On November 3, DSET met with Freedom House in Washington, D.C., for an in-depth discussion on China’s AI global expansion and its implications for democratic governance. The meeting was hosted by Adrian Shahbaz, Vice President for Research and Analysis at Freedom House, and featured presentations by DSET’s Democratic Governance Program on China’s application-driven AI export strategy and the authoritarian characteristics of Chinese AI services’ data practices.
Kai-Shen Huang, Director of DSET’s Democratic Governance Program, outlined China’s approach to AI development, noting that global discussions often focus on frontier model development while overlooking China’s application-driven model that scales AI products domestically before export. Huang emphasized that recent Chinese policy documents explicitly link data governance, model development, and commercialization to promote overseas expansion, with state support including taxation incentives and data centers built in free trade zones. Using the TikTok case as an example, Huang highlighted two critical regulatory obstacles for democratic nations: China’s opaque data laws restricting outbound data transfers, and export controls prohibiting the transfer of core AI algorithms. These restrictions make it extremely difficult to verify the occurrence of technology transfers, even when licensing agreements appear in place.
You-Hao Lai, Deputy Director of DSET’s Democratic Governance Program, shared key findings from his forthcoming policy report, which analyzes the privacy practices of major Chinese AI service providers.
He noted that many services store user data directly in China or enable internal data sharing across corporate affiliates, providing broad channels for government access. He warned that China’s data governance model, marked by its “all in, rarely out” data flow, is not only asymmetrical but also institutionally authoritarian in nature.
To mitigate the risks posed by this expanding data regime, Lai proposed three key policy recommendations for democratic countries: (1) ban the use of Chinese AI services in government agencies and critical infrastructure; (2) establish a formal inbound review mechanism for Chinese AI providers; and (3) build coordinated frameworks for cross-border data governance among democratic allies.
In response to DSET’s presentation, Shahbaz raised several critical questions. He emphasized the need to communicate the urgency of data governance to the public by highlighting real-world examples of data misuse, especially as public concern over China’s cross-border data practices continues to normalize. Shahbaz also asked how regional trade agreements, such as the Regional Comprehensive Economic Partnership (RCEP), might be used to regulate cross-border data transfers.
Lai responded that China is actively promoting its model of data governance through bilateral and multilateral agreements. In response, democratic allies must develop their own coherent frameworks, establishing unified standards to ensure secure data flows and to limit transfers to China. He also shared that Taiwan has recently established an independent data protection authority following a Constitutional Court ruling. However, he acknowledged that current laws on cross-border data transfers remain underdeveloped and urgently require reform to address the specific challenges posed by China’s data practices.
The discussion concluded with mutual recognition of complementary approaches between DSET’s security-oriented analysis grounded in liberal values and Freedom House’s values-first approach that incorporates security perspectives. Both organizations expressed strong interest in future collaboration examining democratic responses to authoritarian AI expansion and exploring multilateral frameworks for trustworthy data governance among like-minded nations.
