By Julia Bergström and Alex Myslinski
Taiwan’s draft AI Basic Act aims to position the country as a global leader in ethical AI governance, but questions about whether it can effectively balance innovation and regulation remain.
The global race to regulate artificial intelligence (AI) is intensifying as governments work to address the ethical, legal, and societal challenges associated with these rapidly advancing technologies.
In May 2024, the EU led the charge by introducing its AI Act – a groundbreaking legislative framework that classifies AI systems based on their risk levels, from unacceptable to minimal. The legislation also imposes corresponding requirements, such as transparency and human oversight for high-risk AI systems, while prohibiting practices like government social scoring. The act came into force on August 1, with provisions that will come into operation gradually over the following 6 to 36 months.
The United States is similarly advancing toward comprehensive federal AI regulation. In October 2023, President Joe Biden issued an executive order for the safe, secure, and responsible development and use of AI, promoting competition, civil rights, and ethical technology use across federal agencies.
The UK, by contrast, has chosen a sector-specific approach, relying on existing laws rather than implementing a singular regulatory framework. Meanwhile, countries like Australia have been debating whether existing legal frameworks are sufficient to address the complexities of AI governance.
Taiwan, the world’s largest producer of chips and a prominent player in AI and the Internet of Things (IoT), is also making strides. In July 2024, Taiwan’s National Science and Technology Council (NSTC) unveiled a draft of the AI Basic Act, marking an important milestone in the country’s efforts to regulate AI.
Since its release, however, the draft has sparked debate regarding its adequacy, oversight, and compatibility with existing regulations – issues that will ultimately determine whether Taiwan’s AI governance will foster or stifle innovation.
A balancing act
An alignment with global standards is crucial, given Taiwan’s export-driven economy. Following widely recognized frameworks not only facilitates smoother integration into international markets but also reassures partners and investors of Taiwan’s commitment to regulatory transparency and consistency.
However, the path to international alignment is obstructed by a slow and fragmented flow of implementation. The act is meant to take precedence as the national standard, and provisions of other Taiwanese regulations are to fill in the gaps when relevant and necessary. If no such national standard or relevant regulation is in place, the standards of other progressive countries may then be considered and submitted for approval by the “competent authorities,” which are loosely defined by the draft.
Adopting multiple international standards into the hierarchy of Taiwan’s regulations could potentially lead to an overly complex framework. Instead, a unified approach – such as the OECD’s definition of “AI systems,” which has already been adopted by the EU – could provide the necessary clarity without excessive complexity.
One significant concern regarding this area of the draft AI Basic Act involves its alignment with human rights protections, especially those related to data privacy. The Taiwan Association for Human Rights (TAHR), the oldest independent human rights organization in Taiwan, in September voiced reservations about the act’s scope, suggesting that it may conflict with Taiwan’s Personal Data Protection Act (PDPA).
The AI Basic Act emphasizes the government’s role in balancing individual rights to privacy and information autonomy with the growth of the AI industry, suggesting that the government may need to allow flexibility in how data is used to enable AI advancements. Conversely, the PDPA strictly requires organizations to protect personal data and limit its collection and use without consent. Specific criticisms have centered on the government’s use of AI, such as employing algorithms to draft criminal verdicts and predict reoffending risk, without fully disclosing how such decisions are made.
Transparency issues have also arisen with AI projects in collaboration with international technology companies. One example is a project between the National Health Insurance Administration (NHIA) and industry to develop an AI system to assess diabetes risk. According to TAHR, a lack of transparency around data sources and processing has raised significant concerns regarding data privacy, consent, and the potential for bias.
AI bias arises when algorithms reflect or amplify prejudices present in the data used for their training. Prejudices can occur due to imbalanced datasets, poorly designed algorithms, or systemic societal inequities that influence the data. For example, an AI system trained on recruitment data that favors male candidates might perpetuate gender discrimination in hiring processes, undermining diversity and equity in workplaces.
Instances of AI bias have been widely documented across industries and can have profound implications for labor and human rights. Automated hiring systems biased against certain demographics can limit access to economic opportunities. Furthermore, the use of biased AI in surveillance or law enforcement may exacerbate discrimination and erode civil liberties.
