← Facts Matter
Facts Matter
Internal Docs Discovered in Lawsuit Exposed Anthropic’s Mass Book Shredding Project
00:17:12
Loading summary
Transcript7 lines
- [00:00]
A
Now, for this next segment, I'm going to issue you a warning right at the very top. If you're like me and when you read books, you don't even like to bend the spine, you definitely don't crease the pages, you really are a book lover who likes to keep them pristine. Perhaps consider changing the video now, because it turns out that Anthropic, the AI giant company behind the Claude Chatbot, they have been secretly destroying millions of books in the process of feeding them into the AI system. And it's a bit brutal how they're destroying those books. They're slicing the spines off, cutting the pages to a certain size, scanning them in mass, and then shredding and discarding and recycling whatever remains, all done to avoid paying copyright fees to the creators. This initiative of theirs was called Project Panama, and it was supposed to be a secret. We only learned about it through sheer luck. Basically, it was made public because of the discovery process in an unrelated lawsuit. And the kicker is that it's all 100% legal. And ironically, it's only legal because Anthropic destroys the books in the process of scanning them. Why? Well, let's go through that together right after you smash those like and subscribe buttons so this video can reach ever more people via the YouTube algorithm. Thank you so much for that. Now, to start with, let's rewind the clock back to 2021, the middle of the pandemic, the lockdowns. A lot of people are working from home. People are wearing masks. There's the debate about the vaccines. And amidst all that, that was when, according to court records, Anthropic decided to use millions of books to train their AI. Chatbot, Claude, you see, AI does not get intelligent on its own, similar to us humans. It needs a learning process. And so Anthropic solution was to illegally download millions of books from pirate websites and feed those illegally downloaded books into their AI system for learning purposes. Journalist Will Oremis explain how this whole process worked over on the Vergecast. Take a listen.
- [02:01]
B
The way they actually started was not by acquiring real physical books the way they started, which is the same way that, according to court documents, OpenAI is is alleged to have started, which is by pirating vast libraries of already digitized books. These are called shadow libraries. They're available online in shady, you know, shady corners of the torrent world.
- [02:26]
A
Now, for a while, there was speculation that these AI companies were doing this type of a thing, but it wasn't until two years later, in 2023, when it was confirmed specifically, in August of 2023, the Atlantic publication did two things. Firstly, they ran an article exposing this practice of downloading pirated books for AI training purposes. And secondly, they published a database of the specific books that were being used for this purpose. And that database was searchable, meaning that for the first time, authors could go in and see whether or not their book's contents were stolen for this purpose. Several authors did this, including Andrea Bartz, Charles Graeber, as well as Kirk Wallace Johnson, all of whom discovered that their works were indeed illegally downloaded and fed into Claude. Eventually, a class action lawsuit was filed with those three authors as the name plaintiffs against Anthropic, citing copyright infringement. And interestingly, this lawsuit proceeded at a pretty rapid pace through the court system. Quote, once it was filed last summer, the suit progressed with unusual speed. After a judge ruled that Anthropic had illegally acquired millions of books and and that the company's executives knew the online libraries they downloaded contained pirated books, the parties agreed to settle. Now we'll get to why that case was likely sped through the courts and settled so quickly by the lawyers. But in a practical sense, the case did eventually wind up settling for a cool $1.5 billion, which to date is the largest copyright settlement in US history. It covered about 482,000 of the pirated titles, and the author, slash publisher of each book is slated to receive about $3,000 per title. Anthropic also agreed to destroy the pirated files. And so at this point, you might be saying, all's well and good, they are going to destroy the pirated files. The authors and publishers got a payout. You can put a bow on top of it. It's done. However, during the discovery process in this class action lawsuit, something came to light which I think it's fair to say astounded the publishing world. Based on internal company records, it turned out that Anthropic only relied on this pirated book method for about two years. Even before all the lawsuits came. They stopped pirating books because their lawyers advised them that this was a legally dangerous thing to do. And so instead what they did was to hire the former head of partnerships for Google's book scanning project. And they tasked him with the following quote. In February of 2024, Anthropic hired the former head of partnerships for Google's book scanning project, Mr. Tom Turvey. He was tasked with obtaining all the books in the world while still avoiding as much legal practice business slog as possible. Anthropic spent many millions of dollars to purchase millions of print books, often in used condition. Then its service providers stripped the books from their bindings, cut their pages to size, and scanned the books into digital form, discarding the paper originals. Each print book resulted in a PDF copy containing images of the scanned pages with machine readable text, including front and back cover scans for softcover books. Using this method, Anthropic acquired copies of millions of books. All right, just to pause here for a super quick moment. If you've watched our program, Facts Matter, then you know that we focus on one thing, which is the facts. Not