Taiwan's FSC survey finds 43% of 383 financial institutions use AI, and the FSC says it will draft risk-based AI rules when appropriate

Key takeaways

• The FSC says 164 of 383 surveyed institutions (43%) use AI, up from 29% in 2024.
• Among AI users, generative AI adoption rose from 27% to 76%; 33 firms use agentic AI, which the FSC describes as early stage.
• The FSC says it is inventorying financial AI use cases and potential risks and will draft risk-based management rules when appropriate. It states no draft or date.
• Firms named technology and model risk, data security and privacy, and compliance as the main challenges.

What did the FSC survey find about AI use in Taiwan's financial sector?

The FSC says 164 of 383 institutions, 43%, now use AI. The survey covered all financial institutions and peripheral units in the banking, securities and futures, and insurance sectors. Among individual institution types, the FSC says domestic banks lead at 87%, followed by life insurers at 65% and securities investment trust enterprises at 56%.

The release gives the survey years as ROC years 113 and 115, which correspond to 2024 and 2026. It also says 214 firms, 56% of those surveyed, plan to adopt or expand AI, with internal administration, intelligent customer service, risk management and legal compliance named as priorities.

What do the results say about generative AI, agentic AI and purposes of use?

Generative AI is the headline figure. The FSC reports adoption among AI-using firms rising from 27% to 76% between 2024 and 2026. The release says use remains concentrated in internal administration and intelligent customer service, and that the top two purposes are higher operating efficiency and labour saving.

The third purpose changed. In 2024 and 2025 it was improving customer experience; in 2026 it is improving analysis and decision-making. On that basis the FSC says AI is gradually becoming a tool that assists operational decisions. That reading is the FSC's, drawn from the ranking shift, and the release gives no further detail on it.

The survey added agentic AI for the first time. The FSC says 33 firms use it, mainly for data collection, research report drafting and code checking in internal work, and that it is at an early stage. Firms generally named technology and model risk, data security and privacy protection, and compliance as the main obstacles. The FSC adds that a Cambridge Centre for Alternative Finance report on global AI in financial services found a broadly similar pattern.

Does the release create rules, and what does the Article 16 statement mean?

No rules are created. The release is a press statement, and the FSC does not publish draft text, a consultation or a timetable in it. It says the AI Basic Act took effect in January 2026 and that the FSC is acting under the Act's Article 16. The FSC's words are: "盤點金融業AI應用場景及潛在風險，並適時研訂以風險為基礎之管理規範".

In plain terms, the FSC says it is inventorying financial AI use cases and potential risks and will draft risk-based management rules when appropriate. The release says the aim is to guide institutions to use AI to improve service quality safely, while promoting innovation. The release does not say what the rules would cover, which institutions they would reach, or whether they would be binding. We did not open the text of Article 16, so we do not describe its content. Our article on the AI Basic Act covers the Act itself, and our piece on FSC financial AI guidelines covers an earlier FSC instrument.

What should Taiwan financial institutions do now?

Nothing in the release requires action. The steps below are the desk's suggestions for a firm that expects the FSC's draft rules.

• List every AI use case, including generative and agentic tools, with the business unit that owns it.
• Note for each use the data it touches and the risk category you would assign, so your inventory can be compared with what the FSC later publishes.
• Check your position against the three challenges firms reported: technology and model risk, data security and privacy, and compliance.
• Watch the FSC's news releases for a draft or consultation, since the release gives no date.

What are the key dates and parties?

What we did not verify

What we opened: the FSC press release page in Chinese, including its dated header, body text and survey-scope note, on 2026-10-25.

What we did not open: the two attachments (the survey results summary and the list of AI uses firms are willing to describe publicly), the text of Article 16 of the AI Basic Act, the AI Basic Act itself, and the Cambridge Centre for Alternative Finance report the release cites. Percentages are the FSC's own and we have not recomputed them from underlying data. The survey method and response rate beyond the 383 institutions stated are not described in the release body we read.

What we refuse to claim: that the FSC has issued or scheduled any rule, or what any future rules would contain or whom they would bind, or that the survey results change any existing duty.

Related briefings

• AI Regulation News hub
• Asia-Pacific AI regulation tracker
• CPAs and finance professionals and AI regulation
• Taiwan FSC financial sector AI guidelines
• Taiwan AI Basic Act framework duties
• Taiwan banking AI accountable officer and bot disclosure
• Taiwan insurer AI underwriting and claims self-regulation