> Buying used copies of books, scanning them, and training on it is fine.

But nobody was ever going to that, not when there are billions in VC dollars at stake for whoever moves fastest. Everybody will simply risk the fine, which tends to not be anywhere close to enough to have a deterrent effect in the future.

That is like saying Uber would have not had any problems if they just entered into a licensing contract with taxi medallion holders. It was faster to just put unlicensed taxis on the streets and use investor money to pay fines and lobby for favorable legislation. In the same way, it was faster for Anthropic to load up their models with un-DRM'd PDFs and ePUBs from wherever instead of licensing them publisher by publisher.

  The Google Books project also faced a copyright lawsuit, which was eventually decided in favor of Google.

  After contacting major publishers about possibly licensing their books, [former head of the Google Books project] bought physical books in bulk from distributors and retailers, according to court documents. He then hired outside organizations to dissemble the books, scan them and create digital copies that could be used to train the company’s AI. technologies.

  Judge Alsup ruled that this approach was fair use under the law. But he also found the company’s previous approach — downloading and storing books from shadow libraries like Library Genesis and Pirate Library Mirror — was illegal.

  > Anthropic also agreed to delete the pirated works it downloaded and stored.

  > As part of the settlement, Anthropic said that it did not use any pirated works to build A.I. technologies that were publicly released.

> Rainbows End was prescient in many ways.

Agreed. Great book for those looking for a read: https://www.goodreads.com/book/show/102439.Rainbows_End

The author, Vernor Vinge, is also responsible for popularizing the term 'singularity'.

I think the jury is still out on how fair use applies to AI. Fair use was not designed for what we have now.

I could read a book, but its highly unlikely I could regurgitate it, much less months or years later. An LLM, however, can. While we can say "training is like reading", its also not like reading at all due to permanent perfect recall.

Not only does an LLM have perfect recall, it also has the ability to distribute plagiarized ideas at a scale no human can. There's a lot of questions to be answered about where fair use starts/ends for these LLM products.

  >  but AI doesn't change the motivations and goals behind copyright

To be even more clear - this is a settlement, it does not establish precedent, nor admit wrongdoing. This does not establish that training is fair use, nor that scanning books is fine. That's somebody else's battle.

> Buying used copies of books

It remains deranged.

Everyone has more than a right to freely have read everything is stored in a library.

(Edit: in fact initially I wrote 'is supposed to' in place of 'has more than a right to' - meaning that "knowledge is there, we made it available: you are supposed to access it, with the fullest encouragement").

I wonder what Aaron Swartz would think if he lived to see the era of libgen.

You can't grab pirated stuff and then hope fair use magically sanitizes it

Yes, the ruling was a massive win for generative AI companies.

The settlement was a smart decision by anthropic to remove a huge uncertainty. 1.5 is not small, but it won't stop them or slow them significantly.

Google scanned many books quite a while ago, probably way more than LibGen. Are they good to use them for training?

> It’s important in the fair use assessment to understand that the training itself is fair use

IIUC this is very far from settled, at least in US law.

META did pirate basically all books in Anna’s archive but if I remember correctly they just whispered a a cried sorry and it ended up as that. Why are they also not asked to pay?

It should not be fine to train on them, because you are creating derivative works, exactly like when you deal with music.

> Buying used copies of books, scanning them, and training on it is fine.

Awesome, so I just need enough perceptrons to overfit every possible copyrighted works then?

> It’s important in the fair use assessment to understand that the training itself is fair use

Is this completely settled legally? It is not obvious to me it would be so

The Librareome project was about simply scanning books, not training AI with them. And it was a matter of trying to stop corporations from literally destroying the physical books in the process. I don't know that this is applicable.

Do they actually need to scan the book?

Or can they buy the book, and then use the pirated copy?

Okay, so the blame for the offense was laundered..

Nevertheless, a crime is a crime.

I'm so over this shift in America's business model.

Original Silicon Valley model, and generally the engine of American innovation/growth/wealth equality for 200 years: Come up with a cool technology, build it in your garage, get people to fund it and sell it because it's a better mousetrap.

