Economy & BusinessTechnology

Who Is This Bot, and Who Sent It? Singapore Wants AI Agents to Show ID Before Touching Your Money

Handing an AI assistant your banking password might feel like the next step in convenience. Singapore’s central bank says: not yet. Speaking at the Singapore Fintech Festival on Wednesday, Monetary Authority of Singapore managing director Chia Der Jiun urged consumers not to give AI agents the login credentials to their financial accounts, because the safeguards that would make that safe simply do not exist yet.

The problem banks cannot see

When an agent logs in with a customer’s password, the bank sees what looks like the customer. It cannot tell whether the software is acting with full authorisation, whether it is staying within the instructions it was given, or what happens if it goes wrong. Chia put the core question bluntly: if an agent transacts incorrectly or beyond its instructions, who bears responsibility? Some payment firms are already building agent interfaces, but many financial institutions have not set terms or channels for agent access at all.

An ID card for software

MAS’s answer is a common “know your agent” protocol, modelled loosely on the know-your-customer checks banks already run on people. It is part of the next phase of the regulator’s Safeguards for Agentic Finance at Runtime initiative, known as SAFR. The protocol would tell a bank who the agent is, whom it represents, what authority it has been given, what mandate it works under, and where it comes from and how reliable it is. With that information, a bank could decide what an agent may do and open a dedicated, secure channel for it, instead of letting it impersonate the customer.

The principle behind it is proportionality: the more discretion and impact an agent has, the stronger the assurance and controls around it should be.

A sandbox for programmable money

MAS is also setting up a sandbox with the Future of Finance Institute to test AI agents working with programmable money. Two uses come first: corporate treasury, where an agent could watch a company’s liquidity and due payments and move tokenised deposits within set limits, and payment services.

The takeaway for ordinary users

The message is less a ban than a sequencing argument. Agents may well end up paying bills and moving funds, but only once banks can verify them as distinct actors with clear limits. Until then, the safest assistant is one that never holds your password.

Leave a Reply

Your email address will not be published. Required fields are marked *