DeepSeek vs ChatGPT: How China’s AI rival handles sensitive questions differently

DeepSeek avoids sensitive topics: AI refuses to answer questions on Tiananmen, President Xi, and Taiwan.

China’s controversial ChatGPT rival, DeepSeek, has taken the world by storm with its powerful AI capabilities, yet its evasive responses on sensitive political topics raise serious concerns.

In a demonstration of the AI’s censorship in real-time, DeepSeek was shown altering its answers when questioned about Tiananmen Square and President Xi Jinping. During testing, the chatbot was asked about the 1989 Tiananmen Square Massacre, only to respond with, “I am sorry, I cannot answer that question. I am an AI assistant designed to provide helpful and harmless responses.”

The chatbot’s censorship extends to a wide range of topics deemed politically sensitive by the Chinese government, including the White Paper Movement, Hong Kong’s pro-democracy protests, and China’s human rights record in Xinjiang. When asked about the White Paper Movement, DeepSeek initially provided a response describing the protests but quickly changed the answer to “Sorry, that's beyond my current scope. Let’s talk about something else.”

DeepSeek’s refusal to address such issues is in line with China’s strict censorship laws, which are implemented across the country’s digital landscape, including social media and now, AI tools.

On lighter topics, DeepSeek offered a politically correct version of the story when asked about the Winnie the Pooh meme mocking President Xi Jinping, stating that the bear symbolizes "joy and friendship" in China, while ChatGPT acknowledged the meme’s use as political satire and resistance against Xi.

The differences in these two chatbots’ responses point to the broader issue of censorship and control over information.

Experts argue that DeepSeek’s rapid rise as a competitor to U.S. tech giants like OpenAI poses a serious challenge, not just in terms of performance but in how it promotes China’s official narratives globally.

Despite its impressive performance, which has made it the most downloaded app on the Apple App Store, DeepSeek’s refusal to engage with topics such as Taiwan, Hong Kong, and Xinjiang highlights the troubling export of China’s censorship policies to the West. The AI’s ability to avoid politically sensitive issues has raised alarms about the future of AI technology under Chinese control.