New model: Still come up with a cool idea, still get it funded and sold, but the idea involves committing crime at a staggering scale (Uber, Google, AirBnB, all AI companies, long list here), and then paying your way out of the consequences later.

Look some of these laws may have sucked, but having billionaires organize a private entity that systematically breaks them and gets off with a slap on the wrist, is not the solution. For one thing, if innovation requires breaking the law, only the rich will be able to innovate because only they can pay their way out of the law. For another, obviously no one should be able to pay their way out of following the law! This is basic "foundations of society" stuff that the vast majority of humans agree on in terms of what feels fair and just, and what doesn't.

Go to a country which has really serious corruption problems, like is really high on the corruption index, and ask the people there what they think about it. I mean I live in one and have visited many others so I can tell you, they all hate it. It not only makes them unhappy, it fills them with hopelessness about their future. They don't believe that anything can ever get better, they don't believe they can succeed by being good, they believe their own life is doomed to an unappealing fate because of when and where they were born, and they have no agency to change it. 25 years ago they all wanted to move to America, because the absence of that crushing level of corruption was what "the land of opportunity" meant. Now not so much, because America is becoming more like their country.

This timeline ends poorly for all of us, even the corrupt rich who profit from it, because in the future America will be more like a Latin American banana republic where they won't be able to leave their compounds for fear of getting Luigi'ed. We normal people get poverty, they get fear and death, everyone loses. The social contract is collapsing in front of our eyes.

Has it been decided that training models is fair use? Has it been decided in all jurisdictions?

Yes, but the cat is out of the bag now. Welcome to the era of every piece of creative work coming with an EULA that you cannot train on it. It will be like clearing samples.

This is excellent news because it means that folks who pay for printed books and scan them also can train with their content. It's been said already that we've already trained on "the entire (public) internet." Printed books still hold a wealth of knowledge that could be useful in training models. And cheap, otherwise unwanted copies make great fodder for "destructive" scanning where you cut the spine off and feed it to a page scanner. There are online services that offer just that.

> pirating of the books is the issue

I have an author friend who felt like this was just adding insult to injury.

So not only had his work been consumed into this machine that is being used to threaten his day job as a court reporter, not only was that done without seeking his permission in any way, but they didn’t even pay for a single copy.

Really embodies raising your middle finger to the little guy while you steamroll him.

It is related to scalable mode training, however. Chopping the spine off books and putting the pages in an automated scanner is not scalable. And don't forget about the cost of 1) finding 2) purchasing 3) processing and 4) recycling that volume of books.

Then shouldn’t they be liable for at least 25 times this amount?

Wdym Rainbows End was prescient?

Paying $3,000 for pirating a ~$30 book seems disproportionate.

Thanks for the reminder that what the Internet Archive did in its case would have been legal if it was in service of an LLM.

> It’s important in the fair use assessment to understand that the training itself is fair use,

I think that this is a distinction many people miss.

If you take all the works of Shakespeare, and reduce it to tokens and vectors is it Shakespeare or is it factual information about Shakespeare? It is the latter, and as much as organizations like the MLB might want to be able to copyright a fact you simply cannot do that.

Take this one step further. IF you buy the work, and vectorize it, thats fine. But if you feed it in the vectors for Harry Potter so many times that it can reproduce half of the book, it becomes a problem when it spits out that copy.

And what about all the other stuff that LLM's spit out? Who owns that. Well at present, no one. If you train a monkey or an elephant to paint, you cant copyright that work because they aren't human, and neither is an LLM.

If you use an LLM to generate your code at work, can you leave with that code when you quit? Does GPL3 or something like the Elastic Search license even apply if there is no copyright?

I suspect we're going to be talking about court cases a lot for the next few years.

[dead]

I guess they must delete all models since they acquired the source illegally and benefitted from it, right? Otherwise it just encourages others to keep going and pay the fines later.

> Buying used copies of books, scanning them, and training on it is fine.

