How AI is transforming fraud prevention in banking

Banks are increasingly making use of technologies in a bid to thwart criminals. With a need to balance fulfilling customers’ desires for instant gratification in terms of having items delivered ASAP with the extra time it takes them to verify transactions, new solutions such as artificial intelligence (AI) are increasingly coming to the fore. 

Adrian Schofield, veteran ICT commentator noted that “putting your money into someone else’s hands for safe-keeping and for enabling transactions has inherent risks that offset the planned advantages”. Fraud is nothing new, Schofield said, given that third parties with criminal intentions have constantly found ways to insert themselves into the relationship between banker and client to deprive them of their “cash”. “As fast as the banks adopt new technologies to secure the transactions, the criminals use the same or newer technologies to divert funds to themselves,” said Schofield. 

The widespread adoption of AI in the banking environment has improved banks’ ability to detect and prevent crime in real-time, as they can monitor behaviour, identify anomalies, and authenticate transactions more effectively, said Schofield. Mark Walker, who leads the telecoms data and analytics practice at IDC Middle East and Africa, explained that one of the biggest issues in banking is managing risk, especially since it is a highly regulated sector. “You're dealing with massively confidential data.” 

Banks have, Walker said, always used technology to monitor transactions and key aspects of managing risk through aspects such as anti-money laundering and know-your-customer. “They all involve the use of technology to monitor and react to any unauthorised transaction. So, they do a lot of pattern recognition. This is where AI comes in. AI is good at pattern recognition,” he said. Rufaida Hamilton, Standard Bank’s head of payments in South Africa, has written that, while AI is not a silver bullet, it is a powerful tool in the fight against fraud as “its ability to process and analyse data at unprecedented speed and scale offers the financial sector, particularly in payments, a fresh approach to identifying and preventing fraudulent activity”. Hamilton said that “payments are the first line of defence against fraud losses, and as we continue to refine and develop AI tools for this area, we can expect to see even more sophisticated fraud detection and prediction systems that will help the entire sector curb the alarming rise in major fraud losses”. 

Chris Wood, managing executive of product at Absa, told Personal Finance that some AI-specific uses include when it comes to fraud management and monitoring. He noted that, with machine learning, AI can pick up patterns and is already monitoring transactions from the point at which a customer taps a card all the way through to the settlement of the deal. 

In a real-life example, Wood said that when the point-of-sale machine says it’s authorising a transaction, what it is doing is sending a message to the issuing bank checking whether there is money in the account as well as checking whether spending limits are being breached. “Those rules are all checked in an instant in that transaction.” 

With the addition of various forms of AI, Wood said that transactions can be declined in less than a second if there is something out of the ordinary, such as the card being in a different country than what the bank would ordinarily have expected. “AI is getting smarter and smarter around how it does those checks so that more transactions that shouldn't go through are blocked,” he said. If, for example, a customer’s largest transactions usually occur at the beginning of the month, it would flag any spending outside of this behaviour pattern, Walker noted. This, he said, could be an unusual purchase of a large appliance around the middle of the month, or a credit card being tapped in New York when recent transactions were in South Africa.

Hamilton explained that AI identifies unusual patterns and behaviours that humans might miss, significantly reducing the risk of unnoticed fraud. Yet, said Walker, banks are under pressure to process transactions as quickly as possible to meet customers’ needs for instant gratification while still keeping as much of a lid on fraud as possible.

While Wood noted that while the biggest use of AI in banking and fraud management currently is in performing point-of-sale checks, the banks also seek to, as far as possible, track phishing websites and report these addresses to virus protection companies so they can be added to databases of suspicious links.

Walker added that internet security companies are constantly updating their databases and adding new suspicious websites to them. Each time AI is used, it gets better based on the type of information it receives, he noted. Hamilton said that AI’s capabilities are rapidly advancing. “Its continuously self-evolving nature allows ongoing refinements,” she said. “In fraud risk mitigation, AI has the potential to distinguish between legitimate anomalies and actual fraud, reducing the number of legitimate transactions mistakenly blocked.

By learning from fraud trends, AI-powered systems can adapt and evolve, providing robust protection against new threats,” she said. “That's where I guess the AI world is quite interesting because as more information is being fed into these engines, these databases, this means more likelihood of accuracy. We're probably not there yet,” said Wood.

Yet, as Schofield noted, the “human factor will always present the highest risk as people will always be the weakest link in any security environment”.

PERSONAL FINANCE