MAS Unveils SAFR Guardrails for Financial AI Agents

Financial institutions are moving AI beyond recommendations toward autonomous agents capable of initiating payments, trades and compliance workflows, often too quickly for humans to approve each step. That shift creates operational and accountability risks because conventional controls typically assess models before deployment or review transactions after completion. The Monetary Authority of Singapore developed SAFR to place a real-time governance checkpoint between an agent’s proposed action and the financial system that executes it.

MAS published the Safeguards for Agentic Finance at Runtime white paper on July 3, 2026, with eight industry participants: Ant International, Circle, HSBC, J.P. Morgan Chase, Manulife, Mastercard, OCBC and Visa. SAFR checks identity, authority, policies and risk limits before payments or trades proceed, while preserving an audit record. Its disposition engine can allow automatic execution, observe an action, escalate it for review or deny it. The framework is an industry reference rather than supervisory guidance.