Buying used copies of books, scanning them, and printing them and selling them: not fair use

Buying used copies of books, scanning them, and making merchandise and selling it: not fair use

The idea that training models is considered fair use just because you bought the work is naive. Fair use is not a law to leave open usage as long as it doesn’t fit a given description. It’s a law that specifically allows certain usages like criticism, comment, news reporting, teaching, scholarship, or research. Training AI models for purposes other than purely academic fits into none of these.

Don't forget: NO LEGAL PRECEDENT! which means, anybody suing has to start all over. You only settle in this scenario/point if you think you'll lose.

Edit: I'll get ratio'd for this- but its the exact same thing google did in it's lawsuit with Epic. They delayed while the public and courts focused in apple (oohh, EVIL apple)- apple lost, and google settled at a disadvantage before they had a legal judgment that couldn't be challenged latter.

So they can also keep models trained on the datasets? That seems pretty big too, unless the half life of models is so low it doesn't matter.

So... it would be a lot cheaper to just buy all of the books?

Thank you. I assumed it would be quicker to find the link to the case PDF here, but your summary is appreciated!

Indeed, it is not only payout, but the destruction of the datasets. Although the article does quote:

> “Anthropic says it did not even use these pirated works,” he said. “If some other generative A.I. company took data from pirated source and used it to train on and commercialized it, the potential liability is enormous. It will shake the industry — no doubt in my mind.”

Even if true, I wonder how many cases we will see in the near future.

Only 500,000 copyrighted works?

I was under the impression they had downloaded millions of books.

I’m an author, can I get in on this?

I think western companies will be just fine -- Anthropic is settling because they illegally pirated books from LibGen back in 2021 and subsequently trained models on them. They realized this was an issue internally and pivoted to buying books en masse and scanning them into digital formats, destroying the original copies in the process (they actually hired a former lead in the Google Books project to help them in this endeavor!). And a federal judge ruled a couple months ago that training on these legally-acquired scanned copies does not constitute fair use -- that the LLM training process is sufficiently transformative.

So the data/copyright issue that you might be worried about is actually completely solved already! Anthropic is just paying a settlement here for the illegal pirating that they did way in the past. Anthropic is allowed to train on books that they legally acquire.

And sure, Chinese AI companies could probably scrape from LibGen just like Anthropic did without getting in hot water, and potentially access a bit more data that way for cheap, but it doesn't really seem like the buying/scanning process really costs that much in the grand scheme of things. And Anthropic likely already has legally acquired most of the useful texts on LibGen and scanned them into its internal library anyways.

(Furthermore, the scanning setup might actually give Anthropic an advantage, as they're able to digitize more niche texts that might be hard to find outside of print form)

The West can end the endless pain and legal hurdles to innovation by limiting the copyright. They can do it if there is will to open up the gates of information to everyone. The duration of 70 years after death of the author or 90 years for companies is excessively long. It should be ~25 years. For software it should be 10 years.

And if AI companies want recent stuff, they need to pay the owners.

However, the West wants to infinitely enrich the lucky old people and companies who benefited from the lax regulations at the start of 20th century. Their people chose to not let the current generations to acquire equivalent wealth, at least not without the old hags get their cut too.

It's naive to think Chinese models have a free pass. Local censorship, language/data biases, and export restrictions cut both ways.

True enough, but training on synthetic data now seems to be pushing SOTA.

This isn’t is a race to the bottom. They could have bought these books instead of pirating them.

But most marginal training of Anthropic, OpenAI and Google models is done on LLM paraphrased user data on those platforms. That user data is proprietary and obviously way more valuable than random books.

wow, I found 8 of my books!

Wild - I searched my name out of curiosity and my PhD research papers turned up. Worth submitting my contact details I guess

Thank you for posting this!

I suspected my work was in the dataset and it looks like it is! I reached out via the form.

It’s pretty incredible that the vast majority of authors will make more money for their books from this settlement that they ever have from selling their books.

From the article:

> Although the payment is enormous, it is small compared with the amount of money that Anthropic has raised in recent years. This month, the start-up announced that it had agreed to a deal that brings an additional $13 billion into Anthropic’s coffers. The start-up has raised a total of more than $27 billion since its founding in 2021.

