
Singapore has set out a plan to use shared AI capabilities to detect financial scams, improve small-business lending and test transactions carried out by AI agents. The Monetary Authority of Singapore and the Future of Finance Institute announced the initiatives at the Singapore FinTech Festival on 18 November.
The most immediate challenge is that scam payments can move between banks quickly. One institution may see only a fragment of the activity. Singapore’s plan is to explore whether shared information and AI models can identify patterns across those boundaries earlier, while preserving controls around the data involved. This is a test programme, not a claim that financial fraud can be eliminated by deploying a model.
The Future of Finance Institute will work with the regulator on an AI Innovation Arena. Its initial dataset is synthetic but designed to reflect interbank transaction and scam patterns drawn from a proof-of-value involving five banks. Developers and researchers can use it to build and evaluate proposed detection systems without being handed raw customer transactions. The regulator says it will begin a design study for a broader scam-detection utility next year.
Synthetic data is useful for experimentation, but a model that performs well in a controlled setting may struggle with real-world false positives, changing criminal tactics or gaps in the participating institutions’ coverage. Any eventual live system would need clear governance, a way to challenge mistakes and safeguards against unnecessary data sharing.
A second mission looks at whether richer information from accounts, invoices, payments and business records can make lending decisions more useful for small and medium-sized enterprises. Some young firms lack the long credit histories or collateral that traditional assessments favour. Singapore intends to invite technology firms to test ideas with realistic synthetic data before deciding what infrastructure could support wider deployment.
The third mission is more experimental. The institute plans an Agentic Money Sandbox where financial institutions can test AI agents making payments or managing treasury tasks within defined authority and limits. The regulator stresses that institutions and businesses remain accountable for actions performed by agents on their behalf. Initial trials will look at the controls needed before any such activity enters live financial environments.
Singapore’s plan is notable for its sequence. It begins with shared data and controlled testing, then asks whether the tools are effective and governable enough to scale. For banks and fintech firms elsewhere, including in Africa, the useful lesson may be less about copying a particular AI model than about creating a framework in which institutions can collaborate without surrendering accountability.







