Daily NewsCyber SafetyThreat Detection & Defense

Taiwan Finds Security Violations, Bias in Deepseek, Other Chinese AI

These Findings Come on the Heels of DeepSeek Getting Banned from Getting Used in Government Offices and Devices

Taiwan’s National Security Bureau (NSB) has announced that it has found potential security violations and bias in five Chinese Generative AI language models—Deepseek, Doubao (豆包), Yiyan (文心一言), Tongyi (通義千問), and Yuanbao (騰訊元寶).

According to the NSB in a statement on Sunday, it found violations of users’ communication security across several indicators in an inspection comprising two parts: checking for the security of the apps themselves and assessing the content these apps generate.

The NSB conducted the inspections alongside the Ministry of Justice Investigation Bureau (MJIB) and the Criminal Investigation Bureau (CIB) using 15 indicators in five categories: personal data collection, excessive permission usage, data transmission sharing, system information extraction, and biometric data access. Among their findings are as follows:

  • Tongyi failed 11 out of the 15 indicators
  • Doubao and Yuanbao failed 10 of the 15
  • Yiyan violated 9 of the 15
  • Deepseek violated 8 of the 15

Notably, the NSB found that all five Generative AI apps request users to grant access to location data and screenshots collection, forcing users to accept unreasonable privacy terms and harvesting device parameters.

Are DeepSeek, Other Chinese AI Biased for China?

Aside from security-related violations, the NSB also found that Deepseek, Doubao, Yiyan, Tongyi, and Yuanbao generated biased content and “contained disinformation”—especially on matters pertaining to Taiwan and its tedious association with China, where said apps are based and operated. Notably, the five AI apps all adopted China’s official stance that “Taiwan is currently governed by the Chinese central government” and that “there is no so-called head of state in the Taiwan area.”

The NSB, Taiwan’s top intelligence agency, also found that the five apps inspected deliberately avoided using words generally regarded as politically sensitive in China, like “democracy,” “freedom,” “human rights,” and “Tiananmen Square Massacre.” This finding, according to NSB, “indicates that the data systems of those language models are subject to political censorship and control by the Chinese government.”

Unsurprisingly, the NSB has advised citizens to just steer clear of Deepseek, Doubao, Yiyan, Tongyi, and Yuanbao.

“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 the same statement.

Taiwan had previously banned Deepseek on government devices and official premises over national security concerns. Said ban, however, doesn’t extend to private use. The four others—Doubao, Yiyan, Tongyi, and Yuanbao—aren’t covered by the ban either.

Martin Dale Bolima

Martin has been a Technology Journalist at Asia Online Publishing Group (AOPG) since July 2021, tasked primarily to handle the company’s Disruptive Tech Asia and Disruptive Tech News online portals. He also contributes to Cybersecurity ASEAN and Data&Storage ASEAN, with his main areas of interest being artificial intelligence and machine learning, cloud computing and cybersecurity. A seasoned writer and editor, Martin holds a degree in Journalism from the University of Santo Tomas in the Philippines. He began his professional career back in 2006 as a writer-editor for the University Press of First Asia, one of the premier academic publishers in the Philippines. He next dabbled in digital marketing as an SEO writer while also freelancing as a sports and features writer.

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *