On October 23, 2025, the "Artificial Intelligence Race: Capitalist Dystopia and Decolonization Resistance" seminar series, hosted online by the Institute for Humanitas Unisinos (Instituto Humanitas Unisinos) in Brazil, was held. Xiong Jie, Secretary-General of the Global South Academic Forum at East China Normal University, was invited to attend and delivered a keynote speech titled "Bias in Large Language Models (LLMs) and the Importance of Digital Sovereignty." After the event, he accepted an exclusive interview, where he elaborated on the challenges posed by artificial intelligence development to the national security of the "Global South" nations. (Interview link: IA: “Viés centrado no Ocidente representa ameaça direta à segurança cultural e nacional da China e do Sul Global”. Entrevista especial com Jeff Xiong）
In his speech, Xiong Jie pointed out that Artificial Intelligence systems, represented by Large Language Models, due to the deeply embedded "Western-centric bias" in their training and design processes, already pose a direct and profound threat to the cultural security and even national security of "Global South" nations, including China. He emphasized that this ideological bias is not an isolated incident but a systemic problem that runs through the entire lifecycle of model design, development, training, and application. Xiong Jie's team's quantitative research found that even China's domestically developed LLMs, due to their reliance on mainstream English content, the high weighting of Western knowledge sources (such as Wikipedia), and the influence of Western cultural datasets during the post-training phase, generally exhibit a moderate to high pro-Western ideological inclination.
Regarding Digital Sovereignty, Xiong Jie analyzed the structural dilemma faced by "Global South" nations, namely their high dependence on US-dominated digital infrastructure. Taking China as an example, he emphasized that China, through seven decades of sustained investment and construction, has been the first among the BRICS countries to achieve relative digital sovereign independence, an experience that holds significant reference value for developing countries.
In response to the current dilemma, Xiong Jie proposed multi-level strategies and recommendations. He advocated that Chinese enterprises and "Global South" nations should jointly explore the construction of new digital infrastructure to effectively reduce reliance on Western technology. Concurrently, he suggested that countries should develop and explore digital governance models that combine China's experience with their respective national realities, ensuring that digital space governance adheres to the core principle of respecting national sovereignty. On the technical level, he introduced his team's LLM bias identification and correction mechanism based on Marxist philosophy, emphasizing the possibility of effective intervention at the model level through methods like prompt adjustments and knowledge enhancement.
Xiong Jie concluded by summarizing that the world is currently at a special juncture where the old order is giving way to the new, and Western hegemony is being questioned and challenged by non-Western cultures. He urged all nations to fully recognize the inherent ideological risks within Artificial Intelligence systems, treat them as a core national security issue, and jointly advance the global digital governance system toward a direction that is more equitable and respects sovereignty.