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

16 financial institutions in Taiwan are promoting the FinLLM project, investing nearly 70 million yuan to build a Taiwan-specific financial large language model. By internalizing local regulations, the project aims to address pain points where general AI is prone to errors, and the first version of the bank-specific model is expected to be launched by the end of this year.

16 financial institutions join forces to develop Taiwan’s financial AI FinLLM

With the generative AI wave sweeping across the globe, general large language models often face pain points when handling specialized financial fields—namely, they are not sufficiently localized and are also difficult to align with Taiwan’s financial industry knowledge and regulations.

In response, the Financial Technology Industry Alliance announced yesterday (4/22) that it has officially launched the Financial Large Language Model project (FinLLM), bringing together 16 domestic financial institutions and combining government-industry-academia resources such as the National Development Council, the Ministry of Digital Development, and the Financial Supervisory Commission.

According to reports from 《Economic Daily》 and 《iThome》, Financial Supervisory Commission Chairman Peng Jinlong said that the financial industry is a highly regulated sector involving a large amount of complex local regulations. Currently, most off-the-shelf general large language models are trained on international data; if applied directly, there is a risk of making mistakes in how regulations are applied.

The Ministry of Digital Development Minister Lin Yijing also mentioned that when general models face specific national financial professional issues, they often quote foreign laws, resulting in incorrect information. Developing models with knowledge of Taiwan’s regulations and the ability to understand them in a localized way has become an important endeavor to ensure risk control and compliance.

Image source: Financial Technology Industry Alliance news photo. Lin Yijing, Minister of the Ministry of Digital Development, delivers remarks at the press conference for Taiwan’s financial industry AI FinLLM financial large language model

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

The Financial Technology Industry Alliance also disclosed the list of participants in the project: CTBC Financial Holding, Chunghwa Post, Taishin Financial Holdings, E.SUN Financial Holding, Taoyuan Cooperative Bank, Mega Financial Holding, First Commercial Bank, Next Bank, Cathay Financial Holding, Fubon Financial Holding, Hua Nan Financial Holding, KGI Financial Holding, Changhua Bank, Bank of Taiwan, Land Bank of Taiwan, and Taiwan Business Bank.

FinLLM development timeline: Training in May, first version by year-end

As for when the financial industry’s FinLLM will be completed, officials revealed that the project is expected to officially begin model training in May this year.

In the first phase, it will focus on the banking sector, where the regulatory and data foundations are more complete. It is expected to complete an initial version of the model in the third quarter this year, and launch the final bank-specific model by the end of this year. Then it will gradually expand into the insurance and securities sectors. 《Business Weekly》 noted that the entire project is expected to cost nearly 70 million yuan.

CTBC Financial CIO Jia Jingguang disclosed that the FinLLM project will combine the Ministry of Digital Development’s “Taiwan Sovereign AI Corpus” and Financial Supervisory Commission regulations to establish a legal training foundation. The model tuning and optimization will be handled by the local technology team Asia Pacific Intelligent Machines, and National Chengchi University will establish a standardized evaluation mechanism to determine the compliance of outputs.

The goal is to enable the system to reach the professional level of junior banking practitioners, able to handle tasks such as credit assessment and financial analysis. In the future, it will also be handled with third-party assistance to support model licensing, iteration, and the development of an application ecosystem.

Image source: Financial Technology Industry Alliance news photo. Group photo of invited guests at the Taiwan financial industry AI FinLLM financial large language model press conference

What’s different about FinLLM from current approaches?

At this stage, most banks that adopt generative AI generally use a retrieval-augmented generation architecture.

Jia Jingguang pointed out that the current approach is to build a knowledge base outside the general model, so the model queries data in real time and then generates answers. Although this can reduce the error rate to some extent, information can be missed during the process of splitting data for retrieval. And when the amount of knowledge increases substantially, the project will face technical bottlenecks such as decreased query efficiency and unstable responses.

The difference between this jointly developed dedicated FinLLM for the financial industry and the previous retrieval-augmented architecture is that Taiwan’s financial regulations and industry knowledge are directly internalized into the model. The system does not need to rely on external queries; it can directly understand financial logic and generate answers, clearly improving response completeness and reasoning and analysis capabilities.

This is also an important step for Taiwan’s financial industry after the AI basic law went into effect and the Financial Supervisory Commission issued guidelines for AI applications in finance.

In the future, AI models used in finance are expected to adopt a hybrid approach: a localized model trained in-house as the core, supplemented by external knowledge bases to provide the latest real-time information, and safeguarded through a human-AI collaboration model to review decisions—driving an upgrade in the overall quality and efficiency of financial services.

Further reading:
Central News Agency reports on what happened next after the NTU students’ lawsuit! They created Traditional Chinese datasets for AI involving alleged infringement; both sides have reached a settlement

People go crazy for raising lobsters! The Ministry of Digital Development: AI agents will definitely be integrated into public service; Foxconn is interested in investing in Taiwan’s computing power