Vals AI finds wide variation in models’ China-related political alignment
Vals AI tested Chinese, American, and other language models with hundreds of questions about China, finding substantial variation in how closely responses aligned with official Chinese government positions. Prompt language, response format, tool access, and fine-tuning changed results, with Simplified Chinese generally producing more pro-government responses. The Decoder reported that some Chinese models adopted official positions rather than refusing, while Western models occasionally did the same.
The study challenges the assumption that Chinese models uniformly follow one political line, but its targeted questions, neutral scoring of refusals, and use of third-party providers limit how directly the results apply to official chatbots or everyday use.
Chinese AI models are politically complex, not just government mouthpieces, study finds
Nov 1, 2026
Nano Banana Pro prompted by THE DECODER
A study by Vals AI shows that Chinese AI models respond to political questions less uniformly than expected. Western models, too, occasionally adopt official Chinese government positions.
#Vals AI#DeepSeek#GLM-5.3 Flash#Kimi K3#MiMo V2.6 Pro#Qwen3.8 Max#Qwen3.6 35B#Thomson 1.0 Small#OpenAI#GPT Image 2.5 Sunburst#Thomson Reuters