The use of Chinese artificial intelligence chatbots presents challenges in terms of personal data security and information censorship
The Chinese artificial intelligence (AI) chatbots, DeepSeek and Qwen, are being promoted as alternatives to Western AI models, such as ChatGPT. While this demonstrates China’s ambition and capability to compete with the West, it also poses challenges for users in terms of information censorship and personal data security when using these chatbots.
INFORMATION CENSORSHIP
Both chatbots provide users with information that is consistent with the regime’s ideology, avoiding responses that would portray China negatively. For example, they avoid discussing Taiwanese independence or the Tiananmen Square protests. DeepSeek repeats the statements of Chinese politicians and uses the pronoun ‘we’ when answering questions about China’s position to show its identification with China.
In February 2025, Politico published a chat with DeepSeek asking why China had blocked imports of goods from Lithuania. The chatbot expressed hope that Lithuania would correct its mistakes and cooperate with China. This response essentially echoed the call made by the Chinese Ministry of Foreign Affairs spokesperson Wang Wenbin in 2023 for Lithuania to ‘correct its mistakes and return to the right path by supporting the one-China principle’. Such responses, alongside with the nature of Chinese chatbots, suggest that biased information is presented to users.
PERSONAL DATA SECURITY
DeepSeek collects the following types of information: data created by users themselves, such as query texts and various documents shared with the chatbot; automatically accessible information, such as device data and IP address; third-party data, such as information linked to a Google account. According to Chinese law, the government is entitled to access data collected by technology companies. Chinese-developed AI software possibly lacks adequate cybersecurity measures necessary to protect the user data. Shortly after its launch, more than a million pieces of data collected by the DeepSeek chatbot were leaked, including chat system metadata and chat transcripts.

Reuters photo