Generative AI has produced the most important copyright litigation in a generation, and the first merits rulings are now in. This UpLaw research article maps the whole field: the New York Times MDL, the Getty judgment in London, the twin June 2025 training-fair-use rulings in Bartz v. Anthropic and Kadrey v. Meta, the record $1.5 billion Anthropic settlement, the Thomson Reuters v. ROSS loss for a non-generative developer, and Germany's GEMA v. OpenAI decision holding that a model's memorization of lyrics is itself a reproduction. It separates the input question (does training infringe?) from the output question (did a generation copy a specific work?), walks the four fair-use factors as courts actually apply them after Warhol v. Goldsmith, and details the derivative-works, secondary-liability, and DMCA Section 1202 theories in play. It compares the U.S. transformative-use regime with the EU text-and-data-mining opt-out and the unsettled UK position, and closes with practical playbooks for content owners, AI developers, and their counsel. Written as a research resource, not legal advice, in a field that changes month to month.