A recent legal decision has reaffirmed the power of fair use in the digital age, and it’s a big win for libraries and the future of public access to knowledge.
On June 24, 2025, Judge William Alsup of the United States District Court for the Northern District of California ruled in favor of Anthropic, finding that the company’s use of purchased copyrighted books to train its AI model qualified as fair use. While the case centered on emerging AI technologies, the implications of the ruling reach much further—especially for institutions like libraries that depend on fair use to preserve and provide access to information.
What the Decision Says
In the case, publishers claimed that Anthropic infringed copyright by including copyrighted books in its AI training dataset. Some of those books were acquired in physical form and then digitized by Anthropic to make them usable for machine learning.
The court sided with Anthropic on this point, holding that the company’s “format-change from print library copies to digital library copies was transformative under fair use factor one” and therefore constituted fair use. It also ruled that using those digitized copies to train an AI model was a transformative use, again qualifying as fair use under U.S. law.
This part of the ruling strongly echoes previous landmark decisions, especially Authors Guild v. Google, which upheld the legality of digitizing books for search and analysis. The court explicitly cited the Google Books case as supporting precedent.
While we believe the ruling is headed in the right direction—recognizing both format shifting and transformative use—the court factored in destruction of the original physical books as part of the digitization process, a limitation we believe could be harmful if broadly applied to libraries and archives.
What It Means for Libraries
Libraries rely on fair use every day. Whether it’s digitizing books, archiving websites, or preserving at-risk digital content, fair use enables libraries to fulfill our public service missions in the digital age: making knowledge available, searchable, and accessible for current and future generations.
This decision reinforces the idea that copying for non-commercial, transformative purposes—like making a book searchable, training an AI, or preserving web pages—can be lawful under fair use. That legal protection is essential to modern librarianship.
In fact, the court’s analysis strengthens the legal groundwork that libraries have relied on for years. As with the Google Books decision, it affirms that digitization for research, discovery, and technological advancement can align with copyright law, not violate it.
Looking Ahead
This ruling is an important step forward for libraries. It reaffirms that fair use continues to adapt alongside new technologies, and that the law can recognize public interest in access, preservation, and innovation.
As we navigate a rapidly changing technological landscape, it’s more important than ever to defend fair use and support the institutions that bring knowledge to the public. Libraries are essential infrastructure for an informed society, and legal precedents like this help ensure they can continue their vital work in the digital age.
Let’s hope this means that the Internet Archive books that got removed from borrowing by non-print-disabled individuals get restored for borrowing by regular members again.
What could be the implications of this ruling for Archive.org? Does this mean that we will see the return of some of the books that are lost to us now on Archive.org?
Archives and libraries were already safe before this ruling, and most likely would’ve been just fine if it had gone in a different direction. There’s a difference between the archival of copyrighted material that the Internet Archive does, and Anthropic using said copyrighted content to train AI models for commercial usage. While the first case is for the public good, the latter is the end result of corporate greed and legal double-standards allowed to run free at the cost of workers’ rights. This case is setting a dangerous precedent that could have very negative ramifications for the future of workers’ livelihoods and rights in many industries (including most media industries) in the near future if this decision isn’t appealed against.
It’s honestly disappointing to see the representatives of IA take this short-sighted stance in support of Anthropic. This isn’t a win for fair use, it’s a loss for workers who have been fighting hard for years against tech companies taking their work without compensation or permission.
Clara TB, I could not agree more. I was astounded at the myopic decision of the judge in the Anthropic case, and I think Internet Archive is being similarly myopic in seeing this decision as a net positive. I was really taken aback when I read Chris Freeland’s response to this ruling.
Judge Alsup compared Anthropic’s Claude to aspiring human writers: “If someone were to read all the modern-day classics because of their exceptional expression, memorize them, and then emulate a blend of their best writing, would that violate the Copyright Act? Of course not. … Like any reader aspiring to be a writer, Anthropic’s LLMs trained upon works not to race ahead and replicate or supplant them—but to turn a hard corner and create something different.” Judge Alsup remains willfully ignorant of the fact that an aspiring writer will labor for months or years to produce a single book-length work, while Claude could spit them out by the dozen without breaking a digital sweat, and then Anthropic could publish and sell them to benefit its own profit margin (not Claude’s).
Even more shocking to me was Judge Alsup’s comparison of the Anthropic LLM to schoolchildren: “Authors contend generically that training LLMs will result in an explosion of works competing with their works … This order assumes that is so. But Authors’ complaint is no different than it would be if they complained that training schoolchildren to write well would result in an explosion of competing works.” Really? A bunch of children who learned to write well (and who must grow up and decide to write professionally, and then do the years of labor it takes to produce even just one book-length work from which they are unlikely to see any profit at all) are just as threatening to authors as an LLM deployed by a commercial entity to churn out competing books and win on volume of sales?
Please, Chris Freeland, think again.
What’s the dangerous precedent? A dangerous precedent would have been the court deciding that anyone (human or machine) isn’t allowed to learn from the books we buy!
