A new Monetary Authority of Singapore (MAS)-backed industry framework proposes tighter guardrails for AI agents acting inside financial systems.
The framework, called Safeguards for Agentic Finance at Runtime, or SAFR, sets out how financial institutions can authorise AI agent actions, trigger human review and record decisions before actions are executed.
SAFR was developed by MAS, leading financial institutions and fintechs under MAS’ BuildFin.ai initiative, which supports the responsible development and deployment of AI solutions in the financial sector.
It is aimed at AI agents that can take action on their own, rather than only produce recommendations for human review.
These agents may be used to initiate payments, submit trading orders, approve credit applications, file regulatory reports or settle insurance claims.
The paper notes that existing governance processes were largely built for human decision-making and may not be enough for autonomous systems operating at high speed and scale.
Checks Before Execution
SAFR is designed to sit between an AI agent and the systems it acts on.
Before an action goes through, the framework checks the agent’s identity, the authority given to it, institutional controls and risk thresholds.
The framework provides governance checkpoints that verify and record an AI agent’s proposed actions before execution, helping ensure they remain within predefined mandates, policies and risk boundaries.
The framework has four main components: agent identity, a controls repository, a disposition engine and an audit log.
The disposition engine determines whether an action should be approved for automatic execution, rejected, sent for human review or allowed to proceed while being flagged for monitoring.
The audit log creates a tamper-evident record of each governance decision, including the proposed action, the rules applied and the outcome.
SAFR builds on MAS’ Project MindForge AI Risk Management toolkit, with a focus on how safeguards can be applied at the point of action for AI agents.
The paper covers areas such as policy-bound execution, real-time validation, auditability and interoperability, and how these safeguards can be embedded into system operations.
Industry Pilots and Use Cases
Industry members have applied the SAFR framework across use cases including agent-assisted payments and treasury operations, wealth management and advisory workflows, and client engagement.
Insert the SAFR case-study table from pages 17 to 20 here.
The report includes use-case examples from Mastercard, Ant International, Visa, Circle, OCBC and Bank of Singapore, and Manulife.
SAFR is not regulatory guidance or a managed service. It is presented as an industry reference model that institutions can adapt to their own technology, risk and compliance systems.
Institutions can implement the framework through native integration, where the AI agent produces a governance record before each proposed action, or through a gateway model that intercepts outbound API calls from existing agents.
Interested industry partners have been invited to join the BuildFin.ai work group to contribute to future versions of SAFR.
The recently announced Future of Finance Institute will support future adoption of the SAFR framework through industry pilots and sandbox experimentation.
Featured image: Edited by Fintech News Singapore, based on image by farknot via Magnific