As fast-developing artificial intelligence (AI) technology increasingly penetrates various industries, boosting efficiency, it brings with it a threat to the financial security of individuals, corporates and financial institutions, especially banks, which are responsible for detecting fraud and reducing client monetary loss.

As an example of such threats, in January of this year, multinational engineering company Arup lost HK$200 million (US$25.58 million) due to a deepfake video that was generated using AI. The video targeted one of the company’s employees at its Hong Kong office by using a digitally cloned version of its chief financial officer to demand specific monetary transfers. The fraud case is still under investigation by Hong Kong authorities.

The number of deepfake scams is rising sharply. In the Asia-Pacific region, the number of deepfake cases detected in 2023, research from tech firm Sumsub finds, increased 1,530%, the majority of which occurred in Vietnam (25.3%) and Japan (23.4%). Other countries notably affected by such fraud were Australia (9.2%), China (7.7%) and Bangladesh (5.1%).

Banks, to maintain a competitive edge in their fight against AI-enabled fraud, Sumsub argues, should focus on coupling modern technology with human intuition to determine how technologies may be used to pre-empt attacks by fraudsters.

In banking, AI algorithms are best tasked with the continuous monitoring of accounts to analyze transaction patterns to detect signs of fraud – for example, unusual large withdrawals or unexpected overseas transactions.

And AI-powered advanced machine learning models, a report by US tech firm DigitalOcean explains, can delve into credit and loan applications to root out synthetic identity fraud by uncovering anomalies that may suggest fabricated identities, preventing financial loss before it occurs.

As a result, the operational efficiency of banks is bolstered as AI takes on the initial detection workload, allowing human investigators to focus on the in-depth analysis of the most high-risk alerts.

When building up the AI fraud detection systems, there is also a set of challenges that need to be solved. Among the challenges, data quality and availability are significant considerations as privacy concerns and regulations may limit the availability of data, thus requiring the maintenance of a balance between data integrity and privacy laws.

In addition, AI systems can still generate false positives – that is, legitimate transactions flagged as fraudulent. These can create friction with customers, leading to frustration and potentially damaging customer-business relationships.

As a result, for banks, the crucial ongoing challenge  – among others, like integrating AI with existing systems and keeping up with evolving threats – is to balancing sensitivity to fraud with the need to provide a smooth customer experience, according to Digital Ocean.