Singapore AI fraud detection test uses cross-bank data to flag scam accounts

Singapore AI fraud detection is being tested across banks by the Monetary Authority of Singapore (MAS), which is working with the banking industry and a law enforcement agency to trial artificial intelligence models using cross-bank and public-private data, Managing Director Chia Der Jiun said in a speech at the Global FinTech Fest. The experiment aims to detect potential fraud sooner, intervene faster and reduce losses, according to MAS.

Singapore AI fraud detection moves from bank-level monitoring to system-level detection

The test addresses a limitation of institution-level fraud controls: scam proceeds may move through accounts at several banks, leaving each institution with only part of the transaction chain. Combining information across participating organisations could help identify suspicious relationships earlier. MAS has not disclosed which banks are participating, the datasets being tested or the criteria that will determine whether the models proceed to deployment.

MAS is also running PathFin.ai, a programme that matches financial institutions with validated AI solutions. Chia said the platform now has more than 300 participants and a growing number of successful matches. It is designed partly to reduce the cost and technical work required for smaller institutions to find and implement AI tools.

Other financial regulators are applying AI and advanced analytics to fraud at different points in the process. The Hong Kong Monetary Authority has directed banks to consider AI and network analytics when monitoring authorised payment scams, including detecting complex networks of suspicious and mule accounts. It has also tested analysis of information from multiple banks in collaboration with the banking sector and law enforcement. The UK Financial Conduct Authority uses machine learning and web scraping to find potentially fraudulent websites. Australia’s ASIC focuses on disrupting the online infrastructure used to attract victims, coordinating the removal of 11,964 phishing and investment scam websites in 2025, a 90% annual increase. FINRA applies machine learning at another layer, analysing hundreds of billions of US market events for potential fraud and manipulation rather than retail payment scams.

MAS expects findings from the test by the end of 2026. The result will show if banks, police and the regulator can use shared data quickly enough to interrupt scam flows before the money is dispersed.