It took me a few days to get the time to read the actual court ruling but here’s the basics of what it ruled (and what it didn’t rule on):
- It’s legal to scan physical books you already own and keep a digital library of those scanned books, even if the copyright holder didn’t give permission. And even if you bought the books used, for very cheap, in bulk.
- It’s legal to keep all the book data in an internal database for use within the company, as a central library of works accessible only within the company.
- It’s legal to prepare those digital copies for potential use as training material for LLMs, including recognizing the text, performing cleanup on scanning/recognition errors, categorizing and cataloguing them to make editorial decisions on which works to include in which training sets, tokenizing them for the actual LLM technology, etc. This remains legal even for the copies that are excluded from training for whatever reason, as the entire bulk process may involve text that ends up not being used, but the process itself is fair use.
- It’s legal to use that book text to create large language models that power services that are commercially sold to the public, as long as there are safeguards that prevent the LLMs from publishing large portions of a single copyrighted work without the copyright holder’s permission.
- It’s illegal to download unauthorized copies of copyrighted books from the internet, without the copyright holder’s permission.
Here’s what it didn’t rule on:
- Is it legal to distribute large chunks of copyrighted text through one of these LLMs, such as when a user asks a chatbot to recite an entire copyrighted work that is in its training set? (The opinion suggests that it probably isn’t legal, and relies heavily on the dividing line of how Google Books does it, by scanning and analyzing an entire copyrighted work but blocking users from retrieving more than a few snippets from those works).
- Is it legal to give anyone outside the company access to the digitized central library assembled by the company from printed copies?
- Is it legal to crawl publicly available digital data to build a library from text already digitized by someone else? (The answer may matter depending on whether there is an authorized method for obtaining that data, or whether the copyright holder refuses to license that copying).
So it’s a pretty important ruling, in my opinion. It’s a clear green light to the idea of digitizing and archiving copyrighted works without the copyright holder’s permission, as long as you first own a legal copy in the first place. And it’s a green light to using copyrighted works for training AI models, as long as you compiled that database of copyrighted works in a legal way.
FTA:
Anthropic warned against “[t]he prospect of ruinous statutory damages—$150,000 times 5 million books”: that would mean $750 billion.
So part of their argument is actually that they stole so much that it would be impossible for them/anyone to pay restitution, therefore we should just let them off the hook.
The problem isnt anthropic get to use that defense, its that others dont. The fact the the world is in a place where people can be fined 5+ years of a western European average salary for making a copy of one (1) book that does not materially effect the copyright holder in any way is insane and it is good to point that out no matter who does it.
Funny how that kind of thing only works for rich people
Ah the old “owe $100 and the bank owns you; owe $100,000,000 and you own the bank” defense.
This version of too big to fail is too big a criminal to pay the fines.
How about we lock them up instead? All of em.
In April, Anthropic filed its opposition to the class certification motion, arguing that a copyright class relating to 5 million books is not manageable and that the questions are too distinct to be resolved in a class action.
I also like this one too. We stole so much content that you can’t sue us. Naming too many pieces means it can’t be a class action lawsuit.
Ahh cant wait for hedgefunds and the such to use this defense next.
Hold my beer.
Lawsuits are multifaceted. This statement isn’t a a defense or an argument for innocence, it’s just what it says - an assertion that the proposed damages are unreasonably high. If the court agrees, the plaintiff can always propose a lower damage claim that the court thinks is reasonable.
You’re right, each of the 5 million books’ authors should agree to less payment for their work, to make the poor criminals feel better.
If I steal $100 from a thousand people and spend it all on hookers and blow, do I get out of paying that back because I don’t have the funds? Should the victims agree to get $20 back instead because that’s more within my budget?
You think that 150,000 dollars, or roughly 180 weeks of full time pretax wages at 15$ an hour, is a reasonable fine for making a copy of one book which doe no material harm to the copyright holder?
No I don’t, but we’re not talking about a single copy of one book, and it is grovellingly insidious to imply that we are.
We are talking about a company taking the work of an author, of thousands of authors, and using it as the backbone of a machine that’s goal is to make those authors obsolete.
When the people who own the slop-machine are making millions of dollars off the back of stolen works, they can very much afford to pay those authors. If you can’t afford to run your business without STEALING, then your business is a pile of flaming shit that deserves to fail.
Except it isnt, because the judge dismissed that part of the suit, saying that people have complete right to digitise and train on works they have a legitimate copy of. So those damages are for making the unauthorised copy, per book.
And it is not STEALING as you put it, it is making an unauthorised copy, no one loses anything from a copy being made, if I STEAL your phone you no longer have that phone. I do find it sad how many people have drunk the capitalist IP maximalist stance and have somehow convinced themselves that advocating for Disney and the publishing cartel being allowed to dictate how people use works they have is somehow sticking up for the little guy
None of the above. Every professional in the world, including me, owes our careers to looking at examples of other people’s work and incorporating their work into our own work without paying a penny for it. Freely copying and imitating what we see around us has been a human norm for thousands of years - in a process known as “the spread of civilization”. Relatively recently it was demonized - for purely business reasons, not moral ones - by people who got rich selling copies of other people’s work and paying them a pittance known as a “royalty”. That little piece of bait on the hook has convinced a lot of people to put a black hat on behavior that had been considered normal forever. If angry modern enlightened justice warriors want to treat a business concept like a moral principle and get all sweaty about it, that’s fine with me, but I’m more of a traditionalist in that area.
Nobody who is mad at this situation thinks that taking inspiration, riffing on, or referencing other people’s work is the problem when a human being does it. When a person writes, there is intention behind it.
The issue is when a business, owned by those people you think ‘demonised’ inspiration, take the works of authors and mulch them into something they lovingly named “The Pile”, in order to create derivative slop off the backs of creatives.
When you, as a “professional”, ask AI to write you a novel, who is being inspired? Who is making the connections between themes? Who is carefully crafting the text to pay loving reference to another authors work? Not you. Not the algorithm that is guessing what word to shit out next based on math.
These businesses have tricked you into thinking that what they are doing is noble.
That’s 100% rationalization. Machines have never done anything with “inspiration”, and that’s never been a problem until now. You probably don’t insist that your food be hand-carried to you from a farm, or cooked over a fire you started by rubbing two sticks together. I think the mass reaction against AI is part of a larger pattern where people want to believe they’re crusading against evil without putting out the kind of effort it takes to fight any of the genuine evils in the world.
What is means is they don’t own the models. They are the commons of humanity, they are merely temporary custodians. The nightnare ending is the elites keeping the most capable and competent models for themselves as private play things. That must not be allowed to happen under any circumstances. Sue openai, anthropic and the other enclosers, sue them for trying to take their ball and go home. Disposses them and sue the investors for their corrupt influence on research.
I think this means we can make a torrent client with a built in function that uses 0.1% of 1 CPU core to train an ML model on anything you download. You can download anything legally with it then. 👌
…no?
That’s exactly what the ruling prohibits - it’s fair use to train AI models on any copies of books that you legally acquired, but never when those books were illegally acquired, as was the case with the books that Anthropic used in their training here.
This satirical torrent client would be violating the laws just as much as one without any slow training built in.
But if one person buys a book, trains an “AI model” to recite it, then distributes that model we good?
No. The court made its ruling with the explicit understanding that the software was configured not to recite more than a few snippets from any copyrighted work, and would never produce an entire copyrighted work (or even a significant portion of a copyrighted work) in its output.
And the judge specifically reserved that question, saying if the authors could develop evidence that it was possible for a user to retrieve significant copyrighted material out of the LLM, they’d have a different case and would be able to sue under those facts.
I don’t think anyone would consider complete verbatim recitement of the material to be anything but a copyright violation, being the exact same thing that you produce.
Fair use requires the derivative work to be transformative, and no transformation occurs when you verbatim recite something.
“Recite the complete works of Shakespeare but replace every thirteenth thou with this”
I’d be impressed with any model that succeeds with that, but assuming one does, the complete works of Shakespeare are not copyright protected - they have fallen into the public domain since a very long time ago.
For any works still under copyright protection, it would probably be a case of a trial to determine whether a certain work is transformative enough to be considered fair use. I’d imagine that this would not clear that bar.
A court will decide such cases. Most AI models aren’t trained for this purpose of whitewashing content even if some people would imply that’s all they do, but if you decided to actually train a model for this explicit purpose you would most likely not get away with it if someone dragged you in front of a court for it.
It’s a similar defense that some file hosting websites had against hosting and distributing copyrighted content (Eg. MEGA), but in such cases it was very clear to what their real goals were (especially in court), and at the same time it did not kill all file sharing websites, because not all of them were built with the intention to distribute illegal material with under the guise of legitimate operation.
And this is how you know that the American legal system should not be trusted.
Mind you I am not saying this an easy case, it’s not. But the framing that piracy is wrong but ML training for profit is not wrong is clearly based on oligarch interests and demands.
This is an easy case. Using published works to train AI without paying for the right to do so is piracy. The judge making this determination is an idiot.
