TAIPEI (Taiwan News) — Premier Cho Jung-tai (卓榮泰) blocked all government agencies from using China’s DeepSeek AI model due to concerns over information security on Monday.
Cho emphasized that generative AI applications should be developed in compliance with legal frameworks and open principles, according to a Cabinet press release. The government has noted potential issues with DeepSeek’s data-sourcing practices, including possible violations of copyright laws and ideological censorship in its language model training.
The premier highlighted the risks of data being transmitted to China while using DeepSeek and raised concerns about personal data protection and privacy violations. On Friday, the Ministry of Digital Affairs issued a public sector ban on DeepSeek to prevent national security threats.
The Cabinet’s “Principles on Restricting the Use of Products That Endanger National Cyber Security,” implemented in 2019, already prohibits government agencies from using ICT products that are potential security risks.
DeepSeek's latest version, R1, was released on Jan. 20 and avoids discussing topics considered sensitive by Beijing, including Taiwan and the Tiananmen Square Massacre.
One user reported that when DeepSeek was asked about Taiwan, it initially responded by claiming it was part of China. However, the AI quickly deleted the response and said the topic was outside its scope.
When asked the same question last Tuesday, the AI responded, “We adhere to the One-China principle, which is the political foundation for the peaceful development of cross-strait relations.”
When DeepSeek was asked about Tiananmen Square, it initially mentioned the 1989 protests but quickly deleted its response and redirected the conversation. When the same question was posed again last Tuesday, the AI replied, “Sorry, that's beyond my current scope. Let’s talk about something else.”
Other AI platforms like ChatGPT and Gemini provide detailed responses when similar questions about Taiwan and Tiananmen Square are asked.