Risk assessments and supervision mechanisms remain unclear. The draft consults the EU’s act, which regulates AI based on risk levels involved in its use, to identify the highest risks, such as the use of real-time remote biometric identification systems in public places for the purpose of law enforcement. However, Taiwan’s draft AI Basic Act tasks the Ministry of Digital Affairs (Moda) with developing an AI risk framework to regulate and supervise the lesser risk levels, without clearly defining how the ministry will interact with other government agencies to mitigate risks.
Despite the draft acknowledging that AI’s applications may aggravate social or economic isolation and cause involuntary unemployment, the act lacks provisions for basic human rights protections or remedial measures. Disadvantaged groups receiving “priority consideration” in AI applications and workers being protected under “relevant labor laws” raises questions about the act’s ability to safeguard privacy, prevent discrimination, and provide accountability in AI use.
The NSTC has defended the draft, stating that it does not allow AI developers to ignore existing regulations concerning data protection and financial or medical oversight. Nonetheless, ambiguity remains about whether the proposed framework is sufficiently robust to address human rights concerns and align with Taiwan’s established legal frameworks.
As the global AI landscape continues to evolve, Taiwan’s draft AI Basic Act must strike a careful balance between innovation and regulation. A basic law’s primary function is to establish general principles while allowing the flexibility necessary for technological advancements. Some industry voices have advocated for the creation of a regulatory sandbox – similar to the “financial regulatory sandbox” used for financial technologies – where AI products, services, and business models can be tested within a controlled environment to help the government refine future AI laws.
The draft act aims to encourage AI development by maintaining a flexible framework, with a primary focus on mitigating AI risks rather than imposing strict prohibitions. However, stakeholders like AmCham Taiwan’s tech-related committees recommend that the NSTC serve as the primary coordinating body to oversee administrative regulations, ensuring consistency and avoiding conflicting interpretations by different government entities. In particular, establishing an experimental environment could be crucial for stimulating innovation while effectively managing risks.
The NSTC’s oversight would be particularly effective, given its expertise in research, development, and innovation, which aligns with the draft’s focus on addressing key trends in AI. This oversight is critical to reinforcing Taiwan’s position in the global AI industrial value chain, as it allows for strategic coordination and alignment with global standards. Furthermore, the NSTC’s role in deepening industry-academia collaboration and cultivating AI professionals is essential for building the foundational infrastructure needed to implement the AI Basic Act effectively.
The path to effective governance
For Taiwan’s AI regulatory framework to be both practical and future-proof, meaningful stakeholder engagement is essential. Continuous dialogue with industry players, academics, civil society groups, and international stakeholders can help fine-tune the regulatory approach, ensuring it is relevant and adaptable to the rapidly changing technology landscape. Public hearings and consultations, for instance, can gather valuable input on how best to strike a balance between ethical considerations and technological innovation.
An AmCham Taiwan member and leader in a global technology company underscores the importance of stakeholder engagement, noting that collaboration with trade associations and industry experts will allow Taiwan’s AI governance framework to reflect real-world needs and challenges accurately.
“Early communication with stakeholders is crucial,” they say. “It ensures that businesses understand their responsibilities and can provide input that helps shape regulations that are practical and effective.”
Stakeholder engagement also plays a crucial role in aligning Taiwan’s AI regulations with international standards. A well-structured public consultation process would not only enhance the quality of the AI Basic Act but also strengthen Taiwan’s position as a leading technology hub.
Moreover, early communication and education on AI regulations would aid businesses in understanding their compliance responsibilities, especially in the development of high-risk AI applications. This approach will not only support local companies but also boost Taiwan’s appeal as a trustworthy and proactive technology hub for global investors.
Effective AI governance requires a targeted approach that allows for the careful differentiation of risks across applications. AmCham Taiwan supports a risk-based framework that aligns with international standards, such as the U.S. National Institute of Standards and Technology (NIST) AI Risk Management Framework or ISO/IEC 42001. By focusing regulatory attention on high-risk AI applications while maintaining flexibility for less critical applications, Taiwan’s framework can help safeguard public welfare without hampering technological progress.
