Virtual University Journals
Search
Search
Framing China: A Comparative Study of Geopolitical Bias in U.S. And Chinese Generative AI Systems
Shahid Hussain Soomro, Chaudhary Musawar Sharief Sandhu
Abstract:
The rapid global spread of large language models (LLMs), including Microsoft’s Copilot, OpenAI’s ChatGPT, Google’s Gemini, Anthropic’s Claude, and China’s Deep Seek, has challenged long-held assumptions about neutrality in knowledge production. Although these systems are often treated as objective information sources, their outputs reflect the political and ideological contexts in which they are developed. This paper examines five leading models, four Western and one Chinese, to assess how they frame sensitive topics related to China. Through qualitative content analysis, the study reviews their responses on leadership, human rights, state surveillance, and foreign policy. A clear divergence emerges: Western models tend to portray China as authoritarian and strategically confrontational, frequently raising concerns about rights and governance, while DeepSeek aligns with the official positions of the Chinese state, emphasizing stability, sovereignty, economic development, and security. Grounded in constructivist and Foucauldian perspectives, the study argues that LLMs shape global discourse and calls for more transparent and inclusive approaches to AI governance.
Keywords:
Large Language Models (LLMs), AI Governance, Geopolitical narratives, China–West relations, Algorithmic Bias
Full Text:
PDF