US Judge Rules AI Training on Purchased Books Fair Use, Pirated Books Infringement

A U.S. federal judge, William Alsup, ruled that the AI company Anthropic's use of legally purchased books to train its large language models, like Claude, constitutes fair use, comparing it to how a human reads and learns from texts. This decision, stemming from a class-action lawsuit by authors alleging unauthorized use of their works, dismissed claims regarding training on purchased books but found that Anthropic's use of pirated books for AI training infringed copyright laws, potentially incurring financial penalties. Judge Alsup emphasized that the AI's transformative use fosters creativity and scientific progress without supplanting original works, and no infringing copies were distributed to users. The ruling highlights the complex legal landscape for AI-generated content and the need to distinguish between authorized learning and unauthorized data acquisition. Meanwhile, Indian copyright law lacks clear guidelines on AI-generated art, raising questions about authorship and originality when AI learns from existing copyrighted materials. Together, these developments underscore ongoing debates about copyright protections, fair use, and AI's role in creative industries worldwide.