AI engines known as large language models (LLMs) ingest huge numbers of copyrighted works to train the engines, and operate them in response to user queries. The ingestion and use have generally been done without permission from copyright owners, sometimes through mass ingestion of works from pirate sites. Numerous lawsuits have been filed against LLMs, and so far most are at early stages. Significant legal issues remain unresolved. Should it matter if LLMs copy from pirate sites? Do LLMs infringe if copyrighted material is used solely for training purposes, and no copyright-protected material is embodied in output? What kind of damages do copyright owners incur, and what types of damages are legally cognizable? Should there be licensing mechanisms to protect copyright owners from use of their works by LLMs?
Speakers: Regan Smith, News Media Alliance and Sy Damle, Latham & Watkins
