Some argue Chinese AI models’ political bias can simply be retrained away since they’re open-weight. But even a stripped-down version of DeepSeek kept repeating Beijing’s line on Taiwan and Xinjiang in its Chinese-language answers.
The Global Danger of China’s State-Controlled AI ModelsThe spread of Chinese AI models with built-in censorship mechanisms could quietly shape the global information environment for the worse.www.worldpoliticsreview.com