You’re right. When you’re doing it for commercial gain, it’s not fair use anymore. It’s really not that complicated.
If you’re using the minimum amount, in a transformative way that doesn’t compete with the original copyrighted source, then it’s still fair use even if it’s commercial. (This is not saying that’s what LLM are doing)
The judge making this determination is an idiot.
The judge hasn’t ruled on the piracy question yet. The only thing that the judge has ruled on is, if you legally own a copy of a book, then you can use it for a variety of purposes, including training an AI.
“But they didn’t own the books!”
Right. That’s the part that’s still going to trial.
The order seems to say that the trained LLM and the commercial Claude product are not linked, which supports the decision. But I’m not sure how he came to that conclusion. I’m going to have to read the full order when I have time.
This might be appealed, but I doubt it’ll be taken up by SCOTUS until there are conflicting federal court rulings.
If you are struggling for time, just put the opinion into chat GPT and ask for a summary. it will save you tonnes of time.
Judges: not learning a goddamned thing about computers in 40 years.
You’re poor? Fuck you you have to pay to breathe.
Millionaire? Whatever you want daddy uwu
That’s kind of how I read it too.
But as a side effect it means you’re still allowed to photograph your own books at home as a private citizen if you own them.
Prepare to never legally own another piece of media in your life. 😄
Gist:
What’s new: The Northern District of California has granted a summary judgment for Anthropic that the training use of the copyrighted books and the print-to-digital format change were both “fair use” (full order below box). However, the court also found that the pirated library copies that Anthropic collected could not be deemed as training copies, and therefore, the use of this material was not “fair”. The court also announced that it will have a trial on the pirated copies and any resulting damages, adding:
“That Anthropic later bought a copy of a book it earlier stole off the internet will not absolve it of liability for the theft but it may affect the extent of statutory damages.”
So I can’t use any of these works because it’s plagiarism but AI can?
My interpretation was that AI companies can train on material they are licensed to use, but the courts have deemed that Anthropic pirated this material as they were not licensed to use it.
In other words, if Anthropic bought the physical or digital books, it would be fine so long as their AI couldn’t spit it out verbatim, but they didn’t even do that, i.e. the AI crawler pirated the book.
Does buying the book give you license to digitise it?
Does owning a digital copy of the book give you license to convert it into another format and copy it into a database?
Definitions of “Ownership” can be very different.
Does buying the book give you license to digitise it?
Does owning a digital copy of the book give you license to convert it into another format and copy it into a database?
Yes. That’s what the court ruled here. If you legally obtain a printed copy of a book you are free to digitize it or archive it for yourself. And you’re allowed to keep that digital copy, analyze and index it and search it, in your personal library.
Anthropic’s practice of buying physical books, removing the bindings, scanning the pages, and digitizing the content while destroying the physical book was found to be legal, so long as Anthropic didn’t distribute that library outside of its own company.
It seems like a lot of people misunderstand copyright so let’s be clear: the answer is yes. You can absolutely digitize your books. You can rip your movies and store them on a home server and run them through compression algorithms.
Copyright exists to prevent others from redistributing your work so as long as you’re doing all of that for personal use, the copyright owner has no say over what you do with it.
You even have some degree of latitude to create and distribute transformative works with a violation only occurring when you distribute something pretty damn close to a copy of the original. Some perfectly legal examples: create a word cloud of a book, analyze the tone of news article to help you trade stocks, produce an image containing the most prominent color in every frame of a movie, or create a search index of the words found on all websites on the internet.
You can absolutely do the same kinds of things an AI does with a work as a human.
You can digitize the books you own. You do not need a license for that. And of course you could put that digital format into a database. As databases are explicit exceptions from copyright law. If you want to go to the extreme: delete first copy. Then you have only in the database. However: AIs/LLMs are not based on data bases. But on neural networks. The original data gets lost when “learned”.
If you want to go to the extreme: delete first copy.
You can; as I understand it, the only legal requirement is that you only use one copy at a time.
ie. I can give my book to a friend after I’m done reading it; I can make a copy of a book and keep them at home and at the office and switch off between reading them; I’m not allowed to make a copy of the book hand one to a friend and then both of us read it at the same time.
That sounds a lot like library ebook renting. Makes sense to me. Ty
That’s not what it says.
Neither you nor an AI is allowed to take a book without authorization; that includes downloading and stealing it. That has nothing to do with plagiarism; it’s just theft.
Assuming that the book has been legally obtained, both you and an AI are allowed to read that book, learn from it, and use the knowledge you obtained.
