Feeding the Algorithm

How Nations Shape AI Training Data to Project Power and Influence Global News Narratives

Abstract

Large language models (LLMs) and generative artificial intelligence (AI) systems are emerging as key intermediaries in the global circulation of news and political knowledge. Yet the datasets that feed these systems are neither neutral nor universal. Governments, regulatory bodies, and corporate ecosystems are now treating training data as a strategic resource—one that can encode national priorities, cultural hierarchies, and geopolitical narratives into the algorithmic infrastructures of communication. This conceptual essay develops the notion of algorithmic diplomacy, arguing that nations are actively curating, regulating, and disseminating datasets to project soft power through AI. Drawing on theories of epistemic sovereignty, digital colonialism, and media framing, the paper identifies three main mechanisms: (1) data curation and localization, (2) model fine-tuning and regulatory alignment, and (3) narrative seeding through open-source ecosystems. Comparative illustrations from the European Union, the United States, China, Russia, and the Gulf states reveal how “feeding the algorithm” has become a new instrument of influence within the global information order. The essay concludes that algorithmic infrastructures now constitute a form of epistemic territory—a contested space where values, identities, and political legitimacy are negotiated through data rather than discourse.