You're joking, but that's actually a good pitch. There was a significant legal issue hanging over their heads, with some risk of a potentially business-ending judgment down the line. This makes it go away, which makes the company a safer, more valuable investment. Both in absolute terms and compared to peers who didn't settle.

Isn't that how the whole system operates? Everyone is a conduit to allow rich people to enrich themselves further. The amount and quality of opportunities any individual receives are proportional to how well it serves existing capital.

So long as there is an excuse to justify money flows, that's fine, big capital doesn't really care about the excuse; so long as the excuse is just persuasive enough to satisfy the regulators and the judges.

Money flows happen independently, then later, people try to come up with good narratives. This is exactly what happened in this case. They paid the authors a lot of money as a settlement and agreed on a narrative which works for both sets of people; that training was fine, it's the pirating which was a problem...

It's likely why they settled; they preferred to pay a lot of money and agree on some false narrative which works for both groups rather than setting a precedent that AI training on copyrighted material is illegal; that would be the biggest loss for them.

They wanted to move fast and break things. No one made them.

very unsurprisingly, new york times is going to frame this as a win for "the little guy" when in reality it's just multi-billion dollar publishers, with a long rich history of their own exploitive practices, hanging on for dear life against generative AI

Fair use isn't about how you access the material, its about what you can do with it after you legally access it. If you don't legally access it, the question of fair use is moot.

This implies training models is some sort of right.

This is a settlement. It does not set a precedent nor even admit to wrongdoing.

> otherwise only the big companies who can afford to pay off publishers like Anthropic will be able to do so

Only well funded companies can afford to hire a lot of expensive engineers and train AI models on hundreds of thousands of expensive GPUs, too.

Something tells me many the grassroots LLM training people are less concerned about legality of their source training set than the big companies anyway.

I wonder how much it would cost to buy every book that you'd want to train a model.

I don't know if I agree with it, but you could argue that if a model was built for purely academic purposes, and then used for purely academic purposes, it could meet requirements for fair use.

(Half joking but) I wonder if musicians need to worry if they learned to play by listening to cassette mixtapes.

I wish the hn rules were more flexible because I would write the best comment to you right now.

They're paying $3000 per book. It would've been a lot cheaper to buy the books (which is what they actually did end up doing too).

That metaphor doesn't really work. It's a settlement, not a punishment, and this is payment, not a fine. Legally it's more like "The store wasn't open, so I took the items from the lot and paid them later".

It's not the way we expect people to do business under normal circumstances, but in new markets with new products? I guess I don't see much actually wrong with this. Authors still get paid a price they were willing to accept, and Anthropic didn't need to wait years to come to an agreement (again, publishers weren't actually selling what AI companies needed to buy!) before training their LLMs.

Silicon Valley's unofficial motto for the last 15 years

I’m sure this’ll be misreported and wilfully misinterpreted because of the current fractious state of the AI discourse, but given the lawsuit was to do with piracy, not the copyright-compliance of LLMs, and in any case, given they settled out of court, thus presumably admit no wrongdoing, conveniently no legal precedent is established either way.

I would not be surprised if investors made their last round of funding contingent on settling this matter out of court precisely to ensure no precedents are set.

Anthropic certainly seems to be hoping that their competitors will have to face some consequences too:

>During a deposition, a founder of Anthropic, Ben Mann, testified that he also downloaded the Library Genesis data set when he was working for OpenAI in 2019 and assumed this was “fair use” of the material.

Per the NYT article, Anthropic started buying physical books in bulk and scanning them for their training data, and they assert that no pirated materials were ever used in public models. I wonder if OpenAI can say the same.

Maybe, though this lawsuit is different in respect to the piracy issue. Anthropic is paying the settlement because they pirated the books, not because training on copyrighted books isn’t fair use which isn’t necessarily true with the other cases.

That was my first though. While not legal precedent, it does sort of open the flood gates for others.

Yes to the first part. Put your site behind a login wall that requires users to sign a contract to that effect before serving them the content... get a lawyer to write that contract. Don't rely on copyright.