Both you and the AI need to follow existing copyright laws and licensing when it comes to redistributing that work.
“Plagiarism” is the act of claiming someone else’s work as your own and it’s orthogonal to the use of AI. If you ask either a human or an AI to produce an essay on the philosophy surrounding suicide, you’re fairly likely to include some Shakespeare quotes. It’s only plagiarism if you or the AI fail to provide attribution.
Why would it be plagiarism if you use the knowledge you gain from a book?
Formatting thing: if you start a line in a new paragraph with four spaces, it assumes that you want to display the text as a code and won’t line break.
This means that the last part of your comment is a long line that people need to scroll to see. If you remove one of the spaces, or you remove the empty line between it and the previous paragraph, it’ll look like a normal comment
With an empty line of space:
1 space - and a little bit of writing just to see how the text will wrap. I don’t really have anything that I want to put here, but I need to put enough here to make it long enough to wrap around. This is likely enough.
2 spaces - and a little bit of writing just to see how the text will wrap. I don’t really have anything that I want to put here, but I need to put enough here to make it long enough to wrap around. This is likely enough.
3 spaces - and a little bit of writing just to see how the text will wrap. I don’t really have anything that I want to put here, but I need to put enough here to make it long enough to wrap around. This is likely enough.
4 spaces - and a little bit of writing just to see how the text will wrap. I don't really have anything that I want to put here, but I need to put enough here to make it long enough to wrap around. This is likely enough.Thanks, I had copy-pasted it from the website :)
Personally I prefer to explicitly wrap the text in backticks.
Three ` symbols will
Have the same effectBut the behavior is more clear to the author
It’s extremely frustrating to read this comment thread because it’s obvious that so many of you didn’t actually read the article, or even half-skim the article, or even attempted to even comprehend the title of the article for more than a second.
For shame.
Nobody ever reads articles, everybody likes to get angry at headlines, which they wrongly interpret the way it best tickles their rage.
Regarding the ruling, I agree with you that it’s a good thing, in my opinion it makes a lot of sense to allow fair use in this case
It seems the subject of AI causes lemmites to lose all their braincells.
“While the copies used to convert purchased print library copies into digital library copies were slightly disfavored by the second factor (nature of the work), the court still found “on balance” that it was a fair use because the purchased print copy was destroyed and its digital replacement was not redistributed.”
So you find this to be valid? To me it is absolutely being redistributed
Ok so you can buy books scan them or ebooks and use for AI training but you can’t just download priated books from internet to train AI. Did I understood that correctly ?
Make an AI that is trained on the books.
Tell it to tell you a story for one of the books.
Read the story without paying for it.
The law says this is ok now, right?
The law says this is ok now, right?
No.
The judge accepted the fact that Anthropic prevents users from obtaining the underlying copyrighted text through interaction with its LLM, and that there are safeguards in the software that prevent a user from being able to get an entire copyrighted work out of that LLM. It discusses the Google Books arrangement, where the books are scanned in the entirety, but where a user searching in Google Books can’t actually retrieve more than a few snippets from any given book.
Anthropic get to keep the copy of the entire book. It doesn’t get to transmit the contents of that book to someone else, even through the LLM service.
The judge also explicitly stated that if the authors can put together evidence that it is possible for a user to retrieve their entire copyrighted work out of the LLM, they’d have a different case and could sue over it at that time.
Sort of.
If you violated laws in obtaining the book (eg stole or downloaded it without permission) it’s illegal and you’ve already violated the law, no matter what you do after that.
If you obtain the book legally you can do whatever you want with that book, by the first sale doctrine. If you want to redistribute the book, you need the proper license. You don’t need any licensing to create a derivative work. That work has to be “sufficiently transformed” in order to pass.
As long as they don’t use exactly the same words in the book, yeah, as I understand it.
How they don’t use same words as in the book ? That’s not how LLM works. They use exactly same words if the probabilities align. It’s proved by this study. https://arxiv.org/abs/2505.12546
I’d say there are two issues with it.
FIrst, it’s a very new article with only 3 citations. The authors seem like serious researchers but the paper itself is still in the, “hot off the presses” stage and wouldn’t qualify as “proven” yet.
It also doesn’t exactly say that books are copies. It says that in some models, it’s possible to extract some portions of some texts. They cite “1984” and “Harry Potter” as two books that can be extracted almost entirely, under some circumstances. They also find that, in general, extraction rates are below 1%.
Yeah but it’s just a start to reverse the process and prove that there is no AI. We only started with generating text I bet people figure out how to reverse process by using some sort of Rosetta Stone. It’s just probabilities after all.
