Two Days-Two AI Decisions-One Big Mess

Stephen Carlisle July 11, 2025

A few weeks ago, two days apart, two judges (who occupy the same courthouse) issued summary judgements in favor of AI developers, ruling that the scraping and copying of the Plaintiff’s material was fair use. These cases were Bartz v. Anthropic ¹ and Kadrey v. Meta Platforms Inc. ² While the fair use finding garnered all the headlines, both Judges hedged their rulings in significant ways, making the outcome of future AI cases very murky indeed. Adding to the uncertainty, is that much of the rulings, particularly the Anthropic decision, rested on very shaky legal reasoning, and set up an inevitable circuit split with the Second Circuit.

The judge in the Anthropic case, in a bold display of ipse dixit, proclaims the scraping and copying by Anthropic was “quintessentially transformative.” ³ Even though the output from the AI program was never an issue, the Court rules that the mere format change from print library to digital library copies was “transformative.” ⁴ This is despite the fact, by the Court’s own admission, that “the format was changed but no content was added or subtracted.” ⁵

Wait a second. By the Supreme Court’s own ruling that gave the US the concept of transformative use, it posed the question “[Does the new work add] something new, with a further purpose or different character, altering the first with new expression, meaning, or message?” ⁶ So how in the world, does an AI program, which changes nothing about the work other than the format, “add new expression, meaning or message”? The Court never explains this. And despite including Campbell in a string cite for a procedural point, never really discusses the holding in Campbell at all.

And while a District Court in the Ninth Circuit is not obliged to follow the rulings of the Second Circuit, this decision flies in the face of a number of rulings from that Circuit including the recently decided case of Hachette Book Group v. Internet Archive, ⁷ where the mere acquisition of a copy of a book did not translate into a right to make a digital copy of the book, and then distribute it.

The judge in the Meta Platforms case fares no better. It simply rules that “There is no serious question that Meta’s use of the plaintiff’s books had a ‘further purpose’ and ‘different character’ than the books-that it was highly transformative.” ⁸ OK, but again, where is the “new expression, meaning or message”? None that I can see.

Interestingly enough, and almost unheard of, the Judge in Meta takes the time to make a pointed swipe at the reasoning of the Judge in Anthropic.

“[In] a recent ruling on this topic, Judge Alsup focused heavily on the transformative nature of generative AI while brushing aside concerns about the harm it can inflict on the market for the works it gets trained on. Such harm would be no different, he reasoned, than the harm caused by using the works for ‘training schoolchildren to write well,’ which could ‘result in an explosion of competing works.’ (citation omitted) According to Judge Alsup, this ‘is not the kind of competitive or creative displacement that concerns the Copyright Act.’ But when it comes to market effects, using books to teach children to write is not remotely like using books to create a product that a single individual could employ to generate countless competing works with a miniscule fraction of the time and creativity it would otherwise take. This inapt analogy is not a basis for blowing off the most important factor in the fair use analysis.” ⁹

And here is the point that makes Meta not quite a slam dunk victory that many make it out to be. The market effect of diluting, or perhaps eliminating a market, is a potential harm that would overcome the supposedly “transformative” nature of the use and render the use not a fair use.

“As for the potentially winning argument—that Meta has copied their works to create a product that will likely flood the market with similar works, causing market dilution—the plaintiffs barely give this issue lip service, and they present no evidence about how the current or expected outputs from Meta’s models would dilute the market for their own works.” ¹⁰

And then this:

“And, as should now be clear, this ruling does not stand for the proposition that Meta’s use of copyrighted materials to train its language models is lawful. It stands only for the proposition that these plaintiffs made the wrong arguments and failed to develop a record in support of the right one.” ¹¹

The Judge in Meta is really drawing a road map for future plaintiffs to follow: the negative market effect, market dilution, and market elimination, is a winning argument that will negate whatever favor being a “transformative” use might bring to the table. As the Court in Meta states:

“Of course, not all copyrighted works would have their markets diluted equally by AI-generated competitors. It seems unlikely, for instance, that AI-generated books would meaningfully siphon sales away from well-known authors who sell books to people looking for books by those particular authors. But it’s easy to imagine that AI-generated books could successfully crowd out lesser-known works or works by up-and-coming authors. While AI-generated books probably wouldn’t have much of an effect on the market for the works of Agatha Christie, they could very well prevent the next Agatha Christie from getting noticed or selling enough books to keep writing.” ¹²

And finally:

“In cases involving uses like Meta’s, it seems like the plaintiffs will often win, at least where those cases have better-developed records on the market effects of the defendant’s use. No matter how transformative LLM training may be, it’s hard to imagine that it can be fair use to use copyrighted books to develop a tool to make billions or trillions of dollars while enabling the creation of a potentially endless stream of competing works that could significantly harm the market for those books.” ¹³

Meanwhile, back at the Anthropic Court, it drops the bombshell that the transformative ruling only applies to copies which were lawfully acquired, not those that were acquired from “pirate” databases.

“From the start, Anthropic ‘ha[d] many places from which’ it could have purchased books, but it preferred to steal them to avoid ‘legal/practice/business slog,’ as cofounder and chief executive officer Dario Amodei put it.” ¹⁴

The Court rules:

“[T]he test requires that we contemplate the likely result were the conduct to be condoned as a fair use — namely to steal a work you could otherwise buy (a book, millions of books) so long as you at least loosely intend to make further copies for a purportedly transformative use… without any accountability. As Anthropic itself suggested, ‘That would destroy the [entire] publishing market if that were the case.’” ¹⁵

“The downloaded pirated copies used to build a central library were not justified by a fair use. Every factor points against fair use… A separate justification was required for each use. None is even offered here except for Anthropic’s pocketbook and convenience.” ¹⁶

That’s a mighty big “oops” on the part of the AI companies. Because they all train on pirated works. Anthropic admitted to knowingly downloading almost 8 million pirated copies of books. If the court awarded $1,000 in statutory damages for every work infringed, that means Anthropic would owe $8 billion in damages. But the upper limit on regular statutory damages is not $1,000 per work, it’s $30,000. ¹⁷

Unfortunately, my calculator does not go that high.

So, the fair use fight over AI is far from over, and all indications are that this will wind up back at the Supreme Court.