In Re Google Generative AI Copyright Litigation addresses a class action of authors and visual artists against Google for direct copyright infringement, and its parent company, Alphabet, for vicarious infringement. This particular case stands out because the plaintiffs claim that copyright infringement occurred when the rightsholders’ works were copied into generative AI models, not that they appeared in the models’ output. The plaintiffs also sought to enjoin Google and Alphabet from further copyright infringement.
Ten litigants with infringement claims stipulated to have their cases consolidated in the Northern District of California, which has become a clearing house for AI-related copyright claims against Google. The litigants sought to represent the creators of millions of registered works that were allegedly used to train over a dozen of Google’s generative AI platforms without authorization. Each of the 10 litigants was able to identify specific works and demonstrate that they had been included in datasets used for training large language models and image generators.
Google responded by bringing a motion to dismiss, claiming that the copied works of one plaintiff were not the subject of a valid copyright registration. Plaintiff Andersen’s registration covered only unpublished works, which Google attacked as a reason to throw out the claim. Andersen had previously posted parts of her work on the social media platform X (formerly known as Twitter), which Google claimed voided the registration for the entire collection of purportedly “unpublished” works.
The court declined to take judicial notice of the social media posts, pointing out that such posts had been demonstrated to be easily faked. It went on to say that, even if Andersen posted some of her works on X, there was no reason to infer that other materials in the collection hadn’t remained unpublished.
Google succeeded elsewhere in its defense by arguing that plaintiffs did not sufficiently identify the specific AI platforms that were alleged to have been trained on their works. Rather, plaintiffs listed some 16 of Google’s generative AI models and offered a paragraph explaining the way in which all generative AI models are trained. Relying on a list of only possible instances of infringement did not pass the muster of the court, which found the attempt overly broad absent any allegation as to how infringement actually occurred.
Plaintiffs’ claim of vicarious infringement against Alphabet likewise did not go over well. Courts have consistently ruled that in a typical parent-subsidiary relationship, copyright infringement claims arising from the conduct of the subsidiary cannot be extended to the parent company. The court said it did not see anything about the relationship between Google and Alphabet that stood out from any other parent and subsidiary — even though the parent and subsidiary shared some of the same board members, and Alphabet allegedly had been “involved in” Google’s developing its AI tools.
The court found that, at this motion to dismiss stage, the plaintiffs had plausibly alleged they were entitled to injunctive relief against future copyright infringement. However, the court granted the motion to dismiss for the claims of direct copyright infringement against some of Google’s AI tools and vicarious copyright infringement against Alphabet. Unfortunately for plaintiffs, they were not given leave to amend the complaint.
In denying leave to amend, the judge noted that this case had a long, developed record spanning over two years that had included numerous opportunities to amend. Despite multi-stage meet-and-confer sessions taking place, plaintiffs had failed to cure their pleading deficiencies. Given the number of opportunities plaintiffs had been presented to amend, the court found that granting further leave would have been futile.
So, the judge narrowed the pleadings, ostensibly to focus the parties’ efforts in this massive and complex wave of litigation. Clearly, any judge considering the legal intricacies of a new, viral technology has their work cut out for them. Which explains the court’s desire to streamline the pleadings and “get on with it.”
