Taiwan has begun a project to create a large language model for the financial sector to support domestic institutions, reported Bloomberg.  

The move is intended to avoid the constraints of overseas AI platforms that may not fully capture local rules and market practice. 

Access deeper industry intelligence

Experience unmatched clarity with a single platform that combines unique data, AI, and human expertise.

Find out more

The initiative brings together 16 financial institutions in Taiwan. The lenders have formed a task force under the Taiwan FinTech Alliance to advance the work. 

In a statement, the group said it wants to build AI infrastructure “best suited to Taiwan’s financial industry”. 

CTBC Financial Holding will lead the project, while the Financial Supervisory Commission will supervise it.  

The planned model will be built on an open-source foundation rather than Chinese platforms. 

Training will use specialised data sets, including sector-specific information and regulatory material supplied by the FSC. 

It will also draw on the Digital Ministry’s Sovereign AI database so that the system reflects local language use and cultural context. 

The cost is estimated at between T$40m ($1.3m) and T$70m. Training is due to start in May.  

A prototype is planned for the third quarter, with the completed version expected before the end of the year. 

The effort comes as large international banks including JPMorgan Chase & Co. and DBS Group continue to benefit from early use of AI. 

CTBC Bank chairman James Chen said a gap is appearing between “AI-ready” companies and those falling behind, with the first group recording stronger revenue growth and market valuations. 

The Taiwan FinTech Alliance said general-purpose models often run into problems in financial use, including weak localisation and difficulty matching domestic knowledge. 

The first stage will centre on banking. The project is later expected to extend to insurance and securities. The alliance said it plans to introduce a FinLLM-based AI agent in the first quarter of 2027 to help with complex financial work and customer service needs.