TAIPEI (Taiwan News) — Research Institute for Democracy, Society, and Emerging Technology warned Wednesday that Chinese AI services could compromise privacy and manipulate public opinion.
DSET’s report analyzed privacy policies from 10 Chinese generative AI services, per CNA. It identified three data-transfer pathways: direct storage of overseas user data in China, corporate data sharing, and compliance with government “lawful” data access requests.
At a Taipei briefing, DSET Democratic Governance Program Deputy Director Lai Yu-hao (賴又豪) said the findings expose risks rooted in China’s data governance model. The report warns such practices could place users in democratic countries under authoritarian oversight.
Lai said services including DeepSeek and Baidu AI Search are operated by China-registered entities and store overseas users’ data in China. Others, such as Monica AI and Talkie AI serve overseas users through entities registered in places like Singapore.
With the exception of Talkie, Lai said the privacy policies of the remaining services allow intra-group data sharing, meaning data may still flow to China even if a service is registered elsewhere. All 10 services also state they comply with government data access requests deemed lawful.
He cited Xiaohongshu as an example, noting that after Taiwan restricted access to the app last year, the company routed overseas users through a Singapore-based entity. However, its privacy policy still allows data sharing within the corporate group.
To mitigate the risks, Lai proposed a “DSR” strategy: “Defend” by barring Chinese-controlled AI from government use, “screen” through stricter market entry and data governance rules, and “rally” allies to align cross-border data regulations.