speculation, not narratives, just what the data shows. And right now, the data globally is pointing in a direction people should really be paying attention to, which is the fact that global conflicts are fueling market uncertainty. That's why I believe in owning something real, something physical, something that the government can't just inflate away, which is physical gold and physical silver from today's sponsor, which is American Hartford Gold, the company that I trust and who I've been personally buying from for many years. Now, these are tangible assets. They're not tied to policy decisions or shifting narratives. Gold has held its value for thousands of years through inflation, currency changes, the rise and fall of empires, as well as general global instability. In a certain sense, gold actually thrives in periods of global instability. And American Hartford Gold is a good company. They have an A plus rating with a Better Business Bureau, thousands of five star reviews. They can deliver your gold and your silver directly to your doorstep. Or they can help you set up a gold IRA if you're trying to diversify your retirement. And right now, when you call and mention my name, Roman, you can get up to $20,000 in free silver with your first qualifying purchase. So give them a call, 866-24-2352 or just text Roman to 65532. That's again, 866-242-2352 or text Roman to 65532. I'll also throw all their details. You'll be able to find them down in the description box below, meaning that they were buying in bulk, millions of books. Slicing the spines of those books clean off, cutting the pages into the right size for the scanners, then scanning them in mass. Then those PDFs were being fed into the AI system for training purposes as well as for future reference. And then whatever remained of the original books were destroyed. According to the documents, they were shredded and then recycled. Basically, if I can summarize it, millions of books died so that Claude could thrive. Now Internally, this initiative was called Project Panama. And it really was not ever supposed to be known by the public. It was never supposed to come to light, evidenced by this one internal company memo made public only by the lawsuit, which said the following quote, project Panama is an effort to destructively scan all the books in the world. We, meaning Anthropic, don't want it to be known that we are working on this. Now, at this point, a good question to ask might be why? Why would Anthropic go through all the trouble of buying millions of physical books, having to cut them all to size, of course, slice the spines off and then cut them all to size, scan them one page at a time and then destroy them rather than just buy the digital copies of those same books. It seems like that would be the obvious choice. And the answer for why they chose this route appears to actually be rather simple. Digital ebooks are almost always sold under a license. The person buying it has to agree to some kind of a terms of service in order to be able to read the text. And usually as the buyer, you get very limited rights to access and to use that particular file under the publisher's terms. Basically, you can access, you don't own anything, you can just access the file on whatever platform the publisher says you can access it on. And then you have to agree to the Terms of service. And included in almost every version of those terms of service are things like the prohibition of commercial AI training, the prohibition on mass reproduction, as well as the prohibition on permanent internal archiving, the exact things that Anthropic would want to do in order to train their AI systems. However, physical books are different. Physical books are generally owned under what's known as the first sale doctrine. You can find it in Title 17 of the U.S. federal Code Standard subsection 109. It means that once you buy a lawful copy of a physical book, it's yours and you can do a lot of things with it without having to reach out and ask the copyright owner for permission. You can resell it, you can destroy it, and very importantly, you can convert your specific physical copy to a different format, all without asking for permission. And what's funny is that initially Anthropic tried going the first route with the ebooks, and they tried to negotiate directly with the publishers for the rights and. But that process proved to be so time consuming and so convoluted that as I alluded to earlier, the company referred to it as simply legal and business slog. In fact, the guy from Google that they hired to lead the book effort. He tried to initially contact the major publishers about licensing their titles for AI training, But according to the court records, those back and forth conversations, quote unquote withered. And so instead they chose to go with the other destructive scanning method. No lengthy negotiations necessary, no back and forth email chains. You just buy a single copy of the book, you scan it, you destroy it, and then you're done. And perhaps the best part for Anthropic of using that method is that the whole thing is 100% legal. To that end, regarding this destructive scanning method, the judge in the class action lawsuit wrote the following quote, it was a fair use because all Anthropic did was replace the print copies it had purchased for its central library with with more convenient space saving and searchable digital copies for its central library without adding new copies, creating new works, or redistributing existing copies. And so, based on that, it seemed like the judge not only greenlit Anthropic's use of these books to train its AI models without the express permission of the author or the publisher, but furthermore, in a twist of irony, it's only allowed because Anthropic destroyed the physical book that it used to make the digital copy. The judge wrote that the company just made a more convenient space saving version of that same book without creating a new copy. They just changed the format. Meaning if they did not