Land Bank of Taiwan Partners with Financial Institutions and Taiwan AI Labs to Enhance AI-Powered Fraud Detection

Financial sector AI fraud prevention is being upgraded. Land Bank of Taiwan has partnered with fellow financial institutions and Taiwan AI Labs to advance a joint AI anti-fraud model training initiative. Eight banks are contributing over 26 million transaction records and more than 50 risk indicators to train the AI model, dramatically enhancing the ability to identify suspicious accounts and setting a new milestone in financial fraud prevention.

The eight participating banks—Land Bank of Taiwan, Taichung Bank, Taipei Fubon Bank, Cooperative Bank of Taiwan, First Bank, Sunshine Bank, Cathay United Bank, and Changhua Bank—launched the joint learning program in 2024 and have now formally allied with Taiwan AI Labs. Leveraging complementary strengths in financial expertise, data analytics, and information architecture, they are co-developing a shared AI model for the financial industry.

This AI model employs federated learning, enabling banks to collaboratively train the model without exchanging raw customer data. This approach ensures both data privacy and information security while allowing banks to share fraud detection insights and risk patterns, thereby improving the identification of suspicious accounts and emerging fraud tactics.

As a 100% state-owned public bank, Land Bank of Taiwan actively supports the Financial Supervisory Commission’s fintech innovation and anti-fraud collaboration policies, fully implementing the government’s five-pronged strategy: 'detect, block, prevent, stop, and punish' fraud.

In terms of anti-fraud modeling, Land Bank has deepened its risk management and fraud prevention applications by collaborating with prosecutors, police, financial information companies, and peer institutions. It shares suspicious transaction patterns and risk parameters, establishes real-time alert and information-sharing protocols, and plans to integrate with the Financial Information Services Co.'s 'preliminary fund flow inquiry mechanism' and 'account fund flow early warning system.' Through inter-agency data exchange and real-time alerts, fraud interception efficiency is further enhanced.

Internally, Land Bank employs anti-money laundering systems for monitoring and reporting suspicious transactions, implements high-risk account opening review procedures for individuals and entities, establishes in-branch care mechanisms for potential victims, introduces foreign national risk data for enhanced identification, and sets up monitoring and supervisory systems for employee transaction behaviors. By combining technological surveillance with institutional controls, the bank effectively reduces risks related to mule accounts and financial crimes.

In terms of results, according to statistics from the Ministry of the Interior’s National Police Agency, Land Bank prevented 163 fraud cases in 2025, avoiding losses exceeding NT$130 million. From January to May 2026, it successfully blocked 31 fraud cases, preventing losses of NT$27 million.

Land Bank has long been committed to in-branch vigilance and fraud prevention, earning recognition through awards such as the 'Anti-Fraud Award' from the Chinese Banking Association and the 'Anti-Fraud Collaboration Contribution Award' from Financial Information Services Co. Moving forward, the bank will leverage joint learning to upgrade its fraud detection models, building a comprehensive financial security defense system based on 'prevention-first, real-time alerts, rapid interception, and full-cycle protection' to safeguard customer assets 24/7 and deliver safer, more trustworthy financial services.