Chinese AI models are politically complex, not just government mouthpieces, study finds

| Source: THE DECODER

Tags: Vals AI, GLM-5.3 Flash, model alignment, Simplified Chinese, AI censorship, benchmarks

A Vals AI study of 25 models found Chinese AI models less uniformly pro-Beijing than expected, Western models sometimes adopting official Chinese positions, and prompts in Simplified Chinese shifting answers pro-government.

Details

Vals AI asked Chinese, American and other language models hundreds of questions about China, varying language, usage context and response format. It scored a "pro-China lean" from minus 1 to plus 1, measuring alignment with official government positions rather than factual accuracy. Overall, Chinese models leaned less pro-government than expected, with large differences between models. GLM-5.3 Flash, for example, confirmed that China's human rights violations are well documented. Some non-Chinese models occasionally aligned with official Chinese positions. Prompt language had a strong effect: for most models, answers in Simplified Chinese were more pro-government than in English, Russian or Spanish. The effect appeared in both Chinese and American models but was much stronger in Chinese ones. For teams deploying models across languages, the result suggests political behavior should be tested per language.