Just before the end of 2025, the Legislative Yuan finally passed the third reading of the "Artificial Intelligence Basic Law" (hereinafter referred to as the "AI Basic Law"), designating the National Science and Technology Council (NSTC) as the competent authority. As Taiwan's first regulation specifically targeting AI, its impact on industry and society is being closely watched by government, industry, and academia.

Chang Li-ching, Chair Professor at Shih Chien University’s College of Law and CEO of the International Artificial Intelligence and Law Research Foundation, analyzed that the "AI Basic Law" resembles a "framework law" rather than a "regulatory law" centered on mandatory constraints.

Unlike the strict controls on artificial intelligence usage in the European Union and South Korea, the core value of Taiwan’s "AI Basic Law" aligns more closely with Japan’s approach: balancing promotion with governance principles and preserving space for innovation. "The completion of the Basic Law is merely a stage mission; the real challenge lies in how to concretely implement the legislative spirit and policy guidelines into individual laws," Chang said.

Since the Basic Law does not establish penalties or prescribe specific obligations, more supporting measures are needed for implementation. Taking supervision and data governance as an example, Wen Shao-qun, Senior Executive Vice President of Technology and Transformation at Deloitte Taiwan, pointed out that the core difficulty of AI governance is how to draw the "red line" when abstract legal principles are applied to concrete commercial scenarios.

Wen cited the example of banks using AI to categorize and tag customers for marketing purposes—identifying wealthy customers with good credit to offer better terms. Intrinsically, this contradicts the principle of "fairness and non-discrimination" found in the Basic Law.

Where exactly should the regulatory scale be set? This requires comprehensive thinking by competent authorities during implementation. "Cross-ministry and cross-domain dialogue mechanisms are crucial," observed Wen Yi-ling, CEO of the AI Foundation, noting that AI development involves complex stakeholders.

Despite the third reading of the "AI Basic Law" and its imminent implementation, outsiders perceive significant challenges. Addressing these doubts, NSTC Minister Wu Cheng-wen accepted an exclusive interview with *CommonWealth Magazine*, emphasizing that the "AI Basic Law" is a major step accelerating Taiwan's AI development and will make 2026 the "inaugural year of AI application" in Taiwan. The following is a summary of the interview:

Distinct Missions for the NSTC and MODA

The "AI Basic Law" is a "superior law." From a governance perspective, the NSTC's mission is technology R&D, budgeting for technology, and policy formulation; it does not directly interface with the public.

However, the Ministry of Digital Affairs (MODA) interfaces directly with the public. Therefore, MODA will formulate relevant "action laws." For example, the "Taiwan Sovereign AI Training Corpus License Agreement – Version 1," developed in cooperation with the Ministry of Economic Affairs (MOEA), and measures for preventing AI fraud are under MODA's jurisdiction.

In the future, cross-ministry collaboration will increase. MODA supports the AI software industry, while hardware falls under the MOEA; cooperation between these two departments will become more frequent. Meanwhile, the formulation and development of national policies, such as the "New Ten Major Infrastructure Projects," are joint efforts by the National Development Council (NDC) and the NSTC.

"Since future AI applications will span all ministries, we cannot rely solely on the NSTC. Furthermore, I do not have a legal background. Therefore, we have a platform composed of three Ministers without Portfolio—myself, Minister without Portfolio Lin Ming-hsin, and NDC Minister Yeh Chun-hsien—to assist in communicating digital governance and reviewing bills," said Wu.

Passage of Basic Law to Accelerate AI Development

"After the Basic Law passes, the entire development of AI will accelerate," Wu stated. "For instance, the Ministry of Health and Welfare has already passed several bills, and MODA recently proposed the draft 'Act for Promoting Data Innovation and Utilization.' The goal is to encourage innovation as much as possible while protecting basic human rights. Of course, there are red lines, so each ministry must decide where its red line lies. You cannot draw a single line that bans everything, as that would hinder innovative development."

Wu emphasized the urgency of passing the Basic Law so that manufacturers know what is permissible. "Under existing laws, if interpreted most strictly, manufacturers would not dare to invest. Even if foreign companies come in, they would only dare to do hardware in Taiwan."

Therefore, the goal following the passage of the Basic Law is to encourage manufacturers to develop application software, which is the core focus of AI. "That is why 2026 will be the inaugural year of Taiwan's sovereign AI application," Wu declared.

The progress of the "New Ten Major Infrastructure Projects" will also accelerate. For example, high-speed quantum computing technology can aid development in the biomedical field, while silicon photonics is a key technology for future optical communications, crucial for the interconnection of low-earth orbit satellites and AI data centers.

Talent must also keep up. Regarding the robotics industry—also part of the "New Ten Major Infrastructure Projects"—Taiwan will not only establish R&D teams, but the National Development Fund has also allocated NT$10 billion as an investment and talent cultivation fund. The hope is that once Taiwan's laws are comprehensive, it will attract startups and foreign talent. "In 2026, we hope to push the development of the AI industry step by step through the coordination of laws, research funding, and increasing manufacturers' willingness to invest."

MODA to Provide Risk Identification Framework and Test Tools

Hou Yi-hsiu, Deputy Minister of Digital Affairs, pointed out that Taiwan’s "AI Basic Law" categorizes risks into three types: risks inherent to the system and service, human-machine interaction risks, and social impact risks.

After completing the risk classification framework, MODA will assist various competent authorities in defining risks for specific technologies within specific fields and application scenarios. For example, facial recognition for door access is common in Taiwan, but the EU prohibits its real-time, continuous, large-scale use in public places.

By using the risk classification framework to categorize high-risk items, appropriate policy tools can be adopted, such as requiring transparency, mandating testing before market release, or prohibiting use in specific scenarios.

Under the principles of AI development and the Basic Law, MODA will formulate an actionable risk classification framework template, allowing competent authorities to follow suit and establish risk response measures for their respective units.