There’s a common belief that DeepSeek, a type of artificial intelligence (AI), only limits what information it shares when you’re using it through an app. Some people think that if you download its AI model and run it on your own computer, those limitations will disappear.
However, a detailed investigation by Wired has found that DeepSeek’s censorship is actually part of its design, and it affects both how the app operates and how it was trained from the start.
For instance, when Wired examined a version of DeepSeek that was running locally—meaning it was directly on someone’s personal computer—they discovered that the AI was designed to avoid discussing certain historical events, like China’s Cultural Revolution. Instead, the AI focused on presenting a more positive view of the Chinese Communist Party.
Read Also: Meta May Halt AI Development It Finds Too Risky
Similarly, TechCrunch did a quick test of another locally run version of DeepSeek and found obvious signs of censorship too. When they asked the AI about the Kent State shootings in Ohio, it provided an answer. But when they inquired about the Tiananmen Square protests that occurred in 1989, the AI simply responded, “I cannot answer.”
This shows that even when DeepSeek is run on a personal computer, it still maintains strict controls on the information it shares, reflecting its built-in censorship.
