16 компаний финансового холдинга сотрудничают, чтобы создать ИИ-модель большого языка, разбирающуюся в тайваньских финансах

under guidance of the Financial Supervisory Commission, China Trust Financial Holding Co. convened, and 16 major financial holdings and banks in Taiwan jointly launched the “Financial Large Language Model FinLLM” project, based on the sovereign AI corpus developed by the Data Development Department to create localized financial AI.
（Background: The Taiwanese government is actively布局"sovereign AI", National Science and Technology Council Chairman Wu Chengwen: forming a national artificial intelligence strategic special committee）
（Additional background: Taiwan’s 11 leading insurance companies have formed the “Claims Alliance Blockchain”！Multiple claims from a single policy, making the application process easier）
16 financial holdings, 2 state-owned banks, plus the Financial Training Institute, Institute for Information Industry, NCCU Financial Technology Research Center, jointly aim to realize a goal: Taiwan should have its own financial large language model.
This project is called FinLLM, initiated by the financial technology industry alliance guided by the Financial Supervisory Commission, with China Trust Financial Holding as the convenor, officially launched today (23rd) at the FinTechSpace in the Financial Technology Innovation Park.

From “individual R&D” to “collective collaboration”

In the past, each financial holding wanting to develop AI had to collect data, train, and evaluate on their own. High computing costs, scattered data licenses, compliance risks borne individually, resulting in each doing just enough, burning a lot of money, but with few practical application scenarios.
This time, the structure is different. The 16 participating institutions include China Trust Financial Holding, Cathay Financial Holding, Fubon Financial Holding, Taishin Shin Kong Financial Holding, E.SUN Financial Holding, Union Bank, Mega International Commercial Bank, First Financial Holding, Hua Nan Financial Holding, KGI Bank, Changhua Bank, Bank of Taiwan, Land Bank, Taiwan Cooperative Bank, Chunghwa Post, and Future Bank, almost all major players in Taiwan’s financial industry are brought into one project.
Plus research and technical units like the Financial Training Institute, Institute for Information Industry, NCCU Financial Technology Research Center, and Asia-Pacific Intelligent Machines, FinLLM aims to connect the entire process from data licensing, model training, evaluation, to business model operation.
More importantly, the legal risks of data licensing can also be shared collectively, which is often more challenging for the financial industry than computing costs.

The foundation is the sovereign AI corpus built by the Data Development Department

FinLLM is not a model trained from scratch. Its foundation is the sovereign AI corpus currently being built by the Data Development Department.
Sovereign AI corpus. Translated, it is a national-led collection of local Taiwanese language, regulations, and document data, serving as the training basis for domestic AI models. This project has been included in one of the top ten national AI infrastructure initiatives, with a high status.
The role of the alliance is to further feed this general foundation with financial-specific corpora: Taiwanese Financial Supervisory Commission regulations, corporate governance codes, bank operation manuals, financial product descriptions, risk disclosure documents.
Even the most powerful US-based large models may not fully address the context of Taiwanese financial regulation. For example, asking GPT about a provision of Taiwan’s Financial Consumer Protection Act, it might give a general explanation, but when it comes to Taiwan-specific concepts like “offshore structured products” or “elderly customer suitability review,” the accuracy may drop.
The alliance trains with legally authorized data, and jointly reviews by member banks to establish standardized evaluation criteria.