Singapore’s Central Bank to Develop a ‘Know Your Agent’ Protocol for AI Agents Accessing Financial Services

The Monetary Authority of Singapore (MAS) will facilitate the development of a common “know your agent” protocol so that banks and other financial institutions can verify AI agents acting for their customers, managing director Chia Der Jiun said in a keynote at the Singapore FinTech Festival. The protocol is part of a new phase of Safeguards for Agentic Finance at Runtime (SAFR), an MAS initiative on how AI agents can transact safely with financial institutions. The government is separately exploring whether the domain name system can identify AI agents as it weighs a national registry for them.

Chia said some payment service providers and financial institutions are building agent interfaces, but many have yet to set terms or channels for agents to access financial services. Some users are instead giving agents their login credentials. “I would not advise giving personal agents such access, as there is insufficient assurance of safety,” he said.

He set out the problems a financial institution faces when an agent arrives: it does not know whether the agent is acting with the customer’s full authorization and control, or whether it is transacting within its mandate. “And if agents transact incorrectly or beyond their instruction, who bears responsibility?” he asked.

The protocol is meant to support trusted information exchange between customer agents and financial institutions, covering who the agent is and whom it represents, what authority it has been given, what mandate it operates under, and what assurance exists about its provenance and reliability. Chia said this would give institutions the information needed to set the terms and scope permitted to an agent and to establish a dedicated secure channel for agent access.

The approach draws on lessons from the first phase of SAFR: agents need bounded autonomy, their identity and authority to act must be clearly established, mandates must specify permissible actions, limits and conditions, and controls must be proportionate to the authority given. “The greater the discretion and impact of action given an agent, the stronger the assurance and controls needed,” Chia said.

MAS will also set up a sandbox with the Future of Finance Institute that combines AI agents with programmable money. In one use case, banks could let a corporate agent monitor liquidity and execute permitted treasury actions with tokenized deposits. In the other, banks and payment providers would verify a customer agent’s identity and authority and apply risk controls before executing payment instructions in tokenized deposits or stablecoins.

Sources: Monetary Authority of Singapore, The Straits Times

–

By the ID Tech Editorial Team