Taiwan – Fundamental Act on Artificial Intelligence

On 23 December 2025, the Legislative Yuan (Taiwanese Parliament) passed the Artificial Intelligence Basic Act. The AI Basic Act was promulgated into law on 14 January 2026. The Artificial Intelligence Basic Act creates a legal framework for the development and deployment of AI technologies in Taiwan. In its Article 1, the Act states that its goal is to build a “smart nation” promoting human-centric AI research, and to establish a safe environment for AI applications.

For oversight, the AI Basic Act appoints Taiwan’s National Science and Technology Council (NSTC) as the central authority to oversee its implementation at national level and local authorities to oversee its implementation at local level. 

Definition of “AI” matches the EU AI Act

Regarding its scope, the AI Basic Act defines AI systems as “systems with autonomous operational capabilities that, though 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 closely matches that of the EU’s AI Act and that of the applicable ISO standard, indicating the desire of the Taiwanese government to align with international legal frameworks.   

“Government first” and “principles first” approach

Rather than strictly imposing hard rules on AI system providers and users active on the Taiwanese market, the  AI Basic Act provides for a government first and principles first approach and requires only that certain measures are taken by Taiwanese government agencies to adopt certain policies to preserve seven fundamental principles. These principles are the following:

1. Sustainability and well-being;
2. Human autonomy:
3. Privacy and Data Governance;
4. Cybersecurity and Safety:
5. Transparency and Explainability:
6. Fairness and non-discrimination;
7. Accountability. 

To ensure achievement of these principles, the Act requires government agencies to take the following measures: 

• Put in place safety measures for AI systems posing certain risks: the AI Basic Act requires the Taiwanese government to work out a governance framework for the prevention of certain risks. These risks encompass “life, bodily integrity, liberty, or property, disrupt social order, national security, or the ecosystem, or cause bias, discrimination, false advertising, misinformation or fraud”. Thus, the AI Basic Act is more specific than the EU AI Act in which human rights it protects , specifically naming “life, bodily integrity, liberty, or property, disrupt social order, national security, or the ecosystem, or cause bias, discrimination, false advertising, misinformation, or fraud”, but also encompasses more collective risks than “only” “health, safety and human rights” as mentioned in recital 1 AI Act;
• Establish a National AI Strategy Special Committee, that coordinates, promotes and oversees AI affairs at national level;
• Promote AI education: the government shall promote AI and AI ethics education across schools, industries, organizations, society and government agencies to strengthen digital literacy;
• Allocate budget for AI R&D: the Taiwanese government shall, within its fiscal capabilities, allocate sufficient budget to support AI policy development;
• Precedence of innovation over other laws: the Taiwanese government commits to provide reasonable use, support and subsidies and to improve any relevant regulations when evaluating new technologies. As a rule of interpretation, the Taiwanese government puts forward an important principle: when the interpretation of this Act conflicts with other laws, the interpretation that promotes new technologies and services takes precedence. This rule indicates that risk governance must serve innovation first and foremost. The Taiwanese government may also establish regulatory sandbox environments for new AI products or services;
• Promote international cooperation: the Taiwanese government pledges to promote international cooperation related to AI;
• Establish a framework for the sharing of AI-useable data: the Taiwanese government will set up mechanisms for data openness, sharing and re-use to improve data availability. This comes with a commitment to improve the quality and quantity of AI training data;
• Avoid unnecessary collection of personal data: the AI Basic Act requires authorities to ensure that personal data protection is by design and by default by avoiding any unnecessary processing of personal data in AI R&D and applications;
• Labor rights: unlike the EU’s AI Act, the AI Basic Act requires the government  to use AI to protect labor rights and to reduce skills gaps caused by AI development, safeguard economic security and ensure dignified labor. This includes employment counselling given to those who lose their jobs to AI;
• Develop international standards: the Taiwanese government commits to reference international standards, such as ISO 42001, to develop a risk-classification framework in order to align with global governance systems. Any AI systems that pose risks may be prohibited by law;
• AI liability: the Taiwanese government will clearly define liability and conditions for accountability for high-risk AI applications and establish mechanisms for redress. The Act also provides that any AI R&D is exempt from liability until the system enters real-world testing or is used to provide products and services, thus providing a simplified research exemption, compared to the one in the EU AI Act;
• Risk assessments before using AI: when using AI to perform tasks or public services, the Taiwanese government will first conduct risk assessments and mitigation measures. 

Timeline and phased application

The AI Basic Act has entered into force on the date of presidential promulgation, i.e. 14 January 2026. After the AI Basic Act comes into force, the following measures will follow: 

• The Ministry of Digital Affairs, NSTC, Ministry of Education and Ministry of Health and Welfare will publish a Minors/Human Rights/Gender Impact Assessment;
• The government shall review and complete its risk assessment for existing government AI rules within six months;
• The government shall establish rules for government use of AI within twelve months;
• The government shall review, establish, or ament any relevant laws to comply with the Act’s requirements within twenty-four months after entry into force.