Taiwan passes the Basic AI Law! How to regulate AI scams? Seven principles for unemployment benefits at a glance

Taiwan’s Legislative Yuan passed the “Basic Law on Artificial Intelligence” in a third reading on December 23, establishing seven core governance principles, with the National Science and Technology Council serving as the central competent authority. Regarding AI scams and deepfake images, high-risk applications must be marked with warning labels and a relief mechanism established; for AI-induced unemployment, the government is responsible for employment counseling. However, civil rights groups criticize this law as similar to a white paper, lacking provisions on citizens’ rights and a list of prohibited AI abuses.

Taiwan’s AI Basic Law Defines Seven Development Principles

To ensure that technological development stays on the right track, Taiwan’s AI Basic Law clearly outlines the seven core principles that the government must follow when promoting AI research and application. These principles cover various aspects from technical safety to social fairness, forming a relatively comprehensive governance framework.

The sustainability principle requires balancing social fairness and environmental sustainability, reducing the digital divide. This means AI development should not only serve tech elites or urban areas; rural and vulnerable groups should also benefit from AI. The human autonomy principle emphasizes respecting personal rights and cultural values, allowing human oversight, and implementing a people-centered approach. This acts as a constraint on fully autonomous AI decision-making, ensuring humans retain ultimate control.

Privacy protection and data governance principles demand proper safeguarding of personal data and the implementation of data minimization. This directly addresses public concerns about excessive personal information collection by AI. Security principles require establishing cybersecurity measures to prevent attacks and threats. As AI systems become more complex, security vulnerabilities could lead to catastrophic consequences.

Seven Governance Principles of Taiwan’s AI Basic Law

Sustainability: Balancing social fairness and environmental sustainability, reducing the digital divide

Human Autonomy: Respecting personal rights and cultural values, allowing human oversight, and implementing a people-centered approach

Privacy Protection and Data Governance: Properly safeguarding personal data and implementing data minimization

Security: Establishing cybersecurity measures to prevent attacks and threats

Transparency and Explainability: Appropriately marking or disclosing AI-generated information to enhance trustworthiness

Fairness: Avoiding algorithmic bias or discrimination against specific groups

Accountability: Developers and users should bear corresponding governance and social responsibilities

The transparency and explainability principle requires appropriate marking or disclosure of AI-generated outputs to increase trust. This means AI-produced content must be clearly labeled to prevent misleading the public. The fairness principle aims to prevent bias or discrimination in algorithms. The accountability principle emphasizes that developers and users should assume governance and social responsibilities, providing a legal basis for potential liability in AI accidents.

Regulatory Measures for AI Scams and Deepfake Images

Regarding the public’s primary concern about AI scams, Taiwan’s AI Basic Law provides an initial management framework. The third reading stipulates that the government should prevent AI applications from infringing on people’s lives, property, or safety, and guard against false advertising, misinformation, or forgery. While this principle is directionally correct, specific implementation details still require supplementary laws from various agencies.

For AI products or systems identified as “high-risk applications,” the government requires clear warning labels or caution notices, and the establishment of relief, compensation, or insurance mechanisms to address potential damages. This tiered management approach is similar to the EU AI Act, categorizing AI applications into different risk levels and imposing stricter regulation on high-risk uses.

However, what constitutes a “high-risk application”? How to define the boundary between AI scams and legitimate uses? These key questions are not yet clarified in Taiwan’s AI Basic Law. For example, does using AI to generate celebrity images for product promotion constitute a scam? How to determine liability for AI customer service providing misleading investment advice? These specific scenarios require subsequent legislation or judicial interpretation for clarification.

Civil Rights Defenders Foundation has issued strong criticism, stating that Taiwan’s AI Basic Law lacks a clear list of prohibitions. They call on the government to refer to international standards and explicitly prohibit the use of AI to establish social scoring systems, predict individual criminal risks, or conduct indiscriminate emotional analysis and facial recognition in education and law enforcement. Only with these specific prohibitions can AI misuse be effectively prevented.

What to Do About AI-Induced Unemployment? Employment Counseling Mechanisms Enshrined in Law

In response to concerns about AI causing unemployment, Taiwan’s AI Basic Law states that the government should actively utilize AI to ensure labor rights, reduce skill gaps, and protect dignity in work. For those unemployed due to AI utilization, the government has a responsibility to provide “employment counseling” based on their work abilities. This marks the first time in Taiwan that the impact of AI on the labor market has been explicitly regulated at the legal level.

Employment counseling may include vocational training, AI skills retraining, employment matching services, and transitional living subsidies. However, Taiwan’s AI Basic Law does not specify subsidy amounts, training durations, or eligibility criteria; these details need to be supplemented by the Ministry of Labor’s implementing laws.

蔡明順, President of Taiwan Artificial Intelligence School, believes that the passage of the Basic Law signifies that AI governance in Taiwan has moved from policy declaration to legal implementation. He emphasizes that the next challenge lies in the formulation of relevant laws by various agencies, such as the Financial Supervisory Commission for financial applications, the Ministry of Health and Welfare for medical certification, and the Ministry of Economic Affairs for manufacturing standards.

蔡明順 recommends that industries should not wait for agencies to draft relevant laws but should start internally assessing AI applications and conducting risk self-evaluations. In the future, every company’s board of directors must have someone knowledgeable in both technology and AI regulations, and establish cross-company data alliances to collaboratively experiment with technology. He particularly points out that Article 10 of Taiwan’s AI Basic Law provides real resources for small and medium-sized enterprises, including subsidies, tax incentives, and investment rewards with legal backing.