The Monetary Authority of Singapore (MAS), the nation’s central bank and financial regulator, has announced a new initiative designed to explore the use of artificial intelligence agents for handling key corporate treasury and payment functions using tokenized deposits. This project, termed the Agentic Money Sandbox, is set to be launched in partnership with the Future of Finance Institute.
AI agents to manage tokenized deposits
MAS Managing Director Chia Der Jiun revealed details of the sandbox during his keynote address at the Singapore FinTech Festival on November 19. According to Chia, the initiative will allow AI-powered agents to perform corporate treasury activities within specified authority and risk limits, as determined by the company and its banking partners.
Within this framework, participating banks may authorize a corporate AI agent to oversee liquidity and monitor payment deadlines. When a transaction is required, the agent is empowered to move funds and execute approved treasury actions via tokenized deposits, while always operating within the boundaries set by the institution and its bank.
The company assigns the operating mandate to its AI agent, and banks retain control over risk parameters and transaction terms to manage compliance and security concerns.
Participating companies would delegate authority to their AI agents, with banks maintaining strict oversight through predefined controls and mandate acceptance for each agent-initiated operation.
In the payments context, the sandbox will involve customer agents authorized to carry out payment instructions using digital money, including tokenized deposits and stablecoins. Before executing such transactions, banks or payment providers will verify the agent’s identity and authority, also applying necessary risk controls.
A controlled environment will be created to test these agentic transaction workflows, ensuring they conform to regulatory standards before being introduced for live operations. Initial launch partners include major financial institutions such as DBS, OCBC, Standard Chartered, and UOB.
Chia noted that businesses anticipate operational efficiencies from deploying agents to independently manage liquidity and payments. He also indicated that some larger banks in Singapore already offer tokenized deposit solutions to their corporate clients for advanced treasury operations.
The SAFR Protocol and Know-Your-Agent standards
Looking ahead, MAS plans to introduce additional safeguards through a specialized know-your-agent (KYA) protocol, named Safeguards for Agentic Finance at Runtime (SAFR). This protocol is intended to help financial institutions authenticate AI agents and monitor their activity in real time.
The SAFR protocol will provide banks with critical information about each AI agent, including the agent’s identity, whom the agent represents, its operating mandate, and assurances regarding its reliability and origin. With this information, banks can define authorized transaction parameters and ensure only trusted agents receive access to sensitive financial functions.
The protocol also emphasizes that the higher the risk and impact of an agent’s discretion, the more rigorous the control and assurance measures must be. Chia cautioned that some users are already providing personal financial account credentials to their agents, advising against such actions due to insufficient security assurances and unresolved questions about liability in case of unauthorized transactions.
In collaboration with Visa, Mastercard, and Ant International, the MAS is working through BuildFin.ai—a platform for agent identification—to develop an interoperability framework enabling trusted recognition of agents across multiple card networks, digital wallets, and agent platforms.
Mini dictionary: The SAFR protocol (Safeguards for Agentic Finance at Runtime) is a regulatory framework designed to enhance transparency, control, and risk management when AI agents interact with digital assets and banking services by providing verified identity and authority information during transactions.
Recent MAS developments in tokenized money
Earlier this month, MAS issued S$100 million in tokenized MAS Bills, settling them against the Singapore dollar wholesale central bank digital currency (CBDC). The SGD Testnet project has evolved into the SG Digital Ledger, signaling a move towards practical digital currency applications in Singapore’s wholesale financial systems.
In parallel, as part of the MAS BLOOM initiative, payment solutions provider StraitsX, along with OCBC and Standard Chartered, succeeded in executing US dollar payments between two distinct tokenized deposit networks within a matter of minutes.
| MAS Initiative | Participants | Key Outcome |
|---|---|---|
| Agentic Money Sandbox | Future of Finance Institute, DBS, OCBC, Standard Chartered, UOB | AI agents trial treasury and payment operations with tokenized deposits |
| BLOOM initiative | StraitsX, OCBC, Standard Chartered | USD payments linked across separate tokenized deposit networks |
| SGD Testnet / SG Digital Ledger | MAS | S$100 million tokenized MAS Bills settled with wholesale CBDC |
The implementation of know-your-agent protocols and tokenized money marks a significant step towards integrating AI automation into Singapore’s regulated financial system.






