Taiwan has enacted the Artificial Intelligence Basic Act to establish a legal framework for the technology, but experts warn that the island faces significant hurdles in data availability and computing power as it seeks to build “Sovereign AI.”
While the legislation marks the official start of AI implementation in Taiwan, the transition from legal text to technological autonomy is complicated by a shortage of Traditional Chinese language data and sufficient GPU resources, according to analysts.
The Data Deficit
Chu Chentso (朱宸佐, Wesley Chu), a researcher at the Harvard Kennedy School, said the development of sovereign AI—a global strategic trend aimed at ensuring localized control over computing power and data—relies on training models that reflect a nation's cultural values.
However, Taiwan's push for autonomous large language models, such as TAIDE, is hindered by a limited volume of Traditional Chinese corpora.
"Insufficient data volume and data errors will lead to biased model computation results," Chu said.
Although Article 13 of the AI Basic Act mandates the government to establish data opening and sharing mechanisms, progress has been slowed by commercial barriers, including copyright fees and licensing concerns from private institutions and media outlets.
To address these gaps, the Ministry of Digital Affairs is promoting a draft act that aims to improve the quality of open government data through legal incentives and establish routine public-private data sharing.
Computing Power and 'Friend-Shoring'
Beyond data, hardware remains a critical bottleneck. Chang Li-ching(張麗卿), chair professor at Shih Chien University and executive director of the AI Legal International Research Foundation, noted that while Taiwan is a global leader in semiconductor manufacturing, it lags behind international giants in the allocation of national-level GPU resources.
Chang proposed a "parallel approach" involving the nationalization of infrastructure and the "friend-shoring" of supply chains. She argued that Taiwan must strengthen its computing autonomy while cooperating with trustworthy nations to build an AI ecosystem resilient to geopolitical tensions.
"This is not just a technical issue, but a trust issue," Chu added.
He noted that Article 12 of the AI Basic Act explicitly promotes international cooperation to ensure Taiwan can align with global standards in computing power, critical materials, and firmware development.
Protecting the Vulnerable
As the law moves toward implementation, experts are emphasizing the need for social equity and protections for digitally vulnerable groups.
Chang highlighted that the Basic Act addresses these concerns through principles, procedures, and remedies. Article four requires the consideration of social equity to reduce digital divides.
For high-impact scenarios such as financial credit scoring or medical diagnoses, Chang warned against opaque algorithms.
"Differential treatment absolutely cannot become black-boxed," she said, advocating for bias detection, traceable records, and human review mechanisms to ensure citizens have the right to appeal.
The act also addresses potential labor disruptions. Article 15 stipulates that the government must provide career transition guidance for workers displaced by AI. Additionally, Article 17 establishes liability frameworks for high-risk applications, ensuring that relief and compensation mechanisms are in place.
