TAIPEI (Taiwan News) — Taiwan’s Ministry of Digital Affairs on Friday issued a security alert for AI models developed by Chinese startup DeepSeek, urging government agencies and critical infrastructure operators to restrict their use due to national cybersecurity risks.
DeepSeek, a Hangzhou-based company, faces criticism for its AI model's self-censorship on sensitive topics. The model restricts content related to issues like Taiwan's sovereignty and the 1989 Tiananmen Square Massacre, and avoids detailed responses about Chinese leader Xi Jinping (習近平).
MODA cautioned that DeepSeek AI products could expose sensitive user data to platforms with potential security vulnerabilities, posing a threat to Taiwan’s information security. The ministry advised public institutions to refrain from using the technology.
Under the Cabinet's AI policy, public sector employees are prohibited from inputting confidential, personal, or unauthorized information into AI platforms. They are also barred from asking AI models about classified business matters or personal data.
The ministry emphasized that DeepSeek AI presents cybersecurity threats, including cross-border data transmission and potential leaks. As a result, it has been classified as a national security risk, and public institutions have been urged to remain cautious.
Beyond Taiwan, the US, UK, France, Germany, Italy, and Japan have also raised concerns about DeepSeek, with several countries taking countermeasures. Authorities cite cybersecurity risks, potential misuse of US AI models, and fears that the technology could serve as a tool for Chinese propaganda.
US Commerce Secretary nominee Howard Lutnick has accused DeepSeek of stealing American technology, alleging that the company relies on US AI models and Nvidia chips for its development. Alexandr Wang, founder of the Chinese AI startup Scale AI, told CNBC that DeepSeek has acquired a surprising number of Nvidia’s H100 chips, reportedly around 50,000 units.
The revelation has sparked concerns, as the Biden administration restricted the sale of advanced AI chips to China in 2022. DeepSeek’s ability to obtain such components has raised suspicions about its connections.
UK Technology Secretary Peter Kyle stated that British authorities are assessing DeepSeek’s chatbot for potential security risks. Meanwhile, South Korea and France's data protection agencies are planning to seek clarification from the company on how its AI chatbot operates and the related data protection concerns.
Italy has ordered emergency restrictions on DeepSeek’s AI platform, preventing it from accessing Italian users' data. Following intervention, DeepSeek has been removed from Apple’s App Store and Google Play Store in Italy.