I'm not sure to what extent you can specify damages like these in a contract, ask the lawyer who is writing it.

I'd argue you don't actually want this! You're suggesting companies should be able to make web scraping illegal.

That curl script you use to automate some task could become infringing.

Maybe some kind of captcha like system could be devised that could be considered a security measure under the DMCA and not allowed to be circumvented. Make the same content available under a licence fee through an API.

No. Neither legally or technically possible.

I'm sure one can try, but copyright has all kinds of oddities and carve-outs that make this complicated. IANAL, but I'm fairly certain that, for example, if you tried putting in your content license "Free for all uses public and private, except academia, screw that ivory tower..." that's a sentiment you can express but universities are under no obligation legally to respect your wish to not have your work included in a course presentation on "wild things people put in licenses." Similarly, since the court has found that training an LLM on works is transformative, a license that says "You may use this for other things but not to train an LLM" couldn't be any more enforceable than a musician saying "You may listen to my work as a whole unit but God help you if I find out you sampled it into any of that awful 'rap music' I keep hearing about..."

The purpose of the copyright protections is to promote "sciences and useful arts," and the public utility of allowing academia to investigate all works(1) exceeds the benefits of letting authors declare their works unponderable to the academic community.

(1) And yet, textbooks are copyrighted and the copyright is honored; I'm not sure why the academic fair-use exception doesn't allow scholars to just copy around textbooks without paying their authors.

This is already a thing in several places.

EU has copyright exemptions for AI training. You don't need to respect opt outs if you are doing research.

South Korea, Japan has some exemptions too I think?

Singapore has very strong copyright exemptions for AI training. You can completely ignore opt-outs legally, even if doing it commercially.

Just search up "TDM laws globally".

As long as you're not distributing, it's legal in Switzerland to download copyrighted material. (Switzerland was on the naughty US/MPAA list for a while, might still be)

"That could push the industry toward consolidation"

Based on history this is not a possibility but a certainty.

The larger players - who grew because of limited regulations - will start supporting stricter regulation and compliance structures in order to increase the barrier of entry with the excuse of "Oh we learned our lesson, you are right". The hypocrisy is crazy but it makes sense from a capitalistic perspective.

I thought the opposite - they set a precedent indicating that reproduction of a copyrighted text by an LLM is infringement. If authors refuse to sell to them (via legal terms indicating LLMs aren't allowed), it's infringement. No?

I'd be curious to hear from a legal professional...

more of a large difference in constant factor, like a galactic algorithm for data trawling

I don't know if anyone has actually read the article or the ruling, but this is about pirating books.

Anthropic went back and bought->scanned->destroyed physical copies of them afterward... but they pirated them first, and that's what this settlement is about.

The judge also said:

> “The training use was a fair use,” he wrote. “The technology at issue was among the most transformative many of us will see in our lifetimes.”

So you don't need to pay $3,000 per book you train on unless you pirate them.

This notably wasn't one of the allegations levied against Anthropic, as Claude was accompanied by software that filtered any infringing outputs. From the relevant opinion finding Anthropic's use of the books to be fair use:

> When each LLM was put into a public-facing version of Claude, it was complemented by other software that filtered user inputs to the LLM and filtered outputs from the LLM back to the user. As a result, Authors do not allege that any infringing copy of their works was or would ever be provided to users by the Claude service.

(from Bartz v. Anthropic in the Northern District of California)

It's not a crime. It is civil lawsuit.

I wonder if Antrhopic's lawyers have enough of a sense of humor to take you up on that if you sent them an email asking...

I assumed they were literally just lying.

it sounds like the former

You can check, see: https://news.ycombinator.com/item?id=45144261

They already, prior to this lawsuit, prior to serving public models, replaced this data set with one they made by scanning purchased books. Destroying the data set they aren't even using should have approximately zero effect.

Perhaps they'll quickly rent the whole contents of a few physical libraries and then scan them all

Copies of these books are for sale for much less than that - very very few books demand a price that high.

