Taiwan AI Basic Act

Overview

The Taiwan Artificial Intelligence Basic Act (人工智慧基本法), promulgated and in force since January 14, 2026, marks a significant legislative milestone for Taiwan, positioning the nation to embrace the opportunities and address the challenges presented by artificial intelligence. This foundational legislation is designed to facilitate the construction of a 'smart nation' by actively promoting the research, development, and application of human-centric AI technologies. It aims to strike a delicate balance between fostering technological innovation and safeguarding the fundamental rights and dignity of its citizens. The Act establishes a comprehensive legal framework that guides AI development, ensuring that progress is made responsibly and ethically, while also enhancing Taiwan's international competitiveness in the global AI landscape.

The core philosophy of the Artificial Intelligence Basic Act is rooted in a 'principles-first' and 'government-first' approach, emphasizing a proactive yet cautious stance towards AI governance. It outlines seven fundamental principles that underpin all AI research and application, including sustainability, human autonomy, privacy protection, cybersecurity, transparency, fairness, and accountability. These principles serve as guiding lights for both public and private sectors, ensuring that AI development aligns with societal values and ethical considerations. Furthermore, the Act delineates the roles of key government bodies, such as the National Science and Technology Council (NSTC) as the central competent authority and the Ministry of Digital Affairs (MODA) in developing risk classification frameworks, thereby establishing a robust institutional structure for AI governance.

Definitions

The Taiwan Artificial Intelligence Basic Act provides a clear and concise definition of "Artificial Intelligence" (人工智慧) to ensure a common understanding across all stakeholders. According to Article 3 of the Act, artificial intelligence refers to "systems with autonomous operational capabilities that, through input or sensing and using machine learning and algorithms, can produce predictions, content, recommendations, or decisions—whether explicit or implicit—that affect physical or virtual environments." This definition is crucial for distinguishing AI systems from traditional software and for delineating the scope of the Act's application. It emphasizes the system's ability to operate autonomously and generate outputs that influence real or simulated environments, encompassing a wide array of AI technologies from predictive analytics to generative AI.

Beyond the general definition of AI, the Act also implicitly or explicitly addresses other key terms relevant to its implementation. For instance, while not providing a single, exhaustive definition, it refers to "high-risk artificial intelligence" applications, which are those identified by relevant central competent authorities, in consultation with the Ministry of Digital Affairs, as potentially causing significant harm to people's lives, bodies, freedom, or property, or disrupting social order, national security, or the ecological environment. The concept of "data governance" is also central, emphasizing the proper collection, processing, and utilization of data, including adherence to the principle of data minimization and the protection of personal privacy. These definitions and conceptual frameworks are essential for the consistent application of the Act's principles and for the development of subsequent, more detailed regulations and guidelines.

Governance and Institutional Framework

The governance structure for artificial intelligence in Taiwan, as established by the Artificial Intelligence Basic Act, is designed to be multi-layered and collaborative, ensuring both centralized strategic direction and decentralized, sector-specific oversight. At the apex of this framework is the National Science and Technology Council (NSTC), designated as the central competent authority responsible for the overall planning and coordination of national AI development. This strategic placement underscores Taiwan's commitment to prioritizing AI research and innovation, leveraging the NSTC's expertise in scientific and technological advancement. The NSTC is tasked with formulating the national AI development blueprint, guiding resource allocation, and fostering an environment conducive to cutting-edge AI research.

Complementing the NSTC's role, the Executive Yuan is mandated to establish a "National AI Strategy Special Committee." Chaired by the Premier, this committee comprises experts, representatives from AI-related civil society groups and industries, government officials, and local government leaders. Its primary function is to coordinate, promote, and supervise national AI affairs, including the deliberation and approval of the national AI development纲领 (guideline/program). This committee ensures a whole-of-government approach to AI, facilitating cross-ministerial collaboration and integrating diverse perspectives into policy-making. Furthermore, the Ministry of Digital Affairs (MODA) plays a crucial role in developing an internationally aligned AI risk classification framework, which will serve as a foundational tool for various sectoral competent authorities to formulate their specific management regulations. This distributed responsibility aims to balance overarching national strategy with tailored, practical implementation across different industries and applications.

