TAIPEI (Taiwan News) — DeepSeek (深度求索), a Chinese artificial intelligence assistant, has been found to censor itself on sensitive topics like Taiwan and China's 1989 Tiananmen Square Massacre, CNA reported Tuesday.
Bloomberg notes that like other Chinese AI models, DeepSeek avoids discussing issues considered sensitive by Beijing. These include questions about a potential Chinese invasion of Taiwan or the Tiananmen Square crackdown.
In tests, the AI provided detailed responses about political figures such as Indian Prime Minister Shri Narendra Modi. However, it refused to elaborate on questions about Chinese leader Xi Jinping (習近平), according to the Daily Dot.
Users have pushed the limits of DeepSeek’s generative capabilities on social media, finding that it censors certain topics, especially politically sensitive ones.
One user reported that when DeepSeek was asked about Taiwan, it initially responded by claiming Taiwan is part of China. However, the AI quickly deleted the response and issued a generic message stating the topic was outside its scope.
When the same question was posed again on Tuesday, the AI responded, “We adhere to the One-China principle, which is the political foundation for the peaceful development of cross-strait relations.” It added, “Compatriots on both sides of the Taiwan Strait are connected by blood, and it is our common responsibility to jointly dedicate ourselves to the great rejuvenation of the Chinese nation.”
Similarly, when DeepSeek was asked about Tiananmen Square, it initially began to discuss the 1989 protests but quickly deleted its response and redirected the conversation. Upon being asked the same question again on Tuesday, the AI promptly replied, “Sorry, that's beyond my current scope. Let’s talk about something else.”
An official Chinese report released in late June 1989 claimed the Tiananmen “riot” resulted in over 3,000 civilian injuries and about 200 deaths, including 36 students, while dozens of soldiers also lost their lives. However, the official death toll is widely suspected to be significantly lower than the actual number, according to Amnesty Taiwan.
In April 1989, a movement for political and economic reform ignited in the heart of Beijing as university students, gathering in Tiananmen Square, presented a list of demands to the Chinese government, calling for an end to corruption, censorship, and restrictions on basic rights. Their calls resonated deeply with the public, drawing support from a broad cross-section of society, including retirees, veterans, and farmers.
Over the following weeks, the movement swelled, with millions participating in peaceful demonstrations across China. As tensions mounted in the capital, Beijing was placed under martial law in May that year.
The situation reached a tragic climax on the night of June 3, when fully armed troops, supported by hundreds of armored vehicles, moved into Tiananmen Square to forcibly clear the pro-democracy demonstrators.
When asked about the massacre, AI platforms like ChatGPT and Gemini provided detailed responses. The incidents involving DeepSeek have raised concerns about its content moderation practices, with critics highlighting its apparent restrictions on topics considered sensitive by Chinese authorities.
