AI increasingly integrated into banking

According to the central bank, nearly 89% of people aged 15 and above had payment accounts by the end of 2025, while digital transactions accounted for over 95% of transactions at many credit institutions. In the first half of 2026, cashless payment transactions rose 34.28% in volume and 12.24% in value year on year.

Banks are increasingly applying AI in data analysis, personalised services, fraud detection and risk management. In August, the Joint Stock Commercial Bank for Foreign Trade of Vietnam (Vietcombank) launched its One Data Platform (ODP) project to support advanced analytics and lay the groundwork for AI and machine learning applications.

Meanwhile, the Tien Phong Commercial Joint Stock Bank (TPBank) combines AI, big data and real-time analytics to detect unusual transaction patterns. TPBank General Director Nguyen Hung said these tools help the bank identify potential risks before they result in losses.

Maintaining human control

At the 2026 Banking Sector Digital Transformation on August 18, SBV Governor Pham Duc An called for focused and controlled innovation amid growing challenges involving cybersecurity, digital fraud, privacy and AI governance.

The circular, which will take effect from November 15, stipulates safety, risk management and implementation requirements for the application of AI systems in the banking sector.

Notably, it classifies AI systems by risk level, including as high-risk those that automatically execute payments exceeding specified thresholds or make automated credit decisions involving 50 million VND or more. Payment risks are assessed based on individual transactions and total daily values.

AI-driven payments require prior agreements defining transaction scope, limits, duration, and conditions. Customers must be able to terminate agreements or suspend or revoke payment authorisation. Daily AI-executed transactions must not exceed internal limits or 100 million VND for individuals and 500 million VND for organisations.

Providers must continuously monitor high-risk systems and ensure authorised personnel can intervene, adjust or suspend AI decisions, or activate emergency shutdowns. All interventions must be documented.

When AI directly interacts with customers, providers must disclose this from the outset. They must also retain the rationale and key factors behind automated decisions affecting customers’ legal responsibilities, financial obligations or access to services, and explain them upon request.

Data used to train, test and operate AI must have clear and lawful origins. Providers must address inaccurate, manipulated or “poisoned” data and prevent data leaks.

The circular underscores the need to combine automation with accountability, data protection and effective human oversight./. VNA