Authors, illustrators push for copyright owner class in case against Google AI

SAN JOSE, Calif. (CN) —  A group of authors and illustrators requested to become part of a class Friday in an action against Google to stop using their copyrighted works to train Google’s generative artificial intelligence product “Gemini.”

U.S. District Judge Eumi Lee was forthright with questions to both parties, with eight named plaintiffs arguing a motion to certify a class that would include anyone owning a copyright, including publishers, whose works Google used to train its AI, amounting to potentially millions of claims.

“For an operative class definition, I need to know what you are asking me to certify, I need to understand,” Lee said.

The plaintiffs’ attorney Lesley Weaver said it would be owners of copyrights who knowingly or unknowingly uploaded work to Google, allowing Google, without their consent, to use those “training data sets” for its generative AI models by crawling its webpages.

The plaintiffs seek injunctive relief, saying Google needs to be transparent when it tries to use copyrighted work, and ask for owner’s consent before use.

“How do we know what works are covered?” Google attorney David Kramer asked, adding the case would be impossible to adjudicate on a class-wide basis. He said Google didn’t keep lists on individual works of what Gemini is trained on.

Weaver said that Google not knowing or keeping records on what data sets its AI is training on “doesn’t make sense.”  Weaver said the plaintiffs want additional discovery from Google to identify the training sets and allow it as evidence.

“It’s how they are making a better cake,” she said. “They are keeping track of what is in the recipe.”

Kramer claimed that Google was allowed to develop and improve existing services using uploaded work through Google’s licenses. He admitted the licenses included sweeping language and could be interpreted as broad but claims of potential copyright infringement or fair use could be litigated individually, instead of trying to find everything that was used by AI that possibly was copyrighted.

Lee, a Joe Biden appointee, said that when reading Google’s language about using uploaded work “set me on my heels.”

But Kramer said that the language was clear, and it wasn’t necessary to ask for consent.

“The plaintiffs haven’t shown why we need to do this,” he said of the possible class action.

Kramer also poked holes some of the individual plaintiffs' claims, saying one, Sarah Andersen, uploaded her illustrations to Google while also having a license with Mattel for a children’s book, and that she published her work on Twitter before she registered for a copyright.

Kramer also said another plaintiff, Hope Larson, did not have a copyright of the cover image of her graphic novel “Chiggers” and her rights were not therefore infringed. Google also said Larson mischaracterized her ownership of another of her titles, saying while she was deposed, she admitted it was a “work for hire” and she did not own the copyright.

During the beginning of the four-hour hearing, Jeff Gould, representing two publishers, argued to intervene, saying the publishers have a larger stake and unique reproduction rights in the case, and could help with expert analysis.

Publishing companies Cengage Group and Hachette Book Group seek to join the putative class action, but Lee questioned why the publishers waited so long to join the case, which started in 2023.

“We were awaiting a transition point in the case,” Gould said.

Google said that it would be prejudicial to add the publishers to possible class with the case so far advanced, including the possibility that with the publishers inclusion, discovery would have to be reopened.

Plaintiff attorney Joseph Saveri said there was already adequate representation of copyright owners through the authors, and to include the publishers may delay proceedings.

Lee also heard brief arguments on a motion for sanctions brought by Google, which accuses the plaintiffs of expanding the scope of the class without meeting and conferring with defendants — something that was a part of Lee’s standing orders in the case. Weaver said in her thirty years of practice she never had to meet with the other side when making class definitions.

“I was troubled, I will be frank, by the failure to meet and confer,” Lee said. “Even if you see it as futile is not a reason to follow a judge’s order.”

Lee did not make any rulings on the motions to certify class, for sanctions, relief or to intervene from the bench.