Singapore's central bank lines up a sandbox for AI agents that manage company cash

Singapore's central bank lines up a sandbox for AI agents that manage company cash

The Monetary Authority of Singapore (MAS) plans a sandbox in which AI agents do corporate treasury tasks with tokenized deposits, within limits set by the company and its bank.

MAS Managing Director Chia Der Jiun laid out the Agentic Money Sandbox at the Singapore FinTech Festival on Nov 19.

The MAS and Future of Finance Institute announced on November 18 that the institute will build the sandbox. It melds AI agents with programmable money in two use cases, corporate treasury management and payments.

MAS gives companies the mandate and banks the controls

In the treasury case, banks could permit a corporate agent to monitor liquidity position and due payments. Chia’s keynote explained that the agent decides when and where funds should move and then uses tokenized deposits to execute permitted treasury actions.

The mandate is given by the company. The controls are set by the participating banks along with the terms of acceptance of each agent instruction.

AI agents to run corporate treasury tests in Singapore sandbox.
The Agentic Money Sandbox section of Chia Der Jiun’s keynote at the Singapore FinTech Festival on November 19, 2026.

On the payments track, customer agents would execute payment instructions with digital money, like tokenized deposits or stablecoins. Before the transaction is executed, the bank or payment provider verifies the identity and authority of the agent, and applies risk controls.

A controlled environment will be created for financial institutions to test how agents transact within set authority, limits and safeguards before any live deployment, according to the joint release. The launch partners for the institute were DBS, OCBC, Standard Chartered and UOB.

“Businesses see benefits for agents to manage cash, payments and liquidity,” said Chia. He added that large banks already extend tokenized deposits to corporate customers for treasury management.

SAFR protocol would tell banks whom each bot represents

Chia tied the sandbox to a know-your-agent protocol. In the next phase of its work on Safeguards for Agentic Finance at Runtime, or SAFR, MAS will enable this.

The protocol would give financial institutions four pieces of information: who the agent is and who the agent is acting for, the authority the agent has, the mandate under which it is operating, and the assurance there is about the origin and reliability of the agent. Banks would use it to delineate an agent’s terms and scope, and to open a dedicated secure channel for agent access.

The greater the discretion and impact of an agent, the greater the assurance and controls, under lessons from SAFR’s first phase.

“Some users are giving agents their login credentials to financial accounts. I would not advise giving personal agents such access, as there is insufficient assurance of safety,” said Chia. He also asked who is liable if agents transact incorrectly or not within their instructions.

Separately, Visa, Mastercard and Ant International are formulating a Know-Your-Agent interoperability framework through BuildFin.ai, a platform launched by MAS. Card networks, wallets and agent platforms could use it to identify trusted agents across systems.

This month, MAS issued S$100 million of tokenized MAS Bills and settled them against Singapore dollar wholesale CBDC. SGD Testnet will become a working utility called SG Digital Ledger.

As part of the MAS BLOOM initiative, StraitsX, OCBC and Standard Chartered completed US dollar payments across two separate tokenized deposit networks in minutes, Cryptopolitan reported.

This story was originally published by Cryptopolitan. Read the original article.

More Tech & AI News