A core challenge for the draft AI Basic Act is defining roles and responsibilities within the AI ecosystem. AI developers, deployers, and end-users each play distinct roles in ensuring the safe and ethical use of AI, especially in line with the act’s basic principles: autonomy, confidentiality, safety, inclusion, and transparency. Yet the current draft offers only vague guidelines about accountability, saying that the government will prevent researchers, developers, or users of AI from abusing their dominant positions without clarifying how the government, or which government body, would ensure that human-centered values are fulfilled.
Stakeholders argue that a clearer delineation of roles is essential. For instance, developers could be responsible for ensuring data privacy and model accuracy, while deployers should be accountable for transparency and risk management in real-world applications. End-users, meanwhile, must understand the implications of the AI tools they use, especially in contexts involving sensitive data or decision-making.
Taiwan’s AI Basic Act seeks to enable responsible innovation without compromising essential oversight. The draft proposes a flexible, principle-based framework, particularly evident in provisions such as Articles 4 and 5, which offer broad regulatory guidance rather than strict mandates. This flexibility is crucial given the rapid pace of AI development, where regulations must be adaptable to new use cases and evolving technologies. AmCham Taiwan and other stakeholders caution against overly prescriptive regulations that could stifle growth. Instead, a framework-style approach is recommended, focusing on enabling innovation while setting clear boundaries to mitigate risks, particularly in high-risk domains like healthcare and law enforcement.
Another critical issue is the question of how Taiwan’s draft AI Basic Act will address the complexities of data security and cross-border data flows. Given the global nature of AI development and Taiwan’s integration into the international technology ecosystem, it is essential for the act to provide clear guidelines on managing data privacy, sharing, and localization – issues that are central to maintaining Taiwan’s competitive edge in AI and related fields.
Another AmCham Taiwan member from a leading international technology company commented that the draft places a heavier focus on regulating AI based on its country of origin, which introduces concerns about how international collaborations and technologies will be treated and whether this regulation may hinder innovation.
One of the reasons for restrictions on the place of origin of AI systems is national security. Algorithms can be manipulated to include backdoors, making cybersecurity a central issue in AI governance. For instance, an AI system developed in a country with differing political interests might embed vulnerabilities or biases that could compromise sensitive applications in defense or infrastructure.
While recognizing the importance of national security, a blanket ban on AI based on place of origin may stifle innovation and hinder international collaboration. Policymakers should consider a more nuanced approach, such as setting stringent cybersecurity standards and promoting transparency in data use, to mitigate risks without curbing technological advancements.
Despite these considerations, limited focus is given to the implications for national security. The draft’s provisions on cybersecurity protection (as highlighted in Article 10) should prioritize alignment with international standards to foster seamless global collaboration while incorporating tailored measures to address Taiwan’s national security needs and bolster resilience against emerging threats.
Paving the way
The success of Taiwan’s AI Basic Act will largely depend on its ability to strike a balance between providing sufficient regulatory oversight and enabling innovation. As AI becomes increasingly critical to economic and social development, the need for clear and practical governance frameworks cannot be overstated. The act’s focus on establishing a flexible yet accountable system could pave the way for Taiwan to take a leading role in AI development, not just as a hardware powerhouse but also as a model for thoughtful, human-centric AI governance.
However, this goal will require ongoing dialogue, both domestically and internationally. Engaging with trade associations, academia, and civil society groups will be key to ensuring that Taiwan’s AI regulations are not only robust but also adaptive to the fast-moving nature of technology. As the NSTC refines the AI Basic Act before it is submitted to the Executive Yuan, early and continuous stakeholder engagement will be essential to paving the way for smooth implementation, avoiding ambiguities, and ensuring that the law can accommodate future innovations.
Taiwan’s draft AI Basic Act is an ambitious attempt to regulate a transformative technology that is evolving at a breakneck pace. By focusing on risk-based regulation, accountability, and stakeholder engagement, the draft could lay a solid foundation for AI governance that serves the dual goals of protecting public welfare and promoting responsible innovation. How the draft evolves in the coming months will be critical in shaping Taiwan’s role in the global AI ecosystem – one that balances technological leadership with a commitment to ethical, transparent, and forward-thinking governance.
