Anthropic’s AI Lawsuit Settlement May Not Go Through, But It Exposes A Truth About Copyright
from the copyright-is-a-mess dept
The latest generation of AI systems, based on large language models (LLMs), is perceived as the biggest threat in decades to the established copyright order. The scale of that threat can be gauged by the flurry of AI lawsuits that publishers and others have launched against generative AI companies. Since the first of these, reported here on Walled Culture back in January 2023, there have been dozens of others, catalogued on Wikipedia, and represented visually on the Chat GPT is Eating the World site. One is against Anthropic. Three authors alleged in a class-action lawsuit that the company had used unauthorized copies of their works to train its AI-powered chatbot, Claude:
Anthropic has built a multibillion-dollar business by stealing hundreds of thousands of copyrighted books. Rather than obtaining permission and paying a fair price for the creations it exploits, Anthropic pirated them.
In June of this year, Anthropic won a partial victory. The federal judge considering the case ruled that the training of the company’s system on legally purchased copies of books was fair use, and did not need the authors’ permission. However, Judge Alsup also ruled that Anthropic should face trial for downloading millions of books from sites such as Library Genesis (LibGen) and the Pirate Library Mirror (PiLiMi), both of which held unauthorized copies of works. The potential penalty was huge. Under US law, the company might have to pay damages of up to $150,000 per work. With millions of books allegedly downloaded from the online sites, that could amount to many billions of dollars, even a trillion dollars. Faced with certain ruin if such a penalty were handed down, Anthropic had a strong incentive to settle out of court. On 5 September, the parties proposed just such a settlement. The New York Times had the following summary:
In a landmark settlement, Anthropic, a leading artificial intelligence company, has agreed to pay $1.5 billion to a group of authors and publishers after a judge ruled it had illegally downloaded and stored millions of copyrighted books.
The settlement is the largest payout in the history of U.S. copyright cases. Anthropic will pay $3,000 per work to 500,000 authors.
The agreement is a turning point in a continuing battle between A.I. companies and copyright holders that spans more than 40 lawsuits across the country. Experts say the agreement could pave the way for more tech companies to pay rights holders through court decisions and settlements or through licensing fees.
Some saw the $3,000 per work figure as setting a benchmark for future deals that other AI companies would need to follow in order to settle similar lawsuits (although a settlement would not set a legal precedent). Music publishers were hopeful they could point to the settlement with writers in order to win a similar deal for musicians. Others worried that the overall size of the settlement – $1.5 billion – meant that only the largest companies could afford to pay such sums, shutting out smaller startups and limiting competition in this nascent market. Indeed, big as the $1.5 billion settlement was, it paled in comparison to the $13 billion that Anthropic has recently raised, to say nothing of its nominal $183 billion valuation. But a post by Dave Hansen on the Authors Alliance blog puts all these breathless predictions and impressive numbers into perspective. For example, he points out:
The settlement isn’t a settlement with “authors.” Or at least not just authors. The moment Judge Alsup defined and certified the class in this case to include any rightsholder with an interest in the exclusive copyright right of reproduction in a LibGen/PilLiMi book downloaded by Anthropic, this case became at least as important for publishers as authors.
Crucially, that means only a portion of that $1.5 billion would go to the actual authors. Some of it would go to the usual suspects: the plaintiff’s lawyers. But there are other costs that must be covered too, and Hansen writes: “it’s easy to see that about a quarter to a third of this settlement is being used up before rightsholders see anything.” And then there is the question of who exactly those “rightsholders” are: the writers or the publishers? Probably both in many cases, with a variable split depending on the contract they signed.
Even before those complex questions are addressed, there is a huge assumption that the proposed settlement will go through in its present form. That’s by no means assured. As Bloomberg Law reported, Judge Alsup said he was worried that lawyers were striking a deal behind the scenes that will be forced “down the throat of authors,” and that the agreement is “nowhere close to complete.”
Judge William Alsup at the hearing said the motion to approve the deal was denied without prejudice, but in a minute order after the hearing said approval is postponed pending submission of further clarifying information.
During the first hearing since the deal was announced on Sept. 5, Alsup said he felt “misled” and needs to see more information about the claim process for class members.
