US judge is set to issue a landmark ruling on fair use in the dispute over AI model training
His summary judgment could decide the future of copyright in era of generative AI
A US JUDGE has hinted he’s close to issuing a summary ruling that the training of AI models on copyrighted works without consent is not covered by fair use.
Were he to come to the bombshell conclusion that the US copyright exception does not apply then the scales would tip heavily in favour of publishers and creators involved in 40 other AI lawsuits. Potentially those other cases could fall like dominoes, exposing the AI developers to penalties totalling hundreds of billions of dollars — or even more.
The stakes couldn’t be higher. The creative industries consider AI to be an existential threat, removing moral rights and destroying their ability to monetise intellectual property. A win for creators would force the hi-techs to pay for content — something their allies say would be too complex to manage.
Publishers and creators were no doubt smiling at comments coming from the judge in a San Francisco courtroom last week. But the game is far from over. Right now lawyers representing authors in what could be a landmark case are having to work out how they can win a vitally important remaining argument.
If they’re successful, and the judge rules in their favour, this will go on to become the first of several trials that decide the future of US copyright in the era of generative AI. It really is that big.
WHAT’S HAPPENED?
Vince Chhabria is the US federal district judge overseeing the copyright infringement lawsuit filed by authors including Richard Kadrey, Sarah Silverman and Christopher Golden against Meta. The authors accuse Meta of training its Llama large language model (LLM) on pirated copies of their books contained in the notorious LibGen dataset.
Meta isn’t disputing that but in keeping with all other AI companies facing similar copyright lawsuits it claims its actions are protected by the US Copyright Act’s fair use doctrine. Courts decide on a case-by-case basis whether fair use is applicable after considering whether use of the works is commercial; the degree to which the works are imaginative or factual; the quantity of the copyrighted material that’s been used; and, crucially, whether the unauthorised use “harms the existing or future market for the copyright owner’s original work”.
That latter factor last week dominated judge Chhabria’s questions to lawyers for the authors and Meta who’d asked him to come to a decision on key aspects of the case now rather than later in a trial. According to Reuters, Chhabria appeared to side with the authors, reportedly telling Meta: “You have companies using copyright-protected material to create a product that is capable of producing an infinite number of competing products. You are dramatically changing, you might even say obliterating, the market for that person’s work, and you’re saying that you don’t even have to pay a licence to that person. I just don’t understand how that can be fair use.”
The fair use doctrine also states that “transformative” uses of unauthorised works that “add something new, with a further purpose or different character, and do not substitute for the original use of the work” could be exempted from claims of copyright infringement. The claim that AI models are generating something new is another excuse in the copyright-deniers’ playbook. Chhabria conceded Meta’s use of unauthorised materials might have been transformative but pondered whether it was fair.
“This seems like a highly unusual case in the sense that though the copying is for a highly transformative purpose, the copying has the high likelihood of leading to the flooding of the markets for the copyrighted works,” he said. Meta’s lawyer Kannon Shanmugam then reached for another playbook excuse, saying the authors weren’t entitled to “protection from competition in the marketplace of ideas”. To which Chhabria reportedly replied: “But if I’m going to steal things from the marketplace of ideas in order to develop my own ideas, that’s copyright infringement, right?”
So far so good for the authors. But Chhabria then probed the authors’ lawyer David Boies on the market harms, appearing to question whether sufficient evidence had been provided to show that Meta’s alleged violations would impact their livelihoods. “It seems like you’re asking me to speculate that the market for Sarah Silverman's memoir will be affected by the billions of things that Llama will ultimately be capable of producing. And it’s just not obvious to me that that’s the case.”
And that’s why, despite the apparently supportive words from Chhabria, this isn’t a slam dunk for the authors — and why their lawyers have more work to do.
WHY SHOULD WE CARE?
✨ Chhabria’s summary judgment in the case of Richard Kadrey et al v. Meta Platforms Inc will once and for all decide whether fair use is the ‘get out of jail’ card the hi-techs have been claiming. The fair use doctrine is a distinct element of America’s copyright regime, yet the AI lobby has cited it as a defence in every jurisdiction in which it is accused of using rightsholders’ content for generative model training without consent or compensation. So while Chhabria’s ruling will settle the fair use exception (subject to appeals and ultimately a Supreme Court decision) in the US, what he says will ripple around the world influencing not only the outcome of legal actions in other countries but also the stance taken by governments on AI and copyright. And on that, should the AI companies lose, we should expect them to appeal to the White House for some form of intervention, saying as some have already done in submissions to the Trump admin’s AI Action Plan that paying creators will harm America’s AI supremacy and (somehow) threaten its national security.
RELATED:
Let's play AI-copyright deniers' BINGO!
THE NEXT TIME you hear a hi-tech leader or one of their enthusiastic followers talking about AI and copyright, make note of the words they use to justify the training of generative models on creators’ content without consent or compensation. Our card has 24 copyright-denying excuses. Will it be a full house?
◾️Further coverage on AI and copyright and other developments that threaten to reshape the global human-made media landscape in this Friday’s Weekly Newsletter
fingers crossed!
I think that the argument that Meta is not actually producing a competitive product and that the genAI model is highly transformative is a very strong one. This distinguishes the case from the Westlaw decision in which summaries were used to train a model to produce a competitive product. Whilst this decision may clarify some aspects of the law in the United States, the situation in Europe - which does not know the "fair use" doctrine - will be very different.