The article discusses the emerging controversy over AI companies using library collections to train their language models. Major AI companies like Anthropic and Meta are accessing digital library collections through the Internet Archive and other sources, raising concerns about copyright infringement and fair use. The National Writers Union and other authors’ groups are particularly concerned about AI systems learning from their works without permission or compensation. The Internet Archive defends its position, stating that AI training falls under fair use and that they implement technical measures to prevent misuse. Libraries are caught in a complex situation, trying to balance their mission of providing public access to information while protecting authors’ rights. Some institutions, like the New York Public Library, are developing their own policies regarding AI access to their digital collections. The debate highlights broader questions about intellectual property in the AI era, with some experts arguing that using copyrighted materials for AI training should require permission and compensation, while others maintain it constitutes transformative fair use. The article also notes that some AI companies are now seeking licensing agreements with publishers and content creators, suggesting a potential shift toward more formal arrangements for accessing training materials.