. As artificial intelligence (AI) agents move from systems that support employees to autonomous decision-making and working, they need to be kept under the same strict surveillance and control as bank users.
The Safeguards for Agentic Finance at Runtime (SAFR), released by the Monetary Authority of Singapore (MAS) last July, has put forward such an approach to managing the risks associated with the use of autonomous AI agents in the financial sector.TAG_OPEN_p_8
SAFR, which was developed as part of BuildFin.ai initiative in collaboration with banks and fintech companies, has indicated that AI agents should not be seen as just a general software system. As AI agents begin to operate autonomously in areas such as payments, loan approval, fraud management, and customer service, there will be an impending need for stronger control over their decisions and transactions.
Financial institutions have been asked to verify and document the AI agent in advance before doing any important work. Under this, the agent has to check the proposed work before starting a payment, approving a loan, conducting a transaction or submitting a report to the regulator.
As the use of AI agents continues to expand rapidly, SAFR sends a message to financial institutions to prioritize the speed and capabilities of autonomous AI as well as its risk control, accountability, and monitoring.
