Eigen RadarAI
Analysis

Singapore plans a way for banks to check what AI agents are allowed to do

Singapore's central bank plans to develop a common protocol for checking whether an artificial intelligence agent has a customer's permission to act. A bank could see whom the software represents and what instructions and limits the customer set. The authority plans to test the approach with corporate treasury tasks and payments. Banks would decide which instructions to accept.

Artificial Intelligence··Evening
A bank authority-verification console and cabled hardware permission interlock checking an AI agent's delegated payment limits inside a glass bank vestibule in Singapore's Raffles Place.

Banks would know whose instructions an agent carries

The Monetary Authority of Singapore, the country's central bank and financial regulator, wants banks to have a common way to check the authority of artificial intelligence agents. Its managing director, Chia Der Jiun, outlined the plan on November 19 at the Singapore FinTech Festival. The proposed protocol would tell a financial institution which customer a piece of software represents and which actions that customer has authorised.[1], [2]

Chia said some financial institutions and payment providers are developing ways for agents to access their services. Many institutions still lack dedicated channels for that access.[1]

The authority plans trials in payments and treasury

In a planned treasury trial, a company could ask an agent to monitor its available funds and keep track of upcoming payments. Participating banks would set their own controls and choose which instructions to accept. In the payments trial, a bank or payment provider would check who the agent is and what it can do before allowing a transaction.[1]

Chia calls for tighter controls as discretion grows

Chia asked who would bear responsibility if an agent made a mistake or exceeded the customer's instructions. He called for stronger controls when software has more discretion or can make decisions with greater financial consequences. The proposal also envisages sharing information about an agent's reliability with the institution asked to act.[1], [2]

References

  1. News sourceThe Straits TimesSingapore plans a protocol for banks to verify AI agents’ authority↩1↩2↩3↩4
  2. News sourceFintech News SingaporeSingapore plans a protocol for checking AI agents' authority in finance↩1↩2