Judge Allows Authors’ AI Copyright Case Against Databricks to Move Forward

A federal judge in Northern California has ruled that a group of authors can proceed with their copyright lawsuit against Databricks, marking another development in the growing wave of legal challenges over how AI models are trained.

The case involves writers Stewart O’Nan, Abdi Nazemian, Brian Keene, Rebecca Makkai, and Jason Reynolds, who allege that their works were used without permission to train large language models. According to the complaint, MosaicML—an AI training company owned by Databricks—relied on the Books3 dataset to help build its own model, MPT, as well as Databricks’ DBRX system.

In a recent decision, U.S. District Judge Charles B. Dreyer declined to dismiss the lawsuit, determining that the authors’ revised claims are detailed enough to move forward at this stage. While the court did not rule on the merits of the case, the judge noted that the allegations presented are sufficient to continue litigation for now, leaving open the possibility that the defendants could still prevail later.

This marks a shift from an earlier phase of the case. A previous version of the authors’ complaint had been dismissed, with the court finding that it lacked the specificity required to proceed. The updated filing appears to have addressed those concerns, allowing the case to advance.

The lawsuit is part of a broader legal battle unfolding across the tech and publishing industries, as authors and creators challenge the use of copyrighted material in training artificial intelligence systems. As courts continue to weigh these claims, decisions like this one could help shape how intellectual property is treated in the age of AI.