A San Francisco federal judge certified an author class against Anthropic over stored pirated books; Midjourney separately moved to dismiss Disney and Universal's infringement claims.
A federal judge in San Francisco granted class certification to a group of authors suing Anthropic, the AI company, over its alleged storage of pirated books used to train its Claude models [1]. In a separate proceeding in the same district, defendants in a copyright infringement suit brought by Disney and Universal moved to dismiss the claims against them, targeting an AI image-generation platform [1]. Both developments mark inflection points in the rapidly expanding body of AI copyright litigation concentrated in the Northern District of California.
The Anthropic class action is before Judge William Alsup of the U.S. District Court for the Northern District of California [1]. Alsup had previously ruled that using copyrighted works to train AI models may qualify as fair use, but that storing pirated copies of those works does not, a distinction that created direct exposure for Anthropic's data-retention practices [2]. The class certification order means plaintiffs can now pursue claims on behalf of a defined group of authors whose works were allegedly stored without authorization, a procedural step that dramatically expands the potential damages pool. The Midjourney motion to dismiss, filed in a separate matter involving Disney and Universal as plaintiffs, asks the court to apply the emerging fair-use framework to AI image generators rather than text-based models [1].
The certification ruling carries substantial practical weight. Class actions that survive this threshold impose settlement pressure disproportionate to individual claims, and AI companies storing large corpora of training data now face aggregate liability calculations that individual-plaintiff suits would not generate. The Midjourney motion raises a distinct but related question: whether fair-use reasoning developed in text-model litigation translates to image-generation systems, which operate on different architectures and produce outputs in a different medium [1]. Judge Vince Chhabria, who presided over related proceedings in Kadrey v. Meta and granted Meta a partial dismissal, has also contributed to the analytical framework that Midjourney's counsel will likely invoke [2].
The next procedural steps will determine litigation posture for the broader industry. In the Anthropic matter, the parties will move toward merits discovery on a class-wide basis, with damages experts likely engaged in the near term. In the Midjourney case, the court must rule on the motion to dismiss before any class or merits issues are reached [1]. How Alsup and the assigned Midjourney judge resolve these motions will shape litigation strategy for every AI company that maintains training datasets drawn from commercially published works.