Another important point underlined by Dave Hansen on the Authors Alliance blog is that even if the settlement goes through, it doesn’t really help to resolve any of the larger copyright issues raised by the new LLMs:
The settlement isn’t far-reaching. While the payment is record-setting for a copyright class action ($1.5 billion), the settlement terms are pretty narrow in scope. Anthropic simply gets a release from liability for past conduct – namely, use of the LibGen and PiLiMi datasets. It is therefore unlike the proposed settlement in the Google Books Settlement that would have created a novel licensing scheme for a wide variety of future uses
The Google Books Settlement is discussed in Walled Culture the book (free digital versions available), as is another notable moment in copyright history. This concerns the fate of Jammie Thomas, a single mother of two. In 2007, she was found liable for $222,000 in damages for sharing twenty-four songs on the P2P service Kazaa. The judge, ordering a new trial for Thomas, called “the award of hundreds of thousands of dollars in damages unprecedented and oppressive”, and took the opportunity to “implore Congress to amend the Copyright Act to address liability and damages in peer-to-peer network cases such as the one currently before this Court.” On retrial, Thomas was found liable for even more: $1.92 million.
It is instructive to compare that $1.92 million fine for sharing 24 songs – $80,000 per work – with the $1,500 per work that Anthropic is now offering to pay. This confirms once more that when it comes to copyright and its enforcement, there is one law for the rich corporations, and another law for the rest of us.
Follow me @glynmoody on Mastodon and on Bluesky. Originally posted to WalledCulture.
Filed Under: ai, copyright, fair use, settlement, statutory damages, william alsup
Companies: anthropic
Comments on “Anthropic’s AI Lawsuit Settlement May Not Go Through, But It Exposes A Truth About Copyright”
As an author whose work has been used to train LLMs, I don’t expect to make much of anything off of lawsuits or settlements or licensing deals. The publishers exploit authors and pocket the bulk of the profit already, so pretending we’ll be first in line for a payout is unrealistically optimistic.
Much like Etsy sellers work for Etsy’s profit and gig drivers work for Uber and Lyft’s profits, authors work for publishers’ profits. They’re just brilliantly outsourcing the operational costs to you. If you’re raging against LLM companies exploiting authors but not also raging against the big media companies exploiting the authors for far longer, you’re betraying the fact that you don’t give a fuck about authors.
One of the biggest obstacles to this settlement will, I think, be the limited scope of the authors who qualify as class members. Among other things they have to have registered the copyright (or have had it registered by their publisher) before the class action was filed. A huge number of authors didn’t have copyrights registered because you only need that done before you sue over them. Most publishers don’t bother registering until that point. All the authors who fall into that won’t be covered by the settlement and can sue Anthropic on their own or as another class. Alsup’s already commented on this. If he thinks this will leave that much liability unsettled, he may decide to nix the settlement or expand the class to cover the majority of authors but keep the $3000/work figure (which’d increase the amount Anthropic would have to pay by a lot).
Amusingly it’s the tech companies that put Anthropic in this position. The DMCA explicitly made transient copies and copies necessary for software to work non-infringing. In their quest to get click-through EULAs and ToS enforceable, the tech companies gutted those provisions through court decisions. Absent that, Anthropic probably could have argued it was only liable for those works it’s software actually regurgitated verbatim, not the entire database behind it.
Re:
That’d be unlikely given the particulars, I think. They’re not getting hit for the entire database, just the shadow library from pirated works. And part of Alsup’s ruling explicitly calls out that it was a “forever library”, not a transient one.
Re: Re:
Possible, but they could make the case that it was a cache the same as your browser uses to store copies of (copyrighted) web pages so it doesn’t have to fetch them from the server every time you view them. Just like that browser cache the shadow library wasn’t distributed to anyone else, and caching like that was one of the things the DMCA’s provisions were supposed to protect.
Re: Re: Re:
The problem is that they were using it more generally. As the ruling puts it:
Building a central library of works to be available for any number of further uses was itself the use for which Anthropic acquired these copies. One further use was making further copies for training LLMs. But not every book Anthropic pirated was used to train LLMs. And, every pirated library copy was retained even if it was determined it would not be so used. Pirating copies to build a research library without paying for it, and to retain copies should they prove useful for one thing or another, was its own use
Re:
Most authors don’t have a publisher, and never will. But they’ll also never be on LibGen.
“Tech” meaning what in this case? Publishing? Given that the idea of a “company” is itself technology, “tech company” is little more that a term that means whatever the hell the person using it wants it to mean.
Re: Re:
Tech companies like Microsoft, Oracle, Blizzard and so on. The ones who need copyright to back up their claim that even if you bought their software you had to make a copy of it onto your hard drive to install it and you hadn’t bought a license to do _that, and the only way to get that license was to agree to their EULA first (which you never saw until you tried to install the software).
That is worrying. No company should be able to afford to pay for deliverately breaking the law, it should be a given that their entire capital structure is wiped out overnight. In point of fact, we should skip the matter of fines entirely and just wipe out the capital structure directly.
As for smaller companies, they can go ahead and purchase those books for the market value of $10 million like everyone else.
As opposed to all other aspects of the law, which are scrupulously fair without consideration of wealth or social status. /S
Walled culture
This simmering crap has come to a boil.
