DSET visited the National Democratic Institute (NDI) for a meeting with members of NDI’s Indo-Pacific Team and the Democracy Innovation Lab, focusing on how democracies can respond to the global expansion of Chinese artificial intelligence (AI) services and their growing influence on governance and public information systems.
Dr. Kai-Shen Huang, Director of DSET’s Democracy and Governance Program, presented his research on China’s AI export strategies in a talk titled “Chinese AI Services and the Need for Input Review.” He emphasized that in today’s era of technological geopolitics, it has become unavoidable for democracies to establish review mechanisms for inputs into Chinese AI systems. Dr. Huang observed that global AI governance discussions often focus on the race for cutting-edge model development, while overlooking the application-centered approach that gives China’s AI ecosystem its global reach.
Through large-scale experimentation in smart cities, healthcare, and transportation, China’s AI firms have created a cycle of technological optimization and commercial export, reinforced by state policy that promotes the outward diffusion of Chinese technical standards and governance norms.
Dr. Huang described this as a dual structure of “tight control at home and outward expansion abroad.” This model allows Chinese AI services to be both globally competitive and politically sensitive. The central challenge for democracies, he argued, is how to build effective input review mechanisms to guard against systemic risks and dependency, while maintaining openness and innovation.
He further noted that China’s application-driven model means that most AI products achieve commercial-scale maturity before being exported, creating competitive advantages in speed and adaptability. However, U.S. efforts to regulate data flows from China face major challenges, as China’s Personal Information Protection Law, Data Security Law, and Cybersecurity Law lack transparency, leaving foreign governments with limited visibility and policy tools.
Algorithmic governance presents another obstacle. Since 2020, China has included personalized recommendation algorithms and AI interaction interface technologies in its Catalogue of Technologies Prohibited or Restricted from Export, effectively restricting the sale of these systems abroad except through government-approved licensing.
Following Dr. Huang’s presentation, You-Hao Lai, Deputy Director of DSET’s Democracy and Governance Program, shared findings from his forthcoming policy report, “The Authoritarian Gaze: China’s Global Data Reach and Systemic Risks to Democracy.” Lai argued that democracies must not only focus on their own AI development but also remain vigilant toward the cross-border infiltration of authoritarian AI.
He analyzed the authoritarian features of China’s data governance model and warned that such practices are creating one-way data flows—information that flows into China but rarely out—reinforcing global data asymmetries. Lai emphasized that the data practices of Chinese AI services could be entrenching authoritarian data governance worldwide, posing systemic risks to both individuals and democratic institutions.
He proposed three policy recommendations for democratic governments:
- Prohibit the use of Chinese AI services and derivative applications in government and critical infrastructure systems.
- Establish inbound review mechanisms for Chinese AI and digital services, focusing on privacy, data security, and institutional risk.
- Develop a multilateral framework for democratic data governance to strengthen cross-border oversight and coordination.
During the discussion, participants raised questions about Taiwan’s efforts in personal data protection, China’s potential use of large language models for influence operations, and the environmental footprint of AI models such as DeepSeek. When asked whether existing U.S. mechanisms—such as CFIUS (Committee on Foreign Investment in the United States) and the International Emergency Economic Powers Act (IEEPA)—are sufficient, Lai responded that they are insufficient for addressing current challenges. He explained that CFIUS mechanisms are triggered only by investment transactions, which means they fail to apply when Chinese AI services collect data directly from U.S. users without investment involvement.
Lai concluded that democracies must develop new governance tools capable of addressing digital infiltration and AI-driven asymmetries—an essential step in safeguarding democratic institutions in the age of artificial intelligence.
