‘Know your agent’: Banks face a new compliance challenge as AI agents shop and pay on their own

That was the warning from Zhuoqun Bian, president of Ant Digital Technologies, the enterprise arm of Ant Group, at the Fortune Leaders Forum in Macau on Sept. 8.

“From the financial institutions’ perspective, when we initiate a transaction, we have to do the KYC [know your customer]” she said. “In the agent economy, you need to know your agents. Who’s the agent? Who does it belong to? Who authorized it?”

Agents could orchestrate as much as $5 trillion in global consumer spending by 2030, according to a January report from McKinsey.

Yet the infrastructure in which these agents operate was designed for humans. “Looking forward, all the infrastructure needs to be rebuilt or enhanced for agents,” Bian said.

On Sept. 6, Ant International, Ant Group’s global payments arm, announced it had begun collaborating with Mastercard and Visa on a “know your agent” interoperability framework, meant to let card networks, digital wallets, and marketplaces recognize trusted AI agents across ecosystems. The companies will work through BuildFin.ai, an industry platform convened by the Monetary Authority of Singapore, the city-state’s central bank.

In an April note, the International Monetary Fund observed that AI agents are probabilistic and adaptive, meaning the same prompt can yield different answers, while payment systems must return the same answer every time. “Payment rails, from card networks to real-time gross settlement (RTGS) systems, rely on predictable rules, legal certainty, and clear accountability structures to ensure trust and financial stability,” the note’s authors wrote.

For banks, the bigger question is whether their own systems can keep pace. “I personally have never come across a situation where the technology of an agent failed us—or rather, led to an undesirable outcome,” said Benson Wong, managing director and head of digital at JPMorgan Private Bank. “It’s always around the operating model, the processes, and the compliance and the controls.”

The cost of an agent getting something wrong scales with its autonomy, he added: “If you ask an agent…an information-seeking question and [it answers] wrongly, it’s not good, but you have an embarrassing moment. If you don’t get agentic workflows correct, there are vastly scaled impacts.”

“You need to know your agents much more than you know your customers,” Bian concluded.