The AI company won another victory in court this week. Wednesday's meta is Kadreyv. It won a motion seeking a partial summary judgment in favor of Meta. This was raised by 13 authors, raised by 13 authors who infringed copyright protections by using copyright to train the Lama AI model. This ruling comes two days after a similar victory for Claude Maker humanity.
However, Judge Vince Chhabria emphasized in his order that the ruling should be limited and that it would not exempt meta from future claims from other authors.
“The ruling does not represent the proposition that Meta uses copyrighted material to train language models,” he wrote. “It only represents the proposition that these plaintiffs had made the wrong argument and were unable to develop a record in favor of the right thing.”
The central question of the case is whether AI companies are eligible for fair use for AI training-protected content. Fair Use Doctrines are a fundamental part of US copyright law that allows people to use copyrighted work without the express permission of the rights holder, like education and journalism. There are four important considerations when assessing whether something is used fairly. Human control focused on transformationality, while Meta focused on the impact that AI use had on existing publishing markets.
These verdicts are a big win for AI companies. Openai, Google and others are fighting for fair use, so the frustration of content creators is that they don't have to enter costs and long-term license agreements with content creators. A group of well-known authors signed an open letter on Friday, urging publishers to take a stronger stance on AI and not use it.
“The AI providers also stole our work from us and our publishers,” the letter reads. The author calls out how AI is trained in his work without permission and compensation, but the program never can connect with humans like a real person. For authors who file these cases, they may see some victory in subsequent copyright infringement trials (for mankind) or new cases. However, there are concerns about the overall impact of AI on current and future writers.
(Disclosure: CNET's parent company Ziff Davis filed a lawsuit against Openai in April, claiming it infringed Ziff Davis's copyright in training and operating AI systems.)
In his analysis, Chhabria focused on the impact of AI-generated books on existing publishing markets, which he regarded as the four most important factors needed to prove fair use. He writes extensively about the risk that generative AI and large-scale language models can violate copyright laws and that fair use must be evaluated on a case-by-case basis. Some works, such as autobiography and classical literature such as Rye Catcher, were probably unable to be created with AI, he writes. However, he said, “The market for typical human-created romance or spy novels can be significantly reduced by a surge in similar AI-created works.”
In other words, AI Slops can make human-written books less valuable and undercut the author's motivation and creative abilities.
Still, Chhabria said the plaintiffs did not provide sufficient evidence to prove harm from “the model of meta diluting the market for their work.” The plaintiffs focused discussions on how Meta's AI models can replicate accurate snippets from their works, and how the company's Llama models undermine the ability to license books to AI companies. These arguments were not persuasive in Chhabria's eyes – he called them “clear losers” so he sided with Meta.
This differs from the human ruling, when Judge William Alsap focused on the “very transformative” nature of the plaintiff's book use that the AI chatbot spews out. Chhabria wrote that the use of copyrighted materials is “no controversy” but that the impact of AI systems on the entire ecosystem is a more urgent issue.
Alsup outlined concerns about the way in which humans obtain books through illegal online libraries and then deliberated on purchasing printed copies for the digitization of the “research library.”
The two court decisions do not legalize the use of content under fair use for not all AI companies. What is noteworthy about these cases is that they first published a substantive legal analysis of the issue. AI companies and publishers have been taking it in court for many years.
However, as Chhabria responded with reference to the judgment of mankind, all judges use past cases with similar circumstances as the point of reference. They don't have to reach the same conclusion, but the role of precedent is important. These two judgments may be referenced in other AI and copyright/copyright infringement cases.
However, we must wait for how much of these rulings will have in future cases, and whether it is warnings or green light that will keep the most weight on future decisions.
For more information, see the Copyright and AI Guide.
