Taiwan's financial industry is developing its own AI! The FinLLM project has invested nearly 70 million yuan; a sneak peek at the development timeline and highlights

Taiwan’s 16 financial institutions promote the FinLLM project, investing nearly 70 million yuan to develop a Taiwan-specific financial large language model. By internalizing local regulations, it aims to solve the pain point of general AI being prone to errors, with the first version of the bank-specific model expected to be launched by the end of this year.

16 financial institutions collaborate to develop Taiwan’s financial AI FinLLM

As the generative AI wave sweeps across the globe, general large language models often face challenges in localization when handling professional financial fields, and are difficult to connect with Taiwan’s financial industry knowledge and regulations.

In response, the Fintech Industry Alliance announced yesterday (4/22) the official launch of the financial large language model project (FinLLM), bringing together 16 domestic financial institutions, along with resources from the National Development Council, the Digital Development Department, and the Financial Supervisory Commission, involving industry, government, academia, and research.

According to reports from Economic Daily and iThome, FSC Chairman Peng Jinlong pointed out that the financial industry is a highly regulated sector involving a large amount of complex local regulations. Currently, most general large language models are trained on international data, and directly applying them risks regulatory misapplication.

Digital Development Department Director Lin Yijing also mentioned that when general models face specific national financial issues, they often cite foreign laws, leading to incorrect information. Developing models with knowledge of Taiwan’s regulations and localized understanding has become an important task to ensure risk control and compliance.

Image source: Fintech Industry Alliance press photo. Digital Development Department Director Lin Yijing speaks at the Taiwan Financial Industry AI FinLLM large language model press conference.

By participating in this AI infrastructure, the financial industry hopes to shift compliance management from passive review to proactive protection, driving a comprehensive transformation of financial services and organizational operations.

The Fintech Industry Alliance also revealed the list of participating entities: CTBC Financial Holding, Chunghwa Post, Taishin Financial Holdings, E.SUN Financial Holding, Cooperative Bank, Mega Financial Holding, First Commercial Bank, Next Bank, Cathay Financial Holding, Fubon Financial Holding, Hua Nan Financial Holdings, KGI Securities, Changhua Bank, Bank of Taiwan, Land Bank of Taiwan, and Taiwan Business Bank.

FinLLM development schedule: training in May, first version of the model by year-end

When will the FinLLM for the financial industry be completed? The official announced that the project is scheduled to start model training in May this year.

The first phase will focus on banks with more complete regulations and data foundations, aiming to complete the initial version of the model by Q3 this year, and launch the final bank-specific model by the end of the year. Subsequently, it will gradually expand to insurance and securities sectors. Weekly Magazine pointed out that the entire project is expected to cost nearly 70 million yuan.

CTBC Financial’s CIO Jia Jingguang revealed that the FinLLM project will combine the Digital Development Department’s “Taiwan Sovereign AI Corpus” and FSC regulations to establish a legal training foundation, handled by the local tech team Asia-Pacific Intelligent Machines for tuning and optimization, with NCC establishing standardized evaluation mechanisms to assess compliance of outputs.

The goal is to enable the system to reach the professional level of entry-level banking personnel, capable of handling tasks such as credit assessment and financial analysis, and in the future, to be managed through third-party assistance for model licensing, iteration, and ecosystem development.

Image source: Fintech Industry Alliance press photo. Attendees’ group photo at the Taiwan Financial Industry AI FinLLM large language model press conference.

How is FinLLM different from current practices?

Most banks currently adopting generative AI generally use retrieval-augmented generation architectures.

Jia Jingguang pointed out that the current approach is to build a knowledge base outside the general model, allowing the model to query data in real-time before generating answers. While this can reduce errors to some extent, information may be missed during data segmentation and retrieval, and as knowledge volume increases significantly, query efficiency drops and answer stability becomes a technical bottleneck.

The difference with this joint development of a dedicated FinLLM is that it directly internalizes Taiwan’s financial regulations and industry knowledge into the model, eliminating the need for external queries, allowing the system to understand financial logic and generate answers directly, significantly improving response completeness and reasoning ability.

This is also an important step for Taiwan’s financial industry following the implementation of the AI Basic Law and the FSC’s guidelines for AI applications in finance.

In the future, AI models used in finance are expected to adopt a hybrid approach, with localized trained models as the core, supplemented by external knowledge bases for real-time information, and decision-making overseen through human-AI collaboration, driving overall improvements in financial service quality and efficiency.

Further reading:
Central News Agency reports on the follow-up after NTU students’ lawsuit! Creating Chinese Traditional data sets for AI involved in infringement, both sides have settled

Public enthusiasm for raising lobsters! Digital Development Department: AI agents will definitely be integrated into public services, Foxconn is interested in investing in Taiwan’s computing power