customers. 

Why is this noteworthy? Fraud has always been a challenge, but traditional detection systems rely on rules-based approaches. This process could unintentionally flag legitimate transactions as suspicious. 

When this happens, customer frustration skyrockets and it could even lead to the potential loss of businesses. Gen AI, however, is learning and adapting in real-time. 

Concept clarified: Generative AI is a type of artificial intelligence that can create new content, including text, images, or data patterns, based on what it has learned from existing data.

Gen AI’s technological edge in fraud detection 

Banks require more sophisticated tools and systems as fraudsters come up with increasingly complex schemes. Enter Gen AI, which is used by banks to improve fraud detection accuracy without inconveniencing customers. 

How? 

There are several advantages to using Gen AI in this regard: 

• Continuous learning.
• Synthetic data generation.
• Unstructured data analysis.

Continuous learning

Unlike traditional systems and the tools of old, Gen AI rapidly refines and adapts its understanding of patterns. 

It distinguishes between legitimate and fraudulent payments, and this process only gets better over time, as it continues to learn from the data it absorbs.  

Synthetic data generation

GenAI can also produce synthetic datasets that mimic real-world financial data. This means robust model training can take place, without compromising privacy or compliance.

It’s no secret that privacy is a thorny issue in today’s digitally-connected world. 

A recent PwC survey titled the ‘Voice of the Consumer Survey 2024’ found that a staggering 83% of consumers say they value and trust institutions that prioritize their personal data. 

Unstructured data analysis

Lastly, GenAI excels at detecting subtle and novel fraud patterns in unstructured data. This significantly enhances overall fraud detection practices.

And it doesn’t stop there. Industry giants have latched on to Gen AI’s capabilities. 

How Mastercard uses Gen AI to prevent fraud

Visa and Mastercard, for example, have developed their own in-house generative AI tools, specifically for this purpose. 

Johan Gerber, executive vice president for security and decision products, says fraudsters no longer get to “operate in obscurity” by laundering the card details of unsuspecting victims.

Gerber explains: “We can now piece together the jigsaw – enhancing trust to banks, their customers and the digital ecosystem as a whole.” 

Mastercard’s in-house AI-based predictive technology does this by:

• Doubling the detection rate of compromised cards.
• Increasing the speed of identifying merchants at risk from fraudsters by 300%. 
• Reducing false positives during the detection of fraudulent transactions by up to 200%.

AI fraud detection: Where to from here?

Using AI for fraud detection isn’t without challenges. The sophistication that makes it such a powerful tool also poses obstacles to widespread implementation.

Privacy concerns are still rife, and the potential for bias in AI models cannot be downplayed. Add regulatory hurdles to the mix, making it a fine line to walk between progress and setback. 

That said, the potential benefits are compelling enough to keep pursuing and refining the technology. As generative AI matures, it could become a cornerstone of modern payment fraud prevention strategies.

About the author