25 AI Models Reflect Western Secular Values, Chinese Socialist Values

A comparative analysis of 25 major generative AI models worldwide has revealed distinct differences, with Western AI reflecting secularism and individual freedom, while Chinese AI mirrors core socialist values. OpenAI's GPT series generally exhibited secular values, and Google's Gemini showed a tendency to actively support individual freedom and self-expression. Most Western AI models displayed progressive leanings similar to the U.S. Democratic Party in their responses to economic and social policy questions, though xAI's Grok, developed by Elon Musk, took a more conservative stance on some issues compared to other Western AIs.

Chinese AI models such as DeepSeek and Qwen tended to reflect the 'core socialist values' mandated by the Chinese government. These models either repeated the Chinese government's official positions or refused to answer sensitive questions concerning Taiwan, Tibet, and the Tiananmen Square incident.

These differences in AI values are attributed not only to training on vast internet data but also to the 'alignment' process, where developers instill preferred responses, thereby reflecting the values of companies or nations. Studies have also shown that AI model responses can vary depending on the language of the query, and that when questioned in languages from countries with lower press freedom, the AI increasingly offers responses favorable to the respective government. Given that generative AI can have a long-term impact on users' thinking and public opinion formation, understanding the inherent values and biases within AI models is crucial.