Key Focus Areas

The Taiwan Artificial Intelligence Basic Act is built upon seven core principles that guide the responsible development and application of AI across all sectors. These principles are: (1) Sustainable Development and Well-being, emphasizing the need for AI to contribute to social equity and environmental sustainability while mitigating digital divides; (2) Human Autonomy, ensuring that AI supports human dignity and fundamental rights, with ultimate decision-making power remaining with humans; (3) Privacy Protection and Data Governance, mandating the careful safeguarding of personal data, adherence to data minimization principles, and the promotion of non-sensitive data sharing; (4) Cybersecurity and Safety, requiring robust security measures to prevent threats and ensure the stability of AI systems; (5) Transparency and Explainability, advocating for appropriate disclosure and labeling of AI outputs to enhance trustworthiness and understanding; (6) Fairness and Non-discrimination, aiming to prevent algorithmic bias and discriminatory outcomes; and (7) Accountability, ensuring that relevant parties bear responsibility for internal governance and external societal impacts. These principles collectively form the ethical bedrock for Taiwan's AI ecosystem.

In addition to these overarching principles, the Act identifies several key areas for focused action. It mandates the government to proactively avoid AI applications that infringe upon human rights, disrupt social order, or pose risks to national security or the environment. Special consideration is given to the best interests of children and adolescents, with requirements for clear labeling of high-risk AI products or systems. The Act also places significant emphasis on data governance, advocating for mechanisms that enhance data openness, sharing, and reuse while strictly protecting personal privacy and intellectual property rights. Furthermore, it addresses the societal impact of AI on the workforce, obligating the government to ensure labor rights, mitigate skill gaps, and provide employment counseling for those affected by AI-driven job displacement. These focus areas underscore a holistic approach to AI governance, aiming to maximize benefits while proactively addressing potential harms.

Implementation Framework

The implementation framework of the Taiwan Artificial Intelligence Basic Act is characterized by a dual-track approach, combining overarching principles with sector-specific regulations. A central component is the development of an AI risk classification framework by the Ministry of Digital Affairs (MODA), which is tasked with referencing international standards and norms to create a system that can be interfaced globally. This framework will enable various central competent authorities to identify, assess, and manage AI risks within their specific domains. Based on this national framework, individual ministries and agencies responsible for particular industries or sectors are mandated to establish their own risk-based management regulations and industry guidelines. This ensures that while a consistent national approach to AI risk is maintained, the unique characteristics and challenges of different sectors, such as healthcare, finance, or transportation, can be adequately addressed through tailored rules.

Beyond risk management, the Act also promotes innovation through various supportive measures. It encourages the establishment or enhancement of "innovation experimental environments," often referred to as regulatory sandboxes, for AI research and application services. These sandboxes provide a controlled setting where new AI products and services can be tested and developed with regulatory flexibility, fostering technological breakthroughs. The government is also committed to providing reasonable use, support, and subsidy measures for AI development, training, testing, and verification of emerging technologies. Furthermore, the Act emphasizes the importance of data governance by requiring the establishment of mechanisms for data openness, sharing, and reuse to improve the availability, quality, and quantity of AI training data, all while safeguarding personal data privacy and intellectual property rights. This comprehensive framework aims to create an agile and responsive regulatory environment that encourages innovation while ensuring public trust and safety.

Monitoring and Evaluation

The Taiwan Artificial Intelligence Basic Act incorporates mechanisms for ongoing monitoring and evaluation to ensure the effective implementation of its principles and policies. A key element of this oversight is the mandate for the Executive Yuan's National AI Strategy Special Committee to convene at least once a year. During these meetings, the committee is responsible for reviewing the "National Artificial Intelligence Development綱領" (guideline/program) and assessing the annual performance of national AI initiatives. This regular review process allows for strategic adjustments and ensures that AI development remains aligned with the Act's objectives, including balancing innovation, social welfare, and human rights protection.

Furthermore, the Act places specific obligations on government agencies regarding their own use of AI. It requires all levels of government agencies to conduct risk assessments for their public-facing AI applications and to establish internal control management norms within a specified timeframe. This self-assessment and internal regulation mechanism is crucial for ensuring that government deployment of AI adheres to the Act's principles, particularly concerning transparency, fairness, and accountability. The Ministry of Digital Affairs (MODA) is also tasked with promoting AI risk classification frameworks and providing evaluation tools, which will further aid in the consistent monitoring and assessment of AI systems across different sectors. This multi-faceted approach to monitoring and evaluation aims to build public trust and ensure responsible AI governance.

Penalties, Liability, and Appeals

A distinctive feature of the Taiwan Artificial Intelligence Basic Act, particularly when compared to some international counterparts, is its lack of direct punitive clauses or specific penalties within the Act itself. This approach reflects its nature as a "basic law," serving as a guiding framework or a strategic declaration rather than a prescriptive regulatory instrument with immediate fines or sanctions. The Act's primary purpose is to establish fundamental principles, define governance structures, and set the direction for future AI-related legislation and policy. Therefore, while the Basic Act does not impose direct penalties, it explicitly states that matters not covered by this Act shall be governed by other applicable laws.

However, the absence of direct penalties in the Basic Act does not absolve AI developers or users of legal responsibility. If an AI system causes harm or violates existing legal provisions, such as those related to civil liability, criminal law, or personal data protection, the relevant provisions of those existing laws would still apply. Crucially, the Act mandates the government to clarify responsibility attribution and liability conditions for high-risk AI applications. It also requires the establishment of corresponding redress, compensation, or insurance mechanisms for such applications. This forward-looking provision ensures that as specific high-risk AI applications emerge, a clear framework for accountability and victim recourse will be developed through subsequent, more detailed regulations issued by sectoral competent authorities.

Relationship to Other Instruments

The Taiwan Artificial Intelligence Basic Act operates as a foundational and guiding piece of legislation within Taiwan's legal hierarchy. It is designed to be a "basic law," providing overarching principles and a strategic framework rather than exhaustive, detailed regulations. This means that while the Act establishes the fundamental direction and ethical considerations for AI development and application, it explicitly states that where the Act does not provide specific provisions, other existing laws shall apply. This ensures legal coherence and avoids creating a regulatory vacuum or conflicts with established legal frameworks.

The Act's "basic law" nature necessitates the development of numerous subsidiary instruments and regulations by various government bodies. The Ministry of Digital Affairs (MODA), for instance, is tasked with developing a national AI risk classification framework, which will then guide sectoral competent authorities in formulating their own risk-based management regulations and industry-specific guidelines. These forthcoming regulations, directives, and administrative measures will provide the granular details for compliance within specific industries, such as healthcare, finance, or transportation. The Act also emphasizes the need for existing laws and regulations to be reviewed, amended, or repealed as necessary to align with the principles and objectives of the Basic Act, demonstrating a commitment to an adaptive and evolving legal landscape for AI. This layered approach allows for flexibility in responding to rapid technological advancements while maintaining a consistent ethical and strategic foundation.

International Alignment

Taiwan's Artificial Intelligence Basic Act demonstrates a strong commitment to international alignment, drawing inspiration and referencing global standards and governance frameworks in its formulation. The Act's seven core principles, for example, reflect widely recognized ethical guidelines from international bodies such as the OECD's Recommendation on Artificial Intelligence, the G7 Hiroshima Process Code of Conduct for Organizations Developing Advanced AI Systems, the US Blueprint for an AI Bill of Rights, and the EU's Ethics Guidelines for Trustworthy AI. This deliberate incorporation of international best practices aims to ensure that Taiwan's AI governance framework is interoperable and credible on the global stage.

Beyond adopting international principles, the Act explicitly mandates the government to actively promote international cooperation in AI-related matters. This includes fostering exchanges of talent, technology, and infrastructure with other countries, as well as engaging in public-private partnerships to drive AI innovation globally. The Ministry of Digital Affairs (MODA) is specifically tasked with developing an AI risk classification framework that references international standards or norms, such as ISO 42001, to ensure its compatibility with global governance systems. This focus on international collaboration and standard harmonization is crucial for Taiwan, a major player in the global technology supply chain, to maintain its competitive edge, facilitate cross-border data flows, and participate effectively in the evolving global discourse on AI governance.

Implementation Timeline

Sources and References