Nobody’s “taking” anyone’s work, nor is anyone entitled to additional compensation. Someone is USING the work, in this case to learn. I have a large collection of handicapping books – do you think it’s illegal for me to learn horse race handicapping from reading those books? Would it be illegal to incorporate what I’ve learned into a software program?
This decision is not only a victory for fair use; it’s a victory for common sense.
Funny how back in Sept 2024 in your case, the ruling was is a derivative use rather than a transformative one?
I don’t understand what makes this different, perhaps a future blog could explain? Thanks
Archival is going to be useless if we can’t segregate plagiarized AI slop from genuine works, which is basically an inevitability the more permissive the law becomes on genAI. As a monthly supporter this is not the kind of thing that makes me feel like continuing to support the archive. Please correct your heads on this topic! There is nothing transformative or preservative about LLM based AI tech.
How is it not transformative? You start with a book, you end up with a machine that can pass the Turing test. This is as transformative as the lightning bolt in Mary Shelley’s Frankenstein!
AI doesn’t plagiarize anything; it learns from it just as you do.
Lots of people keep insinuating that all LLMs do is predict the next word or parrot back plagiarized text. To these people I say… some companies, including Meta, are now offering MILLION DOLLAR salaries to top LLM experts! Why don’t you apply, get hired at a million dollar salary, and write an algorithm that just spits back memorized texts or statistically predicts the next word? No? Don’t think you’ll get passed the interview process? Then please stop denigrating the world-changing accomplishments of LLM technology and technologists. The AI revolution will be as world-transforming as the industrial revolution or the dawn of the information age. Nothing transformative? Just human civilization!
Looking at what the ‘disruptors’ have “achieved” I for one am cynical about the value of A.I. The world is a long way from a Star Trek type utopian unity and ethical society, aren’t we? In that canon, they had to have a Third World (nuclear) war before things finally settled down, too. Sometimes change isn’t good. It should merit itself first. The problem with ‘Capitalism’ is that it massively motivates profit and unethical behaviour is rewarded. Specifically, unethical behaviour is presumed to be punished by the law, but it is corrupt not Common Law for the people with equity. So, without the law performing its duty anywhere near properly, who steps-in? Proper checks and balances (these days) are missing. It is getting worse and the rich are getting richer, faster. (inequality is growing).
Arguably due to the automation of production robbing the people of their bargaining power which kept a balance on centralised power. So, nothing stops centralised power manipulating perception to create profit, over truth. Privatized Big Brother, anyone? This has caused so much harm alreadyL Edward Bernays and the Nazis 90+ years ago, with basic psychological manipulation via traditional media (the PR Industry)… through Brexit through Russian Social Media propagandists etc. History teaches us that A LOT can go wrong. More fool us for BOTH ignoring history’s lessons AND/OR ignoring the potential for technology to be used for good – but that appears to REQUIRE CHECKS AND BALANCES which I would argue are SOUNDLY PROVEN TO BE ABSENT FROM MODERN LAW AND CAPITALISM.
A new era of Laissez-Faire? Oh great, will we have another Irish Famine, then? There is no underpopulated America crying out for cheap labour to escape to this time, remember? So the free market freedom of movement migration ‘positivity’ is going to go where this time? What is to stop these unethical people genociding the spare population and using the landspace for something they want? I believe it’s only a matter of time. Ethics and checks and balances are decidedly NOT going to prevent it. Slow genocide – performed by stealth is included in that logic. I am not vaccinated as I could not make an informed decision with the best will in the world, and being coerced into it suggests something wrong. Remember: Coercive control is a concept applied to errant husbands to keep them down (sometimes rightly, sometimes it’s a misandric lie). When huge powers that be perform coercive control, it’s somehow a virtue. This incongruency is why people’s mental health is so bad these days: massive stress caused by the cognitive dissonance caused by constant lying coercive manipulation and genuine gaslighting as crucial components of many unethical business models.
I’m sorry to say you’ve been thoroughly brainwashed by tech bros, Joseph. But you too will eventually come to understand that you’ve bought into a big lie, at least assuming you’re a human capable of learning to the degree that you mistakenly believe any “AI” implementations are doing.
“BIotechnology, ain’t what’s so bad… like ALL technology, it’s in the WRONG HANDS.” Centralised power = AI (despite protests to the contrary). Why are people ignoring this, coming on top of the pandemic Great Reset. The WEF official site calls it The Great Reset, and the evidence on the ground points to it being somewhat contrived if not full-on pandemic-on-purpose – not some spaced-out conspiracy theory – same as trading patterns and the missing 1 trillion from the Pentagon budget immediately pre-9/11 2001 show something odd.
Centralised control is the key and its consolidation and the rich getting FAR richer is what has happened. So Pro-Bono Internet Archive vs private AI by selfish profiteering companies (generalisation but the trend is clear, be real). They are not the same. The enemy of my enemy is not always (long-term at least) my friend.
I’ll have to join the comments that are criticizing the thrust of this blog post, whose author appears to have fallen prey to simplistic “the enemy of my enemy is my friend” thinking. The people who will stand to benefit from this particular ruling if it’s not successfully appealed are *not* going to do things with it that are in line with IA’s mission statement, quite the opposite. Try to support a legal framework that doesn’t endanger IA’s long-term utility for its short-term survival.