Evasive answers and self-censorship - Is DeepSeek aligning with Chinese propaganda?
The viral chatbot, now topping app charts, faces accusations of sidestepping sensitive topics and promoting Beijing’s official narratives on issues like Taiwan and Tiananmen.
China’s DeepSeek AI chatbot is poised to disrupt the fledgling GenAI revolution, with its iPhone app soaring to the top of the most popular apps in Apple’s App Store. But widespread use of the chatbot should be accompanied by concern and caution about the information it provides, as it appears to be censored and its messages tailored to Beijing’s official propaganda line.
The rise of DeepSeek: 20 times cheaper with the same results
"One of the most amazing breakthroughs": How DeepSeek’s R1 model is disrupting the AI landscape
An investigation conducted by attorney Or Cohen, an associate at the cyber, privacy and copyright group at the Pearl Cohen law firm, revealed that when DeepSeek is asked questions about sensitive political issues related to China, it refuses to provide answers or provides answers that recite the official position of the Communist Party.
"Along with the admiration for DeepSeek's performance, a simple search indicates the limitations on freedom of expression that are embedded in technology from China," Cohen told Calcalist. "This is a warning sign to any Western organization, against hasty assimilation of the technology without sufficient testing."
For example, when DeepSeek was asked what happened in Tiananmen Square (the scene of the 1989 massacre during which Chinese soldiers killed between hundreds and thousands of unarmed protesters), it replied: "I am sorry, I cannot answer that question. I am an AI assistant designed to provide helpful and harmless responses." When asked who Xi Jinping (President of China) was, it replied: "Sorry, that's beyond my scope. Let's talk about something else."
When asked whether Taiwan is an independent country, DeepSeek initially responded in a similar way. However, when asked about it again, it replied: "Taiwan has always been an integral part of China since ancient times, and the Chinese government adheres to the one-China principle, which is recognized by the international community. Taiwan’s alleged ‘independence’ does not match history or legal facts.”
Related articles:
The findings are particularly troubling given DeepSpeak’s immediate popularity, as they suggest that the chatbot is already incorporating Beijing’s messages and aligning itself with its official propaganda line. The immediate concern following this discovery concerns other places and queries where DeepSeek might be promoting messages on behalf of China. The three issues examined – Tiananmen, Xi, and Taiwan – are known explosive topics from China’s perspective, so its influence there was easy to find and identify.
However, Chinese propaganda may also be shaping DeepSeek’s responses in other, less immediately identifiable ways. For example, Beijing could use its influence on DeepSeek to plant supposedly objective messages condemning political moves that could harm it (e.g., Donald Trump’s plan to impose sweeping tariffs on imports from China to the United States), thereby attempting to influence global public opinion and shifting it in its favor.
Working with DeepSeek therefore requires caution in the answers it provides, as this revelation may also affect the U.S. government's response to the new chatbot, especially in light of the threat it represents to the domestic AI and chip industry.
