Federal Judge Rules AI Companies Can Use Copyrighted Books for Training

In a major development for the AI industry, a federal judge has ruled that it is legal for companies to use copyrighted books to train their language models without obtaining permission from the authors.

The ruling came in the case of Bartz v. Anthropic, where a group of authors sued the AI company Anthropic for using their copyrighted works to train its large language model (LLM). Judge William Alsup sided with Anthropic, stating that the company's use of the books falls under the "fair use" doctrine of copyright law.

This decision marks the first time a court has explicitly given credence to the AI industry's argument that training on copyrighted materials constitutes fair use. It could have significant ramifications for the many other lawsuits brought by authors, artists, and publishers against companies like OpenAI, Meta, Midjourney, and Google for their use of copyrighted works in AI training.

The fair use doctrine is a legal exception to copyright infringement that allows limited use of copyrighted material without permission, such as for the purposes of commentary, criticism, news reporting, teaching, or research. In this case, Judge Alsup ruled that Anthropic's use of the books to train its AI model fell under the "transformative" aspect of fair use.

"While the ruling is not a guarantee that other judges will follow Judge Alsup's lead, it lays the foundation for courts to side with tech companies over creatives," wrote TechCrunch reporter Amanda Silberling.

However, the judge did note that Anthropic may still be liable for damages related to the way it obtained the books, as the company allegedly downloaded millions of copyrighted books from pirate websites. The court will hold a trial to determine the extent of Anthropic's liability for these "pirated copies."

The decision comes as a major blow to the authors, artists, and publishers who have filed numerous lawsuits against AI companies in recent years. These plaintiffs have argued that the use of their copyrighted works to train AI models without permission or compensation is a violation of their rights.

But the AI industry has pushed back, contending that training on large datasets of text and images, even if copyrighted, is essential for developing advanced language and image generation models that provide significant public benefit. They've claimed this use constitutes fair use.

With this ruling, the courts have now lent credence to that fair use argument - at least in the case of Anthropic. It remains to be seen whether other judges will follow suit, or if this decision will be appealed.

Regardless, the Bartz v. Anthropic case highlights the growing tension between the rights of creators and the needs of the AI industry. As generative AI models become more sophisticated and ubiquitous, the legal battles over their training data are likely to continue.

This latest court ruling represents a major win for AI companies, but the fight over copyrights and fair use in the age of AI is far from over. Creatives will undoubtedly continue to push back, seeking to protect their intellectual property rights. The ultimate resolution of this issue could have sweeping implications for the future of AI development.