The Monetary Authority of Singapore, together with financial institutions and FinTechs, has published an industry white paper setting out a framework for safeguarding AI agents in financial services at the point of execution. The paper, Safeguards for Agentic Finance at Runtime, or SAFR, is designed to help firms ensure that autonomous AI agents act within predefined mandates, policies and risk boundaries through real-time controls rather than relying on human intervention after the fact. The framework introduces governance checkpoints that verify and record an AI agent’s proposed actions before tasks are executed. Building on MAS’ Project Mindforge AI Risk Management toolkit, SAFR focuses on operationalising safeguards in live systems, including policy-bound execution, real-time validation, auditability and interoperability. Industry participants have applied the framework to use cases such as agent-assisted payments and treasury operations, wealth management and advisory workflows, and client engagement activities conducted within approved boundaries. MAS has invited additional industry partners to join the BuildFin.ai work group to shape future iterations of SAFR. The Future of Finance Institute will support adoption through industry pilots and sandbox experimentation aimed at testing and deploying SAFR-aligned solutions.
Monetary Authority of Singapore2026-07-03
Monetary Authority of Singapore publishes SAFR framework for runtime safeguards on AI agents in finance
The Monetary Authority of Singapore has published the SAFR white paper, an industry framework for runtime safeguards on AI agents in financial services. It proposes governance checkpoints that verify and record proposed actions before execution so agents operate within firms’ mandates, policies and risk limits. MAS also said future adoption will be supported through BuildFin.ai collaboration, pilots and sandbox testing.