You own the rights to the content you produce as per the copyright laws or equivalent in your jurisdiction. Copyright maximalists (pirmarly in the US) pretend they own much more and courts are loathe to rule on this.
If you CHOOSE TO MAKE YOUR WORK PUBLIC there are some rights you voluntarily give away. To later claim that someone owes you millions or billions of dollars for using what YOU CHOSE OT PUT OUT isn’t in any way rewarding you for being stupid nor for being creative.
LLMs are a thing now. LLM training on publicly-available materials is a thing too. Piracy is a loaded pejorative term that is NOT a legal term, a term of the art, etc. Somali dudes on a skiff attacking a ship — that’s piracy. Anthropic scraping the net for publicly available documents – not it.
But as in any case where politics interferes with the law, “justice” stands aside for “lobbyists.”
So here we are. Anthropic wanted out quick with a cheap deal. Now it’s in legal limbo. The real problem is that this “piracy” schtick should have been deal with back in the Napster days.
But it wasn’t. Too many pussies eager to “settle up cheap.” And now we reap what they sowed.
Re:
It’s not even clear that it’s illegal. LibGen may be violating copyright, but the people who download from there? The trial didn’t get far enough to find out. In general, I don’t recall anyone getting into much trouble for downloading.
I wonder about that. I mean, did they want “cheap”? They’re valued at $183 billion. Paying 1% of that to lock out their competitors is pretty good value. Even 5 or 10 percent might be. Anthropic just need to make a bit of a show about it being “punishing”, to keep too many people (especially the judge) from noticing that.
Re: Re:
Those valuations are a hilarious, dangerous fucking joke.
Re: Re: Re:
That doesn’t really matter, as long as Anthropic can get $1.5 billion from its investors. Then they can pay off the publishers, effectively booting from the game any would-be competitors who can’t get $1.5 billion each.
Re:
Copyright is not one of them.
Re: Re:
True, because despite the name, copyright is not a right. It’s a government-granted monopoly, a privilege that comes with certain limitations (notably “fair use”, and the similar ideas in other countries).
Re: Re: Re:
Eh, it depends on how you’re using the term. U.S. law explicitly calls it a right (or rights, plural). Essentially, it’s a form of property rights. It’s the correct legal term, at least in the U.S.
Regardless of what you want to call it though, it’s not something that is lost when a work is published, unless they’re confusing it with public domain. Which these works are not, and isn’t applicable here.
Re:
It’s not burglary if the door was unlocked.
If a child borrows a book from a friend (let’s just say it’s a famous one like… oh, the Lord of the Rings, for the purpose of argument), actually heck, let’s say he’s a brazen little thief and steals the Lord of the Rings from Barnes & Noble, reads it, and then goes on to become a well known published fantasy author, can he now be sued for copyright infringement?
His stories are all about elves and dwarves and wizards and dragons, but there’s nothing straight ripped line for line from LOTR. He was however, essentially, trained on a pirated dataset.
Re:
Which is why the actual copyright infringement of downloading and storing the works, unrelated to the training of the LLMs, was the only actionable part of the lawsuit. But the
authorslawyers are using it as a proxy to attack LLM training despite not getting traction on that topicRe:
Probably. Who’d even heard of a “sua sponte” dismissal before Trump’s recent lawsuit was dismissed by the judge? In almost all other cases, anyone can be sued for anything.
Years ago, I started reading “The Sword of Shannara” by Terry Brooks. I found it to be little more than a Lord of the Rings re-telling, with the names changed. Terry was “ripping” from LOTR as much as these “A.I.” services are “ripping”, and got away with it.
Re:
No, but he could be tried as a thief (a literal thief!) if the statute of limitations has not been exceeded.
Re:
I’m beginning to realise that the reason some people get fooled by a thing that does not think is that they have no first-hand experience of the activity.
Authors who didn’t sell out to publishers will get paid.
I may be in a for a bit of a windfall here.
AI slop took over the internet because authors who created quality works weren’t properly compensated.
Although ChatGPT is named in the article, it doesn’t get trained the way other LLMs might be. The conversation I had with it clarifying this follows:
Re:
You’re using an old model or something. It can search the live internet now. Also, you can’t trust it’s claims on how it is trained. It might be hallucinating details.
Re: Re:
TIL: ChatGPT 4 is an old model.
Re: Re:
According to the definition of hallucinations, only animals can have them, making you far more likely to be hallucinating than any AI.
Re:
“I asked the lying machine to tell me how it works and it said it received the truth graven on tablets of gold from the hands of the Almighty.”
Re:
Actually, ChatGPT can access the internet, but it’s limited in how it does so. Here’s the conversation I just had with it: