Taiwan has long been recognized as a pivotal player in the global technology landscape, particularly in semiconductor manufacturing and chip production. As artificial intelligence continues to reshape industries worldwide, Taiwan is positioning itself to become a leader in AI development, especially by creating its own Traditional Chinese large language model.
To this end, after nearly two years of extensive discussion, consultation and debate, Taiwan enacted the AI Basic Act, which went into effect 14 Jan. This landmark legislation establishes the government's foundational stance on AI and serves as a guiding framework for future policies, regulations and legal interpretations.
Defining AI and the institutional framework
The act adopts a definition of AI that closely aligns with the EU AI Act, emphasizing AI systems as technologies that provide humans with predictions, content, recommendations or decisions. This alignment signals Taiwan's intent to harmonize its regulatory approach with international standards, facilitating cross-border cooperation and market integration.
The law designates the National Science and Technology Council as the primary regulator overseeing its implementation. However, many critical operational responsibilities are delegated to the Ministry of Digital Affairs, which will play a central role in managing AI governance, risk assessment and policy promotion.
Seven fundamental principles guiding AI governance
The AI Basic Act articulates seven core principles that will underpin Taiwan's AI ecosystem and guide the formation of future policies and regulations: sustainable development and well-being; human autonomy; privacy protection and data governance; cybersecurity and safety; transparency and explainability; fairness and nondiscrimination; and accountability.
The explicit inclusion of accountability marks a significant milestone in Taiwan's legal landscape, emphasizing the importance of a governance framework for AI deployment.
Responsibilities and obligations under the AI Basic Act
While the AI Basic Act primarily outlines the government's responsibilities, it also imposes important obligations on private enterprises, particularly regarding the classification and management of high-risk AI systems. When an AI product or service is deemed "high-risk" by the government, it must carry appropriate warnings or alerts to inform users of potential risks.
The classification of high-risk AI will be determined jointly by the MODA and other relevant government agencies. To ensure a comprehensive and balanced approach, the development of risk assessment tools and methodologies will involve multistakeholder consultations. These consultations will incorporate feedback from civil society, industry experts, academia and legal professionals, ensuring that diverse perspectives shape the regulatory framework.
Promoting an internationally operable AI risk framework
The MODA is tasked with promoting an internationally interoperable AI risk management framework by referencing global standards and best practices. This initiative aims to facilitate Taiwan's integration into the global AI ecosystem and ensure that its regulatory approach remains compatible with international norms.
Furthermore, the ministry will assist other government agencies in developing sector-specific risk-based AI management frameworks. These agencies, in turn, will guide the industries under their purview to establish tailored guidelines and codes of conduct for AI risk management. This layered approach ensures AI governance is both comprehensive and adaptable to different sectors' unique challenges.
Restrictions and prohibitions on AI applications
The AI Basic Act also delineates circumstances under which AI applications may be prohibited or classified as high-risk with stringent restrictions. These include scenarios where AI: infringes on individuals' life, bodily integrity, freedom or property rights; disrupts social order, national security or environmental sustainability; or engages in bias, discrimination, false advertising, misinformation or fraud, violating applicable laws.
To address liability issues arising from high-risk AI applications, the government will establish clear standards and conditions for attributing responsibility. Mechanisms for remedy, compensation, and insurance will also be developed to protect affected parties and ensure accountability.
Data governance and AI training data
Recognizing high-quality data is the lifeblood of AI, the act mandates Taiwan's government to establish an open data framework and mechanisms to facilitate data sharing and secondary use. This initiative aims to enhance the availability and quality of training data, which is critical for developing robust AI models.
In the meantime, the AI Basic Act emphasizes the importance of protection of personal data and requires the implementation of data minimization in AI development.
The government is actively considering new legislation focused on data governance and open data promotion. This prospective law will initially prioritize opening government-held data and gradually encourage private sector participation in data sharing, fostering a more collaborative data ecosystem.
Regarding copyright concerns, the AI Basic Act does not explicitly relax copyright protections but emphasizes the government's role in increasing the quantity and quality of AI training data. This balanced approach seeks to respect intellectual property rights while promoting AI innovation.
Funding, research and talent development
The AI Basic Act underscores the necessity of sustained investment in AI research and development. The government is committed to allocating increased funding, promoting AI infrastructure and providing financial incentives such as tax deductions to stimulate innovation. The AI Basic Act also mentions the possibility of establishing a regulatory sandbox for AI development.
Moreover, the law highlights the importance of AI education and talent cultivation. It encourages collaboration between public institutions and private enterprises to nurture a skilled workforce capable of driving Taiwan's AI ambitions forward. Protecting labor rights amid AI-driven transformations is also a key consideration, ensuring technological progress does not come at the expense of workers' welfare.
Legislative roadmap and industry implications
One of the most consequential mandates of the AI Basic Act is the requirement for Taiwan's government to review existing laws and enact new legislation within two years to align with the principles and provisions set forth in the act. This legislative overhaul will provide a comprehensive legal framework to regulate AI technologies effectively.
In parallel, the government has already begun establishing the AI risk classification framework and related mechanisms. For private companies, this signals an urgent need to audit their AI usage, develop internal governance frameworks, and stay abreast of regulatory developments to ensure compliance.
Conclusion
Taiwan's enactment of the AI Basic Act represents a bold and forward-looking step in its aim to become a global AI leader. By embedding fundamental principles such as accountability, transparency and fairness into its AI governance framework, Taiwan is setting a high standard for responsible AI development.
The law's risk-based approach will be further implemented and developed within two years from now. For the private sector, the AI Basic Act is both a call to action and an opportunity to innovate responsibly within a clear regulatory environment. Rapid action to establish an internal AI governance framework, along with continuous monitoring of AI regulations, is recommended for the private sector.
Ken-Ying Tseng is a partner at Lee and Li, Attorneys at Law.
