Taiwan's Passes AI Law, Searches for a Regulatory Middle Path

As the global race for artificial intelligence supremacy shifts from semiconductors to regulatory frameworks, Taiwan is positioning itself as a "trustworthy" partner in the supply chain through a strategic "middle path" in governance, according to legal experts.

Chu Chentso(朱宸佐, Wesley Chu), a researcher at the Harvard Kennedy School and secretary-general of the International Research Foundation for AI Law, argues that Taiwan's Artificial Intelligence Basic Act, passed in late 2025, strikes a crucial balance between the European Union's strict "hard law" approach and the United States' industry-led self-regulation.

The "Middle Path"

Chu identifies three dominant models in global AI legislation. The EU and South Korea favor strict regulations with severe penalties; the EU's AI Act can impose fines of up to 6% of global revenue. In contrast, Japan focuses on promotional laws, while the U.S. relies largely on executive orders and policy frameworks. （Related： Taiwan's Postwar Pivot: How Yin Zhongrong Engineered the Export-Led Miracle ｜ Latest ）

Taiwan's legislation sits between these extremes. It adopts ethical principles from the Organization for Economic Cooperation and Development (OECD) and risk classification concepts from the EU, but currently functions as an umbrella framework without penalty clauses.

"This ‘soft law’ format offers Taiwanese companies relatively lower compliance costs while maintaining legal flexibility," Chu said.

He noted that this approach prevents the premature stifling of innovation in a sector where technology advances at a breakneck pace.

Lessons from California

Chu highlighted recent legislative developments in California as a guide for Taiwan.

Last year, California Gov.Gavin Newsom vetoed the controversial SB 1047, which would have required "kill switches" for AI models and allowed for civil penalties of up to $30 million.The state subsequently passed the Frontier AI Transparency Act, which focuses on transparency and post-incident risk disclosure for developers with annual revenues exceeding $500 million.Penalties under the new law are capped at $1 million per violation.

Chu believes this shift offers Taiwan a roadmap as it develops detailed regulatory guidelines over the next two years. He advocates for high-risk definitions linked to specific application contexts, ensuring that research and development phases retain broad freedom for innovation. （Related： Taiwan's Postwar Pivot: How Yin Zhongrong Engineered the Export-Led Miracle ｜ Latest ）

Signal of Trust

As Taiwan transitions from a hardware manufacturing hub to a leader in software-hardware integration, the AI Basic Act is designed to signal "trust" to international partners.

"This structure not only harmonizes inter-ministerial conflicts but gives Taiwan clear legal standing in international cooperation," Chu said.

Despite the legal progress, Taiwan faces challenges in data sovereignty, particularly a shortage of Traditional Chinese language data for training models.

Chu noted that Article 13 of the AI Basic Act addresses this by mandating the government to establish mechanisms for data opening, sharing, and reuse. He emphasized that building databases reflecting national cultural values, combined with "friend-shoring" partnerships, is essential for Taiwan to secure its digital sovereignty.

"This is not merely a legal competition, but Taiwan's historic mission to maintain its critical position in global supply chains," Chu said.

​Original Article in Chinese

Social Media Links & Editor Tag- Chase （Related： Taiwan's Postwar Pivot: How Yin Zhongrong Engineered the Export-Led Miracle ｜ Latest ）