Central bank moves to tighten AI use in banking, requires prior customer notification

The State Bank of Vietnam (SBV) has moved to tighten oversight of artificial intelligence (AI) in the banking sector, requiring lenders and e-wallet providers to notify customers in advance when AI systems are used to interact directly with them.

The proposal is included in a draft circular on safety, risk management and conditions for AI deployment in banking activities, released as financial transactions surge, especially during Tet (Lunar New Year) holiday.

Under the draft, banks and intermediary payment service providers must inform customers before deploying AI-powered tools such as virtual assistants, automated hotlines and chatbots in customer communications.

Institutions must also notify users if AI is applied for emotion recognition or biometric classification. Any content generated by AI, including images, audio or video, must be clearly disclosed to prevent misunderstanding.

Notably, the draft prohibits banks from using AI to exploit customer vulnerabilities linked to age, disability or financial hardship in order to market high-risk or unsuitable financial products.

Customers will have the right to file complaints against decisions made on the basis of AI systems. In such cases, banks are required to assign human personnel to review and handle the matter.

The new regulations are expected to take effect from March. AI systems already in operation before the effective date will have until September 2027 to comply.

The central bank said the tighter framework is aimed at controlling risks arising from the rapid expansion of AI across banking operations, ensuring fair treatment of customers and protecting vulnerable groups.

In recent years, many Vietnamese banks have stepped up the use of AI in areas such as eKYC, credit scoring, loan appraisal and fraud detection.

The growing use of generative AI in customer service, including answering queries on transactions, card services, interest rates and exchange rates, has raised fresh regulatory concerns over data security, output accuracy and privacy protection.

- (VNA/VLLF)