In July 2024, National Science and Technology Council (NSTC) announced the draft of AI Basic Law (人工智慧基本法), marking a milestone in AI regulation in Taiwan. Heated discussion was sparked off by this draft among civil society organizations and political parties.
The draft of AI Basic Law proposed by NSTC features a regulatory framework that aims to balance AI innovation and ethical considerations. According to the NSTC, the draft law primarily seeks to promote responsible AI development while ensuring transparency, fairness, and accountability in AI-related applications. The key aspects of the draft include:
1. Ethical Guidelines and Governance: The law emphasizes AI ethics, requiring developers and users to adhere to principles of fairness, non-discrimination, and human rights protection.
2. Transparency and Explainability: AI developers must ensure that AI decision-making processes are interpretable, allowing users to understand and challenge automated decisions.
3. Public and Private Sector Responsibilities: The draft outlines different levels of accountability for AI developers, businesses, and government institutions, encouraging self-regulation and compliance with international AI governance standards.
While the NSTC’s draft seems forward-looking and draws on relevant EU and US legislation, it has faced strong criticism from civil society organizations, particularly Taiwan Association for Human Rights (TAHR) and Judicial Reform Foundation (JRF).
According to TAHR’s official statement, the draft lacks concrete governance measures. The draft outlines broad principles but does not provide specific implementation guidelines, which could make enforcement difficult and fail to regulate and mitigate possible risks related to AI.
TAHR also warns potential legal conflicts, particularly in Article 5, which suggests that under the premise of complying with the fundamental principles, the interpretation and application of other laws should not obstruct the development of new technologies. This provision could override regulations such as Personal Data Protection Act, weakening current legal safeguards. Furthermore, the draft emphasizes the government’s role in advancing AI development but lacks clear definitions of government responsibilities and the procedural requirements for agencies involved in AI research and application.
Another key issue is the absence of risk-based approach. Unlike the EU’s AI Act, which adopts different levels of regulation based on various risk categories, NSTC’s draft merely tasks the Ministry of Digital Affairs with establishing a framework without specifying oversight mechanisms.
To address these gaps, TAHR recommends reviewing the legal framework, clarifying the relationship between existing data protection laws and this draft, and revising the draft to include safety and human rights protections. Moreover, it urges government agencies to assess their use of AI to prevent potential rights violations. A more comprehensive and structured approach is needed to ensure AI development aligns with legal and ethical standards.
JRF has also raised several concerns about the draft, arguing it lacks a comprehensive approach to digital governance. A key criticism is that the draft lacks enforceable rights and responsibilities. JRF points out that rather than serving as a mere policy declaration that encourages AI development, the law should explicitly define citizens’ rights and corresponding governmental obligations to protect them. A true “Basic Law” should establish accountability mechanisms to ensure these protections are upheld.
JRF suggests that AI regulation should not exist in isolation but rather be part of a broader legal framework that addresses the entire digital ecosystem. Without such an approach, the rapid advancement of AI may outpace legal safeguards, potentially undermining democratic values and fundamental rights. JRF also states that the draft should explicitly prohibit certain AI applications or scenarios that pose significant risks. Following the EU AI Act’s approach, Taiwan could ban AI uses that threaten fundamental rights and democracy, such as social scoring, biometric categorization with sensitive data, and facial recognition. Implementing such restrictions would help prevent potential abuses and ensure AI development aligns with human rights and ethical standards.
Furthermore, JRF criticizes that the exemption clause in Article 12 of the draft is overly broad, as it excludes AI development and research activities from liability regulations until they reach the application stage. Without clear definitions of “development”, "research," and "application," companies and government agencies could exploit this loophole to evade responsibility by claiming their AI systems are still in testing. JRF urges a proactive approach, such as regulatory sandboxes, to balance innovation with accountability from development to deployment.
Legislators from the Democratic Progressive Party (DPP) and the Kuomintang (KMT) have submitted their respective drafts to the Legislative Yuan. The DPP’s draft, proposed in October 2024, is a slightly revised version of the NSTC's draft, and as of the date of this article, it has been referred to committee for review. The KMT’s draft, proposed in March 2025, emphasizes balancing technological development with human rights protection, and as of the date of this article, it has been placed on the agenda for a plenary session. The Executive Yuan’s draft is still under discussion, with implementation primarily overseen by the Ministry of Digital Affairs, which will also promote subordinate laws, including AI risk classification. Given the widespread use of AI today, there is hope that the review and approval of the AI Basic Law will be completed during this legislative session.
Word count: 845
