Taiwan banks team up to build local AI! Finance’s large language model goes live by the end of the year at the fastest

Citi Financial Holdings has teamed up with a fintech industry alliance made up of 16 banking institutions. On April 22, it officially announced the “Financial Large Language Model (FinLLM)” project. The first AI model for banks and the finance sector is expected to be released in August this year, and in Q1 2026 it will also roll out an AI agent (AI Agent) based on FinLLM.

Led by Citi Financial Holdings, 16 institutions launch the FinLLM project

Taipei Times reported that the fintech industry alliance recently announced the launch of the “Financial Large Language Model (FinLLM)” initiative. Led by Citi Financial Holdings, it will be jointly carried out with Taiwan Business Bank, Land Bank, Taiwan Enterprise Bank, Changhua Bank, Cathay Financial Holding, Fubon Financial Holding, Taishin Fubon Financial Holding, KGI Fund Management, Everbank, Chunghwa Post, and Future Commercial Bank, for a total of 16 financial institutions participating together.

The project’s estimated build cost is about NT$40 million to NT$70 million. Model training will officially begin in May this year. The first bank model is expected to be released at the earliest in August. By the end of the year, the final version will be completed and the results will be published externally.

Complex regulations make it hard for overseas models to be used; localizing AI has become a necessary engineering task

The chair of the Financial Supervisory Commission, Peng Jinlong, said that the financial industry relies heavily on language and text processing, yet it is also a tightly regulated industry. It involves a large number of local regulations and supervisory procedures, so international general-purpose models such as ChatGPT and Gemini cannot be applied directly.

He emphasized that because banks have already expanded into a wide range of financial businesses, the relevant regulations and data foundation are more complete. This will help accelerate model training and deployment speed, and serve as a core basis for later expansion into areas such as insurance and securities. The plan expects to roll out an AI agent based on FinLLM in the first quarter of 2026 to further broaden application scenarios.

Peng Jinlong highlighted three major benefits: sovereign AI, shared infrastructure, and inclusive finance

Peng Jinlong said that the FinLLM plan has three core benefits. First, it will strengthen sovereign AI capabilities by internalizing local regulations and industry knowledge into the model. This will reduce reliance on overseas technologies while improving data governance and cybersecurity autonomy.

Second, it will build shared AI infrastructure. By jointly investing through 16 institutions and integrating academic research and technology resources, it can avoid duplicated investment and promote overall technological upgrading in the financial industry. Third, it will expand the spillover benefits of inclusive finance. In the future, services could be extended to small and medium-sized financial institutions, general enterprises, and education units, improving the accessibility and inclusiveness of financial knowledge.

Cross-ministry support: FinLLM is incorporated into the national AI development plan

On the government support side, the FinLLM plan has already been formally included in the “AI Top Ten New Constructions Promotion Program” of the National Development Council. The Ministry of Digital Affairs also provides a sovereign AI database as the source of training data, and officials from the National Science and Technology Council also attended the launch ceremony, showing a high level of determination to drive integration across ministries.

Citi Bank’s chairman Chen Jiawen said in this regard that companies have quickly split into two categories: “AI-related” and “non-AI.” The former has a clear lead over the latter in revenue and market value. He cited JPMorgan Chase as an example: the bank invests as much as $15 billion per year in technology and AI R&D. Royal Bank of Canada and DBS Bank have also maintained high profitability by leveraging digitalization and AI applications, showing that AI adoption has become a competitive foundation for financial institutions globally.

At this moment, Taiwan’s financial industry launching its own AI infrastructure is a key move to respond to this global trend.

This article Taiwan’s banking industry joins forces to build local AI! The fastest large-scale language model goes live by the end of the year; the earliest appearance was on Lian News ABMedia.