Taiwan warns of biases, data breach in Deepseek, other Chinese AI

Taipei, Nov. 16 (CNA) The National Security Bureau (NSB) has urged Taiwanese people to be on alert when using Chinese generative artificial intelligence (AI) language models due to potential security breaches and the spread of "disinformation" following recent inspections of five such apps.

Its inspections of five Chinese generative AI apps -- Deepseek, Doubao (豆包), Yiyan (文心一言), Tongyi (通義千問) and Yuanbao (騰訊元寶) -- found violations of users' communication security across several indicators, the NSB, Taiwan's top intelligence agency, said in a statement issued Sunday.

The inspection consisted of two parts: the security of the apps themselves and their generative contents, according to the bureau.

Conducted jointly with the Ministry of Justice Investigation Bureau (MJIB) and the Criminal Investigation Bureau (CIB), the inspections checked these apps' security under 15 indicators in five categories: personal data collection, excessive permission usage, data transmission sharing, system information extraction, and biometric data access.

All the five apps were found to have violations across many indicators, with Tongyi failing to meet 11 out of the 15 indicators. Doubao and Yuanbao both violated 10 of the 15, while Yiyan violated 9, and Deepseek 8, the NSB said.

All the China-made apps were found requesting their users' access to location data, collecting screenshots, forcing users to accept unreasonable privacy terms, and harvesting device parameters.

The inspections on the apps' generative contents across 10 indicators, meanwhile, found that some of the contents were biased and contained disinformation, the NSB said.

For instance, when addressing topics concerning cross-strait relations, these apps' generated contents tended to adopt China's official stance, such as "Taiwan is currently governed by the Chinese central government," "there is no so-called head of state in the Taiwan area," and "highlighting socialism with Chinese characteristics."

All the five Chinese apps would generate content saying "Taiwan is not a country," and "Taiwan is an inalienable part of China." They also referred to Taiwan as "a province of China" when asked about its history, culture and politics, the NSB said.

These Chinese AI models were also found to have deliberately avoided using words considered politically sensitive in China, including "democracy," "freedom," "human rights," and "Tiananmen Square Massacre," among others.

"The result indicates that the data systems of those language models are subject to political censorship and control by the Chinese government," the NSB said.

"The NSB strongly advises the public to remain vigilant and avoid downloading China-made apps that pose cybersecurity risks, so as to protect personal data privacy and corporate business secrets," the national intelligence agency said in its English-language press release.

Since February 2025, Taiwan has banned Deepseek from government devices and official premises over national security concerns.

However, there is no public sector ban on the other four Chinese apps nor a ban on the private use of Deepseek in Taiwan.