The Monetary Authority of Singapore (MAS) is testing AI scam detection models that could help banks flag risks earlier by analysing transaction data from five banks.

MAS is working with banking industry partners, the Government Technology Agency of Singapore and the Singapore Police Force on a Proof-of-Value to assess how AI and machine learning can improve scam detection.

The MAS AI scam detection project will use historical transaction data, including bank account numbers, from five participating banks to train and evaluate AI and machine learning models.

MAS aims to build more accurate models by combining data from multiple banks, instead of relying only on information held by individual financial institutions.

Earlier detection could support quicker assessment and intervention, and help reduce customer losses from scams.

Data to remain protected during AI trial

MAS has set up a secure data sharing environment for the project, with policies and protocols to protect customer information.

The data will remain confidential and protected using cryptographic techniques.

Bank account numbers used in the project will be hashed, meaning only the bank that contributed the data can identify the actual account numbers.

Access to the data will be limited to authorised personnel in a controlled environment that will be continuously monitored during the Proof-of-Value.

All data used will be deleted once the project ends.

The current phase is intended to lay the groundwork for wider industry collaboration in using AI and machine learning to support financial crime prevention.

MAS may expand the scope and sophistication of the models after assessing the results, including the use of broader datasets and a wider set of use cases.

Featured image: Edited by Fintech News Singapore, based on image by Who is Danny via Magnific