Copyright Law and AI Training in the US and EU

This article explores how copyright laws in the United States and European Union currently address the practice of training artificial intelligence (AI) models using copyrighted materials. By examining recent controversies involving major AI platforms such as ChatGPT, Meta's LLaMA, and Google's Gemini, we analyze the legal principles of fair use and transformative use in the U.S., alongside the EU’s specific text and data mining exceptions. The paper contrasts AI model training to traditional human learning processes, highlighting conceptual inconsistencies in the application of copyright laws. We critically assess whether existing legal frameworks adequately accommodate the realities of AI-driven technological advancements.