Written by: Haim Ravia, Dotan Hammer
Chinese Deepseek services have been gaining traction as cost-effective and “open” alternatives to artificial intelligence tools offered by OpenAI, Meta, and Google. However, despite their rising popularity, these services come with significant challenges—most notably, the censorship imposed by the Chinese government.
One review of Deepseek reveals that its chat service refuses to respond to politically sensitive questions that are censored in China. Topics such as the Tiananmen Square incident, the identity of Xi Jinping (President of the People’s Republic of China), and Taiwan’s status as an independent country are systematically avoided.
In other cases, the service produces answers aligning with the Chinese government’s official stance. For instance, it declares that “Taiwan has always been an integral part of China” and affirms that “the Chinese government resolutely opposes any form of “Taiwan independence” separatist activities.” It further concluds that “any attempts to split the country are doomed to fail”.
These findings raise concerns about bias in Deepseek’s output. Even when used in its “Source Available” configuration, the service remains subject to Chinese law.
Moreover, users of the “Source Available” versions of Deepseek must adhere to the Deepseek License Agreement, which imposes restrictions on the service’s use—even in its on-premises configuration—and limits the distribution of derivative works. API and cloud users are also subject to terms of use that allow Deepseek to use inputs to enhance its services and share the inputs and outputs with “certain entities” within the Deepeek Chinese group of companies.
Users are advised to exercise caution when providing content to Deepseek services and to critically evaluate the outputs produced. Diligence is essential to identifying potential biases or alterations in the AI-generated content that could undermine the quality and integrity of the generated output.
Click here to read the Deepseek License Agreement.
