DSET Commentary in NBR Analyzes Three Core Challenges for Taiwan’s AI Governance

Taiwan completed the legislation of its Artificial Intelligence Basic Act in December 2025, marking a major milestone in the country’s AI governance. DSET Democratic Governance Program Director Kai-Shen Huang recently published a commentary in the National Bureau of Asian Research (NBR) analyzing the three core challenges Taiwan’s AI governance will face in the wake of the Basic Act’s enactment.

The Basic Act: Governance Principles, Institutional Architecture Still Under Construction

The Basic Act sets out seven core principles and a clear policy direction, but leaves much of the implementation — high-risk AI designation, cross-ministerial responsibility, and escalation mechanisms — to subsequent regulations and coordination. Huang notes that an initial legislative framework for dividing labor among ministries and sectoral regulators exists, but its effectiveness will only be tested in real-world disputes.

PDPA Amendments and Data Protection Commission: Taking on the Enforcement Burden of AI Governance

The Personal Data Protection Act (PDPA) amendments have passed their third reading, and the Personal Data Protection Commission (PDPC) is being established. Since AI risks are largely manifested through data practices, PDPA enforcement is effectively becoming a central pillar of AI governance. Huang notes that the key challenge for the PDPC is building AI-aware supervisory capabilities from day one, rather than defaulting to traditional data protection frameworks.

The Copyright Gap: Moving from Interpretive Guidance to Scalable Licensing Mechanisms

The copyright treatment of AI training data remains contested globally. Taiwan currently lacks a dedicated text and data mining (TDM) exception, and although the Intellectual Property Office (TIPO) has issued interpretive guidance, it cannot substitute for a clear legal framework. Huang notes that as major AI models mature and copyright disputes move into litigation and legislation elsewhere, Taiwan’s response faces growing time pressure. For the industry, compliance uncertainty will persist until licensing mechanisms are in place; for rights holders, the legal gray zone offers little protection. Huang emphasizes that moving from case-by-case interpretation to a predictable, operable licensing regime is the most urgent next step. DSET’s Democratic Governance Program will continue to monitor AI governance developments in Taiwan and globally.