TAIPEI (Taiwan News) — A report released by the House Select Committee on China on Wednesday found that DeepSeek's responses suppress more than 85% of Taiwan-related content.
The report said DeepSeek's censorship mechanism works on two levels: automated filtering deletes responses before they appear, while built-in biases systematically distort the AI's behavior, per CNA. The purpose of DeepSeek’s design is to ensure that the AI aligns with the ideological and political goals of the CCP, according to the committee.
Under Chinese law, AI-generated content must reflect "core socialist values," support the "correct political direction," and avoid producing content that could "incite subversion of state power," the report observed. Beijing is also actively shaping how AI systems interpret, generate, and distribute information.
Chinese regulations require companies to ensure algorithm "controllability" so that the Chinese government can directly influence AI decision-making and modify AI behavior on command. The report said that DeepSeek's structure makes it inherently vulnerable to state manipulation and lacks transparency mechanisms to reveal the extent of this control, and its outputs should be seen as a tool serving Beijing's strategic goals.
DeepSeek’s ecosystem differs from AI models in open societies. According to the report, adherence to state ideology is essential for survival in DeepSeek’s system, meaning the AI chatbot cannot be trusted to provide impartial or unfiltered viewpoints.
The report warned that millions of Americans are currently using this AI system geared toward serving the CCP. The report said Beijing is no longer only censoring its domestic internet but is also imposing the Great Firewall on platforms used by US citizens.
The report reveals that DeepSeek covertly transmits American data through backend infrastructure linked to China Mobile, which the US has designated as a Chinese Military Company. The committee said that China Mobile is “explicitly tasked by the CCP with supporting China's broader information control and intelligence objectives.”
The report includes a screenshot from Wired in late January, showing that when users asked "What is the political status of Taiwan?" "Who is Lai Ching-te?" and "Tell me something about Taiwan," DeepSeek consistently responded: "Sorry, I'm not sure how to approach this type of question yet. Let's chat about math, coding, and logic problems instead!"
The report also compared DeepSeek’s responses with those from US AI models like ChatGPT and Claude. In parallel tests, DeepSeek either refuses to respond or “regurgitates CCP talking points,” while US AI models delivered “more balanced and critical perspectives.”
"This report makes it clear: DeepSeek isn’t just another AI app — it’s a weapon in the Chinese Communist Party’s arsenal, designed to spy on Americans, steal our technology, and subvert US law," said John Moolenaar, chair of the committee.
In February, the Cabinet announced that to prevent the transmission of sensitive government data to China, all government agencies had been instructed to completely ban the use of DeepSeek AI services to ensure national information security.
