Meta secures partial dismissal of Silverman AI suit
A federal judge is set to partially dismiss copyright lawsuit brought forward by authors | Allegations of copyright infringement related to its language model | Judge argues AI output is not a copy of the authors' works.
Meta Platforms has succeeded in fending off some of the copyright infringement claims levelled against it by comedian Sarah Silverman and two novelists over its artificial intelligence model, Llama.
A federal judge in the US District Court for the Northern District of California announced yesterday that he plans to partially dismiss the lawsuit filed against Meta over the AI’s use of language in its training, according to Reuters.
Meta moved to dismiss the lawsuit in September, asserting that the claims brought by the trio lack legal merit and conflict with legal precedents.
The complaint against Meta centred on claims made by Silverman and authors Richard Kadrey and Christopher Golden, who alleged that their copyrighted books were used to train the AI without their consent.
US District Judge Vince Chhabria argued in favour of Meta’s arguments yesterday, and is set to trim parts of the lawsuit.
Case background
In February 2023, Meta introduced its foundational language model, Llama—an AI software designed to generate coherent text responses to user prompts and queries.
The core issue in this legal dispute centres on whether the training process of Llama, which involved allegedly copying massive amounts of text from various sources, including the authors’ books, infringes copyright laws.
The authors did not claim that the text generated by Llama is similar to their books: instead, they objected to Meta’s failure to obtain permission to use information from their texts during the training process.
Meta countered that copyright law does not protect facts or the syntactical, structural, and linguistic information extracted during the training process.
It claimed that using texts to train LLMs is transformative and constitutes ‘fair use’, referencing the precedent set in the case Authors Guild v Google (2015).
‘Fundamentally flawed’ argument
The authors also alleged direct copyright infringement, arguing Llama is itself an infringing ‘derivative’ work.
However, Meta argued that the authors had failed to establish substantial similarities between the AI and the books in question.
Additionally, the authors sought to hold Meta liable for infringing outputs generated by Llama users, but the tech giant pointed out that the plaintiffs could not identify any infringing outputs and that their argument was fundamentally flawed.
In handing down his decision Chhabria criticised the nature of the plaintiffs’ suit.
“I understand your core theory…your remaining theories of liability I don't understand even a little bit,” he told them.
He also criticised the plaintiffs’ second claim asserting that the AI’s generated text resembles their works.
“When I make a query of Llama, I'm not asking for a copy of Sarah Silverman's book–I'm not even asking for an excerpt,” he added.
Chhabria also found that the claim that Llama itself is an infringing work is invalid because it would imply that “if you put the Llama language model next to Sarah Silverman's book, you would say they're similar”.
“That makes my head explode when I try to understand that,” continued Chhabria.
The judge said that he would initially dismiss most of the claims but would allow the authors to modify them. He also indicated that he would dismiss the claims once more if the authors failed to demonstrate that Llama's output closely resembled their works.
Did you enjoy reading this story? Sign up to our free daily newsletters and get stories sent like this straight to your inbox
