Singapore Tests AI Scam Detection System Using Data From Five Banks To Flag Fraud Earlier And Reduce Customer Losses

Scam detection across Singapore’s banking system is being reworked with artificial intelligence at the centre, as regulators and industry partners test whether shared data and machine learning can help flag suspicious activity earlier and cut customer losses.

AI Steps Into Singapore’s Fight Against Bank Scams

Singapore’s Monetary Authority is working with five banks, the Government Technology Agency and the Singapore Police Force to explore how AI and machine learning can improve scam detection across the financial sector.

The effort is part of a Proof-of-Value exercise designed to test whether combining data across institutions can produce more accurate models for spotting high-risk transactions and accounts.

The aim is early identification, allowing banks to respond faster before losses escalate.

MAS said the initiative is intended to strengthen existing fraud prevention systems used by individual banks by adding a wider, industry-level view of transaction patterns.

How Will Data From Five Banks Be Used?

The project will draw on historical transaction records from five participating banks, including bank account numbers, to train and assess AI models.

By pooling information across institutions, MAS hopes the models can detect scam patterns that may not be visible when data is analysed in isolation.

The authority noted that this broader dataset could improve accuracy in identifying suspicious behaviour and reduce missed signals.

What Does The Secure Data Setup Look Like?

To manage privacy concerns, MAS has built a controlled data-sharing environment with strict access rules and encryption measures.

The authority said “secure data sharing environment governed by policies and protocols to safeguard customer information”.

Account numbers used in the testing phase are being hashed, meaning they are converted into coded values so only the originating bank can trace them back to real accounts.

Access is limited to authorised personnel within a monitored system, and all data will be deleted once the trial concludes.

Cryptographic techniques are also being used to keep information confidential throughout the process.

Could This Change How Scams Are Caught Earlier?

If successful, the system could help banks detect suspicious activity sooner, allowing quicker intervention and reducing customer losses linked to scams.

MAS said the exercise is also intended to lay the groundwork for wider industry collaboration in using AI to tackle financial crime more effectively.

Depending on the results, the scope of the models could later be expanded to include larger datasets and additional use cases across the financial system.