Game Theory and Inter-Company Data Sharing for AI Model Training

This paper examines how game theory can be used to model and design data-sharing strategies between firms developing artificial intelligence. It explores both competitive and cooperative settings, analyzing how incentives, market structure, privacy concerns, and platform design affect data-sharing outcomes. Two core models are discussed: a non-cooperative game between competing firms and a cooperative model involving Shapley value allocation. The study highlights when data sharing can be stable, fair, and beneficial for innovation while identifying potential risks to competition and consumer welfare.