MAS partners finance industry players to develop safeguards for AI agents

[SINGAPORE] The Monetary Authority of Singapore (MAS) on Friday (Jul 3), along with financial institutions and fintech companies, published an industry white paper on developing safeguards for AI agents in finance.

The paper proposes an industry-developed framework that enables AI agents in the financial services sector to carry out financial tasks securely, reliably and safely.

Also known as Safeguards for Agentic Finance at Runtime (SAFR), it provides for a set of governance checkpoints that verifies and records an AI agent’s proposed actions before the execution of its tasks.

The move comes as AI agents in financial services increasingly carry out tasks autonomously. This means businesses require “real-time safeguards” to ensure that the behaviour of these agents remain within predefined mandates, policies and risk boundaries set by the financial institution.

The paper includes industry case studies of how the SAFR framework operates across different financial domains. They involve the likes of Ant Financial, OCBC, Visa and Mastercard, and cut across the areas of corporate banking, wealth management and insurance.

Industry members have applied the framework in client engagement, as well as agent-assisted payments and treasury operations.

The framework was developed under MAS’ BuildFin.ai initiative – a programme which supports the “responsible development and deployment of AI solutions in the financial sector”, said the central bank on Friday.

MAS said that interested industry partners are invited to join the BuildFin.ai work group to contribute to and help shape subsequent iterations of SAFR.

Earlier in May, DBS CEO Tan Su Shan said in an interview with The Business Times that team agents and enterprise agents – which support complex agentic workflows – are being trialled at the bank’s offices in Singapore, Hong Kong and elsewhere.

Different AI models and agents in the Singapore bank are also deployed in the legal and compliance function, which involves real-time fraud detection for anti-scam efforts and in drafting suspicious transaction reports.

Others negotiate general terms and conditions with third-party suppliers, and systems that classify corporate customers into anti-money laundering risk categories.

DBS chief data and transformation officer Nimish Panchmatia said the bank unlocked S$1 billion in AI initiatives last year.