The News/Media Alliance (N/MA) expressed strong support for the U.S. Copyright Office’s newly released pre-publication report
on Generative AI Training, calling it a critical step in clarifying how copyright law applies to the use of protected content in training artificial intelligence models.
This report, the third in the Copyright Office’s AI-focused series, analyzes whether the use of copyrighted material -- particularly by large language models (LLMs) and retrieval-augmented generation (RAG) systems -- constitutes infringement. While acknowledging the complexity and evolving legal landscape, the Office concluded that AI training involves multiple acts that could infringe copyright if not covered by a license or a legitimate legal defense.