離開
進入《TAIPEI TIMES》 Finance sector to launch localized AI
Financial Supervisory Commission Chairman Peng Jin-lung, front row, fourth right, Minister of Digital Affairs Lin Yi-jing, front row, fourth left, and other guests attend a news conference in Taipei yesterday. Photo: CNA
KEY MILESTONE: Due to complex regulations, foreign AI models cannot be applied to the financial sector, necessitating the development of a local version, an official said
By Crystal Hsu / Staff reporter
Taiwan’s financial sector yesterday launched a joint initiative to develop a localized large language model （LLM）, part of a broader push to build domestic artificial intelligence （AI） capabilities and reduce reliance on foreign technologies.
The first version, focused on banking knowledge, is set to be released by the end of this year, Financial Supervisory Commission Chairman Peng Jin-lung （彭金隆） said at the launch event in Taipei.
Subsequent phases would expand the model to cover insurance, securities and other financial services, he said.
The project is expected to deliver three key benefits, including bolstering sovereign AI by embedding local regulations, supervisory practices and industry expertise into the system, Peng said.
It would also reduce reliance on foreign-developed models such as OpenAI’s ChatGPT, Google’s Gemini and Meta Platforms Inc’s Llama, he said.
Led by CTBC Financial Holding Co （中信金控）, the initiative aims to upgrade the industry’s technological capabilities by pooling resources from participating institutions and avoiding duplicated investment. It could ultimately form a shared AI infrastructure for the local financial sector, Peng said.
In the long term, the platform is expected to support financial inclusion by extending services and knowledge to smaller financial institutions, companies, educational organizations and the general public, improving access to financial information and services, Peng said.
Overall, the project is expected to serve as a key milestone in AI adoption for the financial sector, he said, adding that banks and insurers are particularly well-suited for automation given their heavy reliance on language and text processing.
Peng added that financial services operate under complex local regulatory frameworks, meaning foreign AI models cannot be directly applied.
A localized financial LLM is therefore necessary to ensure compliance and accuracy, he said.
Peng added that the initiative has strong government backing, having discussed the project with the National Science and Technology Council （NSTC）.
Officials from the council and the Ministry of Digital Affairs （MODA） attended the launch.
The project has been incorporated into the council’s national AI development program, while the MODA is providing sovereign AI datasets to support training and development.
Peng said the initiative is only the starting point. After the banking-focused model is released later this year, future versions would expand to incorporate knowledge from insurance, securities and other financial sectors.
Participating institutions include state-backed lenders such as Bank of Taiwan Co （臺灣銀行）, Land Bank of Taiwan Co （臺灣土地銀行）, Taiwan Business Bank Co （臺灣企銀） and Chang Hwa Commercial Bank Ltd （彰化商業銀行）, as well as financial holding companies including Cathay Financial Holding （國泰金控）, Fubon Financial Holding Co （富邦金控）, TS Financial Holding Co （台新新光金控）, KGI Financial Holding Co （凱基金控） and SinoPac Financial Holdings Co （永豐金控）. Other participants include Chunghwa Post Co （中華郵政） and digital lender Next Commercial Bank Co （將來商業銀行）.
新聞來源：TAIPEI TIMES