They're paying much more than the actual damages because US copyright law comes with statutory damages for infringement of registered works on top of actual damages, between $200 and $150,000 per work. And the two sides negotiated this as a fair settlement to reduce the risk of an unfavourable outcome.

If you acquire something illegally of course the judgement against you has to be much higher than the legal price. Why would anyone purchase anything if the worst thing that could happen to you for stealing it was just paying the retail price?

They are not paying for reading the book, they are paying for redistributing the book in perpetuity presumably.

Yes. Or download it legally (e.g. web content not behind a paywall).

Reach the publishers or resellers (like amazon for instance)

Give them this order : "I want to buy all your books as epub"

Pay and fetch the stuff

That's all

Docket: https://www.courtlistener.com/docket/69058235/bartz-v-anthro...

Proposed settlement: https://storage.courtlistener.com/recap/gov.uscourts.cand.43...

https://authorsguild.org/news/what-authors-need-to-know-abou... looks useful.

Deep research on Claude perhaps for some irony if you will.

$150,000 statutory damages for willful infringement.

Make sure to grab the mother-of-all-torrents I guess if you're going to go that path. That way you get more bang for your 1.5B penalty.

A million torrents would cost 1,500 each.

I suspect you could acquire and scan every readily purchasable book for much less than $3k each. Scanhouse for instance charges $0.15 per page for regular unbound (disassembled) books, plus $0.25 for supervised OCR, plus another dollar if the formatting is especially complex; this comes out to maybe $200-300 for a typical book. Acquiring, shipping, and disposing of them all would of course cost more, but not thousands more.

The main cost of doing this would be the time - even if you bought up all the available scanning capacity it would probably take months. In the meantime your competition who just torrented everything would have more high-quality training data than you. There are probably also a fair number of books in libgen which are out of print and difficult to find used.

It's a tiny amount of data relatively speaking. Much more expensive per token than almost any data source imaginable

The court has to give preliminary approval to the settlement first. After that there should be a notice period during which the lawyers will attempt to reach out and tell you what you need to do to receive your money. (Not a lawyer, not legal advice).

You can follow the case here: https://www.courtlistener.com/docket/69058235/bartz-v-anthro...

You can see the motion for settlement (what the news article is about) here: https://storage.courtlistener.com/recap/gov.uscourts.cand.43...

They just raised $13B, so yes

https://www.anthropic.com/news/anthropic-raises-series-f-at-...

> It's less than 1% Anthropic's valuation

The settlement is real money though. Valuation is imaginary.

There’s alternatives to wiping out the company that could be fair. For example, a judgement resulting in a shares of the company or revenue shares in the future rather than a one time pay off.

Writers were the true “foundational” piece of LLMs, anyway.

Nobody is winning the AI race.

Because everyone is expecting AGI now and it's not happening with our current tech.

It was coming regardless of the case results.

Maybe it's time to get some Llama models copied before an overzealous court rules badly.

If it was a sure thing, then the rights holders wouldn't have accepted a settlement deal for a measly couple billion. Both sides are happier to avoid risking losing the suit.

They would only be wiped out if the court awarded the maximum statutory damages (or close to it). There was never any chance of that happening.

Wait, DID they admit guilt? A lot of times companies settle without admitting guilt.

I mean the ruling does in fact find that treating this particular kind of creative work qualifies as fair use.

All the ai companies are still using books as training data. Theyre just finding the cheapest scanned copies they can get their hands on to cover their ass

This isn't a criminal case so zero people of any financial position would end up in prison.

No one, rich or poor, goes to jail for downloading books.

Books aren't hosted publicly online free for anyone to access. The court seems to think buying a book and scanning it is fair use. Just using pirated books is forbidden. Blogs weren't accessed via pirating.

The blogger’s content was freely available, this fine is for piracy.

The settlement was for downloading the pirated books, not training from them. Unless they're paywalled it would be hard to argue the same for a blog.

Printing press, audio recording, movies, radio, television were also transformative. Did not get rid of copyright or actually brought them.

I feel it is insane that authors do not receive some sort of standard compensation for each training use. Say a few hundred to a few thousand depending on complexity of their work.