AI Sovereignty, Data Bias, and Ethical Challenges in International Law

Abstract As artificial intelligence (AI) increasingly shapes global governance, the concept of AI sovereignty raises significant legal and ethical concerns. A central challenge lies in systemic bias within dominant AI models, which are predominantly trained on Western data, languages, and cultural perspectives. Such bias influences content generation, legal interpretation, and historical representation, potentially marginalising non-Western knowledge systems. This paper examines the implications of biased training data for AI sovereignty in Asian legal and geopolitical contexts, focusing on linguistic inequality, restricted access to regional datasets, and AI-driven content moderation. It analyses how these factors may generate information asymmetries and constrain national regulatory autonomy. By comparing Western-led governance frameworks with alternative state-centric regulatory models, the study evaluates whether localised AI development or international regulatory cooperation better addresses algorithmic bias, digital censorship, and human rights protections in an increasingly fragmented global AI order.