That’s possible but it’s not what the authors found.
They spend a fair amount of the conclusion emphasizing how exploratory and ambiguous their findings are. The researchers themselves are very careful to point out that this is not a smoking gun.
Yeah authors rely on the recent deep mind paper https://aclanthology.org/2025.naacl-long.469.pdf ( they even cite it ) that describes (n, p)-discoverable extraction. This is recent studies because right now there are no boundaries, basically people made something and now they study their creation. We’re probably years from something like gdpr for llm.
The LLM is not repeating the same book. The owner of the LLM has the exact same rights to do with what his LLM is reading, as you have to do with what ever YOU are reading.
As long as it is not a verbatim recitation, it is completely okay.
According to story telling theory: there are only roughly 15 different story types anyway.
That’s my understanding too. If you obtained them legally, you can use them the same way anyone else who obtained them legally could use them.
Judge,I’m pirating them to train ai not to consume for my own personal use.
Yeah I have a bash one liner AI model that ingests your media and spits out a 99.9999999% accurate replica through the power of changing the filename.
cpOut performs the latest and greatest AI models
I call this legally distinct, this is legal advice.
mvwill save you some disk space.Unless you’re moving across partitions it will change the filesystem metadata to move the path, but not actually do anything to the data. Sorry, you failed, it’s jail for you.
stupid inodes preventing me from burning though my drive life
This ruling stated that corporations are not allowed to pirate books to use them in training. Please read the headlines more carefully, and read the article.
But, corporations are allowed to buy books normally and use them in training.
Please read the comment more carefully. The observation is that one can proliferate a (legally-attained) work without running afoul of copyright law if one can successfully argue that
cpconstitutes AI.
Check out my new site TheAIBay, you search for content and an LLM that was trained on reproducing it gives it to you, a small hash check is used to validate accuracy. It is now legal.
The court’s ruling explicitly depended on the fact that Anthropic does not allow users to retrieve significant chunks of copyrighted text. It used the entire copyrighted work to train the weights of the LLMs, but is configured not to actually copy those works out to the public user. The ruling says that if the copyright holders later develop evidence that it is possible to retrieve entire copyrighted works, or significant portions of a work, then they will have the right sue over those facts.
But the facts before the court were that Anthropic’s LLMs have safeguards against distributing copies of identifiable copyrighted works to its users.
Does it “generate” a 1:1 copy?
You can train an LLM to generate 1:1 copies
I am training my model on these 100,000 movies your honor.
This ruling stated that corporations are not allowed to pirate books to use them in training. Please read the headlines more carefully, and read the article.
thank you Captain Funsucker!
Unpopular opinion but I don’t see how it could have been different.
- There’s no way the west would give AI lead to China which has no desire or framework to ever accept this.
- Believe it or not but transformers are actually learning by current definitions and not regurgitating a direct copy. It’s transformative work - it’s even in the name.
- This is actually good as it prevents market moat for super rich corporations only which could afford the expensive training datasets.
This is an absolute win for everyone involved other than copyright hoarders and mega corporations.
I’d encourage everyone upset at this read over some of the EFF posts from actual IP lawyers on this topic like this one:
Nor is pro-monopoly regulation through copyright likely to provide any meaningful economic support for vulnerable artists and creators. Notwithstanding the highly publicized demands of musicians, authors, actors, and other creative professionals, imposing a licensing requirement is unlikely to protect the jobs or incomes of the underpaid working artists that media and entertainment behemoths have exploited for decades. Because of the imbalance in bargaining power between creators and publishing gatekeepers, trying to help creators by giving them new rights under copyright law is, as EFF Special Advisor Cory Doctorow has written, like trying to help a bullied kid by giving them more lunch money for the bully to take.
Entertainment companies’ historical practices bear out this concern. For example, in the late-2000’s to mid-2010’s, music publishers and recording companies struck multimillion-dollar direct licensing deals with music streaming companies and video sharing platforms. Google reportedly paid more than $400 million to a single music label, and Spotify gave the major record labels a combined 18 percent ownership interest in its now-$100 billion company. Yet music labels and publishers frequently fail to share these payments with artists, and artists rarely benefit from these equity arrangements. There is no reason to believe that the same companies will treat their artists more fairly once they control AI.
You’re getting douchevoted because on lemmy any AI-related comment that isn’t negative enough about AI is the Devil’s Work.
i will train my jailbroken kindle too…display and storage training… i’ll just libgen them…no worries…it is not piracy
Of course we have to have a way to manually check the training data, in detail, as well. Not reading the book, im just verifying training data.
