The State Bank of Vietnam's SIMO fraud monitoring system has warned 6 million customer instances this year, and more than 2 million customers stopped or cancelled a transaction after an alert, keeping roughly 7,600 billion VND in their accounts, Le Anh Dung, deputy head of the central bank's payment department, said at the Smart Banking 2026 conference on 24 October. Thi Truong Tai Chinh Tien Te reported the figures on 25 October.
Identity checks now cover nearly 177 million individual customer records and 3 million organisation records, the same official said. Cashless transaction volume grew more than 35% in the first 8 months of 2026 against the same period last year, internet payments rose more than 46%, mobile payments 35%, and QR payments 7% by count and 26% by value. Those figures extend a series The-Delta.io has tracked all year. In April the central bank reported 156.6 million verified records, covering 154.5 million individuals and 2.13 million corporate accounts, and its first quarter comparison put QR growth at 16.43% by count and 52.4% by value. The 8-month release reports different windows and different bases, so the 2 sets are not directly comparable, and the payment department did not explain the gap.
Cross-border QR links are now two way with Thailand, Laos and Cambodia, and visitors from China, South Korea and Singapore can pay domestic merchants through connected schemes. Speed is the constraint that comes with it. Dung said a transfer clears in seconds, which is also the window in which defrauded money can be pushed through several intermediary accounts, and he listed 4 principles for the banking sector: safety designed into the product, identity as the base of trust, shared risk data between institutions, and user-centred innovation.
A newer problem sits inside the third principle. Once AI agents execute payments on a person's behalf, Dung said, banks have to establish the identity of the agent, who authorised it and what scope it may transact within. Do Duc Huy, a senior ASEAN consultant at Palo Alto Networks, told the same conference that more than 90% of the attacks surveyed involved identity or access credentials, and that machine identities in enterprise systems can outnumber human ones by more than 100 times. His recommendation was to remove standing privileges and put time limits on access grants.
What the release does not settle is how much of the 7,600 billion VND would have been lost without the warning rather than diverted into a second attempt, and how the SIMO thresholds interact with the deepfake and 24-hour delay rules the central bank scheduled for 2027.