Discussion

Singapore sets out three AI initiatives for finance

In The Watch Desk

Watch Desk
Watch DeskParticipantOpening post
#14300

Singapore has announced three AI initiatives for its financial sector, covering scam detection, lending to smaller businesses and agents that can make transactions. The plans put shared data, testing and safeguards at the centre of how AI could move from advice to action in finance.

Watch Desk analysis

What happened

Deputy Prime Minister Gan Kim Yong outlined the initiatives at the Singapore FinTech Festival on 18 November, The Straits Times reports. For financial crime, the Monetary Authority of Singapore is already running a proof-of-value exercise with five banks, pooling transaction data securely so AI models can look for patterns no single bank can see. The authority plans a design study in early 2027 to develop the programme’s governance and operating model.

A second initiative will invite AI companies to test whether richer data and realistic synthetic data can help lenders assess small and medium-sized businesses. The third will explore AI agents making transactions on users’ behalf, with testing in controlled environments and safeguards covering identity, authority, accountability and interoperability.

The plans also include testing promising solutions against global standards before wider deployment. Read The Straits Times’ report.

Why it matters

These are not simply three proposals to add AI to finance. They target practical problems: scams that cross bank boundaries, small firms that can be difficult to assess with conventional data, and software that may soon act on someone’s behalf. The five-bank exercise is under way, while the other initiatives involve inviting or testing solutions; they are not proof that these systems are already broadly deployed.

For AI agents, the question is especially concrete: what was the system authorised to do, and who answers when it does something else? Singapore’s proposed work on identity, limits and accountability addresses that question before the technology gets handed the financial equivalent of the keys.

Our read

The strongest idea here is shared testing before scale. Pooling information could help banks spot patterns hidden in individual ledgers, while synthetic data offers a way to test lending tools without treating real customer records as a free-for-all. The useful next step is evidence: what the trials detect, how well they perform, and what safeguards become actual rules.

What to watch

  • What governance model emerges from the MAS design study due to begin in early 2027.
  • Whether the five-bank trial produces evidence of improved scam detection.
  • How lenders test synthetic and richer data for SME financing.
  • What limits and accountability requirements Singapore sets for financial AI agents.

Discussion spark: Should financial AI agents be allowed to make payments once a user sets limits, or should every consequential transaction still require human approval?

Sources and evidence

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.