Precedent Set: Court Rules Anthropic Can Train AI on Copyrighted Books Without Author Approval

Federal Judge William Alsup has ruled that Anthropic is permitted to train its models on published books without seeking permission from the authors.

This marks the first instance in which a court has affirmed an AI company’s claim that the doctrine of fair use may shield it from liability when utilizing copyrighted materials in the training of large language models.

In the past, a number of authors, artists, and publishers have filed dozens of lawsuits against AI developers. While Judge Alsup’s decision does not ensure that other courts will adopt a similar stance, it could set a precedent benefiting tech companies.

The outcomes hinge on interpretations of the fair use doctrine—a principle that allows for the use of copyrighted works without permission from the rights holder under certain circumstances.

In the case against Anthropic, a group of plaintiffs accused the company of establishing a «central library of all the world’s books,» claiming that these works were downloaded illegally from pirated websites.

The judge stated that the training of the neural networks with the materials was conducted fairly and that he would further examine the nature of this «central library.»

«We will conduct a trial regarding the pirated copies used to create Anthropic’s central library and the damages associated with this. The fact that the company later purchased a copy of a book that it previously stole from the internet does not absolve it from liability for theft but may influence the calculation of statutory damages,» Judge Alsup noted in his ruling.

Additionally, it is worth mentioning that in June, the AI image generation startup Midjourney became another company accused of violating intellectual property rights.