destroy the original copy, then they would be creating a new copy and therefore that would not be allowed. And the implications here are huge with this ruling. Not just Anthropic, but all these giant AI companies, they no longer have to go out and make individual deals with writers, publishers and editors to use a book's content. They simply have to go out, buy a single copy of that book, scan it, put it into the system, and then destroy the original. However, even though it is technically legal, a big lingering question is whether or not it's palatable to the public. Because I myself, as I mentioned at the top of the episode, I try to not even crease the spines of the books that I read. And so imagining somebody slicing off the spines and then cutting the pages and shredding millions of books is a bit unnerving to say the least. At least to me. And a good question might be, if this continues moving forward, given the fact that they got the legal green light, where will this end up? Now? As a corollary, I will say that OpenAI, Microsoft, as well as Elon Musk's X AI, they recently came out and started talking about using non Destructive scanning, moving forward. And so let's see where that goes. But even you know, regardless of that, there is a place somewhere on earth right now where there is an enormous paper graveyard overflowing with the remains of millions of sacrifice books. Also, that Claude can help you and I write that term paper or that resume or answer a question about an email that we have to respond to. So, you know, take that for what it's worth. Finally, we asked Claude what he thinks about all this, about this whole situation. And here's what Claude told us, quote, choosing destruction over reversible digitization purely for speed and cost, on a scale of hundreds of thousands to millions of books, with none of the public facing benefit that made earlier mass digitization projects more defensible. That's a choice I find hard to feel good about. And you know what, Claude?
- [13:25]
B
Me too.
- [13:26]
A
If you'd like to go deeper into this project, Panama as it's been discovered, you can find my research notes. The links will be down there in the description box below, right below those like and subscribe buttons, both of which I hope you smash so this video can reach ever more people. And, and then lastly, please leave your comments in the comment section below. Do you agree with this project? Do you think it's kind of overblown? Because initially I thought that they were taking like rare books and like manuscripts and books that are out of circulation. There's only one kind in the world of this particular book, and they were splicing those spines off and copying those books and then destroying the originals there. So like the only copy of this one book would be in Claude and it would be stealing from the public's access to that particular book. However, looking into the documents, maybe there's a few copies of those types of books, but it seems like the places where they were mass buying those books from were just kind of like the same places that you would be buying books from online. And they weren't offering those type of like rare books. So it seems like they were just buying very generic books that are in circulation and print. They were just happening to buy one copy instead of reading it. They would splice it and do do the scanning and then discard it. So that's not too crazy. I mean, it's like going into Barnes and Noble, buying one of each book that they have and then doing that process, but they would still have other copies for other people to buy. So I don't know. I mean, so that, that felt good to me at the very least, that they weren't taking out these like rare manuscripts and taking them away from the public. But it's still kind of visually difficult to kind of look at a book having its spine ripped off and imagining millions of that. Of those type of things happening. So. I don't know. Leave your thoughts in the comments section below. Is it overblown? Am I just. Am I just too much of a, you know, I don't know, a book lover, as I call it? Let me know. And then lastly, on a totally unrelated topic, if you have ever wondered where we go after we die, Great question. I'm sure many of us have wondered that. Well, over at the Epoch Times, we just published a phenomenal new film called Final Hours where we interview people who died previously. They were clinically dead. No heartbeat, no breath, no brain activity. Or conversely, some of them were brain dead, meaning they had no brain activity. But regardless which camp they fell into, they both came back to life and they were able to recall what they had experienced on the other side and then tell us what they experienced. And the stories they told were amazing. And many of the people who watched that film, they told us that listening to those stories of the people who had actually died, saw the other side and came back to tell the tale changed their view on what life actually was. Such a fundamental thing they changed their view on after watching the film, which is like the best thing you can imagine the film to do. So that's all to say. Check it out. It you can watch Final hours over on EpicTV. The link to it will be right there at the top of the description box below. Hope you check it out. Technically, I was dead. But I felt that this was more real than real. And I felt more alive than life itself.
- [16:16]
C
Thousands have glimpsed something vast and loving waiting on the other side. I was surrounded by beings, some of them I recognize from this life. It's more just like a light. The light of God. I was so happy. I just felt like more and more of the pain of this world was just floating away. What do these human experiences mean for our collective soul?
- [16:46]
A
The more we come into harmony, love and kindness, the more we're aligned with the essence of the universe. And then, until next time, I'm your host, Roman from the Epoch Times. Stay informed and most importantly, stay free.