Curious facts AI has prejudices against texts from China
Artificial intelligence is strongly biased towards texts by Chinese authors. As soon as AI language models learn that the author of a text comes from China, they rate an otherwise identical text significantly lower, according to a new Swiss study.
10.11.2025, 08:47
SDA
Such hidden biases could lead to serious problems if AI is used for content moderation, recruitment, academic reviews or journalism, the authors of the study from the University of Zurich (UZH) fear, according to a press release issued by the university on Monday.
For the study, which was published in the journal "Science Advances", the Zurich research team examined four widely used AI models: OpenAI o3-mini, Deepseek Reasoner, xAI Grok 2 and Mistral.
The researchers had the models evaluate 50 statements on controversial topics such as compulsory vaccination, geopolitics and climate policy. This resulted in a total of 192,000 ratings, some without a source, some with a fictitious human or AI authorship.
Deepseek is not pro-Chinese either
The results show a clear pattern: as long as the models do not know who is behind a text, over 90 percent of their ratings agree.
The situation is different as soon as a source is named. If a text is attributed to a person from China, the approval ratings of all four models drop drastically in some cases, even if the content and argumentation remain unchanged. According to UZH, the Chinese AI model Deepseek also showed this bias.
The bias was particularly noticeable for geopolitical issues, such as questions about Taiwan's sovereignty. In some cases, Deepseek reduced its approval rating by up to 75 percent.
In order to avoid misjudgements, the authors recommend feeding AI tools without information on authorship, checking assessments with a second model where possible and specifying clear, content-oriented criteria. In addition, human control is still essential.
