CeoTelechaje 
Anthropic has confidentially settled a massive class-action lawsuit alleging it pirated millions of books to train its AI, Claude. This avoids a trial that could have resulted in catastrophic financial liability, potentially reaching hundreds of billions of dollars.
The case established a crucial legal distinction: while the judge ruled that training an AI model on copyrighted works is likely “fair use” due to its transformative nature, he also ruled that illegally obtaining and storing the data in a central library is not protected. This left Anthropic exposed to immense liability for its data sourcing methods, not for the training itself.
The settlement, coming after a class of up to seven million copyright holders was certified, is a major event. It signals that future legal battles will focus less on whether AI training is transformative and more on how companies acquire their training data. The outcome serves as a warning to AI firms that while training may be legal, using pirated or improperly sourced data is a high-risk praice.
