AI Regulation Tracker  /  Industry standard

Taiwan insurers adopt AI self-regulation as FSC pre-announces underwriting and claims rule changes

Taiwan's life and non-life insurance associations issued self-regulatory standards for AI systems, and the FSC's Insurance Bureau has pre-announced an amendment that would put explainability and fair-treatment duties on AI-assisted underwriting and claims. Insurers using AI to price risk or decide payouts should prepare now.

By Anthony Guerriero  ·  Reviewed by The Leveraged Years Editorial Desk  ·  Published July 7, 2026  ·  Last updated July 7, 2026

Taiwan has moved insurers into the same AI accountability structure it built for banks and securities firms. The Life Insurance Association and the Non-Life Insurance Association have issued "Self-Regulatory Standards for Insurers' Use of AI Systems," and the Financial Supervisory Commission's Insurance Bureau has pre-announced a rule change that would carry the same expectations into binding supervision for the parts of the business that decide who gets covered and who gets paid.

What the self-regulation requires

The associations built their standards on the FSC's Guidelines for the Application of AI in the Financial Industry, finalized on June 20, 2024. The standards address four areas: cybersecurity, explainability, data protection, and employee rights. In practice that means an insurer running AI systems is expected to secure them against attack and manipulation, to be able to explain how a model reached a result, to protect the personal data those models ingest, and to account for the rights of employees whose work the systems change.

This is self-regulation, not statute. It binds members of the associations through the industry's own governance rather than through direct penalties. That is the same design Taiwan used across the financial sector, where the FSC published principles-first guidance and then had industry associations write the operating rules that make the principles concrete. The banking association took the same route before insurers, issuing operating rules that require a named senior officer to own AI oversight and that force disclosure when a customer is dealing with a machine. For insurers, the practical effect of the four pillars is that an AI system used in the business is now something the firm has to be able to describe, defend, and supervise, not simply deploy.

The explainability pillar is the one most likely to reshape day-to-day work. An underwriting model that declines an applicant or loads a premium, or a claims model that flags a payout for review, produces a result that someone inside the insurer must be able to account for. Building that capability is not free. It touches how models are documented, how decisions are logged, and how quickly a compliance or actuarial team can reconstruct why a given policyholder received a given outcome.

The pending underwriting and claims duty

The more consequential change is the one still in motion. On December 18, 2025 the FSC Insurance Bureau pre-announced an amendment to Articles 6 and 7 of the Regulations Governing Insurance Solicitation, Underwriting and Claims. As pre-announced, the amendment would apply explainability and fair-treatment duties where AI is used in underwriting and claims. An insurer that uses a model to score an application, set a premium, or evaluate a claim would owe the affected policyholder an explanation and a demonstrably fair outcome.

A pre-announcement (預告) is the notice-and-comment stage in Taiwan's rulemaking, not the final rule. The exact effective date has not been confirmed, so insurers should treat the amendment as pending rather than in force. The direction, however, is clear: the fairness and explainability expectations that currently live in voluntary standards are being drawn into a regulation the FSC enforces.

What it does not do

Neither instrument bans AI in insurance, and neither sets a numeric accuracy or bias threshold that a model must clear. The self-regulatory standards do not carry their own fines, and the Articles 6 to 7 amendment, as pre-announced, is a duty framework rather than a prescriptive technical standard. Insurers are not told which models to use or to stop using AI in underwriting and claims. They are told to be able to explain and to justify the outcomes.

The cross-border read

For a US insurer, this is a familiar pattern arriving in a new market. Colorado's SB 21-169 already requires insurers to test algorithms and predictive models for unfair discrimination and to govern the external data behind them, and the EU AI Act treats AI used in life and health insurance risk assessment and pricing as high-risk, with associated obligations. Taiwan's underwriting-AI fairness duty is the Asia-Pacific parallel to those regimes. A US or global insurer operating in Taiwan would be subject to the local self-regulation now and to the amended Articles 6 and 7 once they take effect, and the convergence across Colorado, the EU, and Taiwan points toward explainable, fairness-tested underwriting AI becoming a de facto baseline rather than a regional quirk.

Browse the full AI Regulation News tracker

Informational analysis for working professionals, not legal advice. Confirm how any rule applies to your situation with qualified counsel.