Taiwan Frontlines: Taiwan sees AI governance as national security

Taiwan Current News (TCN), in partnership with the German Marshall Fund of the United States (GMF), produces the podcast series Taiwan Frontlines, which tracks global trends and explores how Taiwan connects with the world.

In the latest episode, Isabel Hou (侯宜秀), Taiwan's political deputy minister of digital affairs, told hosts Bonnie Glaser and Jason Hsu (許毓仁) that Taiwan sees artificial intelligence governance and national security as inseparable, while the island confronts mounting cyber pressure and the risks posed by increasingly capable AI systems.

Hou, who took office in September 2025 and helped advance Taiwan's AI Basic Act, said the government's approach is intended to support innovation while building safeguards around AI deployment, critical infrastructure and information integrity.

Trust as an AI security issue

Hou said Taiwan can contribute to a trusted open-weight AI ecosystem by working with democratic partners on safeguards, evaluation tools and Traditional Chinese-language data.

As AI agents and robots become more closely integrated into daily life, she said, public trust will become essential.

Hou argued that Taiwan's democratic values, human-rights protections and emphasis on privacy could help make open-source and open-weight models more trustworthy.

She noted that many open-weight models currently come from China, but said their compliance with Chinese regulations has raised concerns over censorship and political value alignment. Taiwan is seeking to develop tools to evaluate the extent to which models may reflect censorship by the Chinese government, she said, while also contributing data to train more trusted open-source models with international partners.

“I think that's what Taiwan can contribute,” she said.

A growing cyber burden

Citing a National Security Bureau report showing an average of 2.63 million cyber intrusion attempts targeting Taiwan's critical infrastructure each day last year, Hou said targeted agencies, virtual private network services and external websites, using common attack methods or automated login attempts to identify weaknesses.

She said, “it is like someone... consistently trying every door handle every day to see which door is left unlocked.”

The priority, she noted, is to ensure that systems are protected and that newly discovered vulnerabilities are patched faster than adversaries can exploit them.

Hou said Taiwan shares threat intelligence in real time through joint defense systems and cybersecurity alerts, allowing agencies to adjust their defenses quickly when an attack is detected. She added that the focus of cyberattacks is increasingly shifting from data theft toward disruption of essential civilian sectors, including energy and healthcare, and threats to social stability.

Taiwan is therefore expanding the scope of cyber-defense drills from its highest-tier critical facilities to a broader range of important infrastructure, she said. 

Regulation has limits

Hou cautioned that regulation alone cannot neutralize global AI-enabled threats. Taiwan's risk-classification and assessment framework applies to government agencies and businesses that legally deploy AI in Taiwan, but it cannot reach malicious actors operating outside the country's jurisdiction, she said.

“Overstating what regulation can do is itself a risk,” Hou said. 

She compared regulation to a public health system rather than an immune system, in that it can help legitimate actors deploy AI responsibly and impose resilience and reporting obligations, but it cannot by itself stop cyberattacks or malicious open-weight models circulating globally.

Countering such threats instead requires defensive capabilities, real-time threat intelligence, international cooperation and engagement with frontier AI laboratories, she said.

Lessons for democratic partners

Hou said Taiwan's experience offers a practical lesson for the US, Europe and other partners because the island is managing AI governance while facing ongoing cyber and information pressure from China.

For Taiwan, she said, testing whether AI models reproduce authoritarian narratives is not only an ethical question but a national security concern. She also argued that citizen participation can strengthen societal resilience against information warfare, citing Taiwan's experience with civic tech communities and fact-checking networks.

Taiwan, Hou said, is seeking a flexible "third path" between a more market-driven US approach and the European Union's heavier regulatory model — a framework law that establishes principles while leaving room for regulators and stakeholders to adapt as technology evolves.

For more in-depth coverage, listen to Taiwan Frontlines on NOWNEWS' official YouTube channel.