Anthropic to pay at least $1.5 billion to authors whose work it knowingly pirated
- Reference: 1757309530
- News link: https://www.theregister.co.uk/2025/09/08/anthropic_settles_author_lawsuit/
- Source link:
News of the settlement emerged late last week in a [1]filing [PDF] in the case filed by three authors - Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson - who claimed that Anthropic illegally used their works.
We’re going to see a lot more of this. AI companies wIll create ‘slush funds’
Anthropic admitted to having bought millions of physical books and then digitizing them. The company also downloaded millions of pirated books from the notorious Library Genesis and Pirate Library Mirror troves of stolen material.
The company nonetheless [2]won part of the case, on grounds that scanning books is fair use and using them to create “transformative works” – the output of an LLM that doesn’t necessarily include excerpts from the books – was also OK. But the decision also found Anthropic broke the law by knowingly ingesting pirated books.
Plaintiffs intended to pursue court action over those pirated works, but the filing details a proposed settlement under which Anthropic will create a $1.5 billion fund which values each pirated book it used for training at $3,000. Anthropic also agreed to destroy the pirated works.
[3]
In the filing, counsel observes that this is the largest ever copyright recovery claim to succeed in the USA and suggest it “will set a precedent of AI companies paying for their use of alleged pirated websites.”
[4]
[5]
This settlement is indeed significant given that several other major AI companies – among them [6]Perplexity AI and [7]OpenAI - face similar suits. It may also set a precedent that matters in Anthropic’s dispute with [8]Reddit over having scraped the forum site’s content to feed into its training corpus.
The filing asks the court to approve the settlement, a request judges rarely overrule.
[9]Boffins detail new method to make neural nets forget private and copyrighted info
[10]Suetopia: Generative AI is a lawsuit waiting to happen to your business
[11]Google admits anticompetitive conduct in Australia, agrees to modest fine
[12]AI data-suckers would have to ask permission first under new bill
“We’re going to see a lot more of this,” according to Daryl Plummer, a distinguished VP analyst at Gartner.
Plummer predicted the AI industry will have to build “slush funds” to handle copyright and other legal claims.
[13]
“Death by AI claims will rise one thousand percent,” he told The Register , suggesting that as people turn to AI for advice and counsel, sometimes with disastrous outcomes, their loved ones will seek recompense.
Anthropic appears not to have commented on the settlement. For what it’s worth, it can cover the cost of the settlement with some of the $13 billion in funding it [14]announced last week. ®
Get our [15]Tech Resources
[1] https://regmedia.co.uk/2025/09/08/pacer_anthropic_settlement.pdf
[2] https://www.theregister.com/2025/06/24/anthropic_book_llm_training_ok/
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_offbeat/legal&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2aL6pN_acdAFAzUHqYwU7fgAAARE&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_offbeat/legal&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aL6pN_acdAFAzUHqYwU7fgAAARE&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[5] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_offbeat/legal&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aL6pN_acdAFAzUHqYwU7fgAAARE&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[6] https://www.theregister.com/2025/08/26/perplexity_asahi_nikkei_lawsuits/
[7] https://www.theregister.com/2024/11/21/new_york_times_lawyers_openai/
[8] https://www.theregister.com/2025/06/05/reddit_sues_anthropic_over_ai/
[9] https://www.theregister.com/2025/09/04/boffins_detail_ai_mind_wipe/
[10] https://www.theregister.com/2025/08/12/genai_lawsuit/
[11] https://www.theregister.com/2025/08/18/asia_tech_news_in_brief/
[12] https://www.theregister.com/2025/07/24/ai_copyright_bill_floated/
[13] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_offbeat/legal&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aL6pN_acdAFAzUHqYwU7fgAAARE&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[14] https://www.anthropic.com/news/anthropic-raises-series-f-at-usd183b-post-money-valuation
[15] https://whitepapers.theregister.com/
For the moment, the seemingly endless supply of morons with more money than sense seem happy to prop it up regardless of what their money is actually being spent on.
Pon Zi will think of a scheme
From the IPO's, obviously.
Bitcoin?
You know what...
I will create a 'slush fund' of my own... I wil put $10 in it and then I'm going to illegally download every movie I want !
Let's see how that goes...
Asinine Idiots...
"transformative works"
Is that transformative in the same way [1]you can make ChatGPT spew out summaries and passages from books ?
[1] https://www.theregister.com/2023/05/03/openai_chatgpt_copyright/
"AI" will get even more expensive
Various providers are already jacking up costs, because those energy- and GPU-devouring monstrosities are becoming insatiable molochs.As are the VCs who want their pound of flesh.
Which will quickly become another case of "too many streaming services costing too much money every month".
Are those employers dreaming of diving through Scrooge McDuck money vaults by replacing all their staff with "AI" going to shell out ever higher amounts for ever less return?
Are those poor sods who talk to "their" GPT as the only friend, or worse, quasi-romantic partner going to keep paying more and more? Or even be able to? There's a breaking point where seeking professional help is more feasible.
I predict that the disconnect between what users are willing to pay for a mediocre assistant who makes boneheaded mistakes more often than not - and will continue to do so - and what these companies need to charge to actually be sustainable companies is going to be when the whole house of cards comes crashing down.
Lawyer: So you don’t deny scanning the books?
Ai: There are worse crimes than burning books. One of them is not scanning them.
Lawyer: Do you think that a jury will believe that you did not appropriate copyrighted materials after scanning them?
Ai: Why, sometimes, I’ve believed as many as six impossible things before boot up.
Lawyer: Wait, are you reciting texts as your deposition answers?
Ai: Yes, but it’s transformative work.
So for clarity....
...they're gonna destroy the pirated books etc, but they aren't going to remove the data they ingested for their models?
So they're basically paying $1.5bn to get away with pirating all the written material they could possibly get their hands on.
Sounds pretty cheap for them. Where's the criminal charges for operating the largest copyright theft operation the world has ever seen, with the entire C-suite charged with the crimes?
Steven R
Re: So for clarity....
The real problem here is the court supporting the claim that ingesting copyrighted material is "fair use". By this reasoning if they had bought the books first then scanned them that would be absolutely hunky dory, would have cost significantly less than the 1.5B and they'd have the same model they do now.
Ingesting *entire* works en-masse where the whole content is retained (even in a derived or encoded form) and all of it is necessary to create the "derived" output is absolutely not what fair use is intended to allow. Quoting a paragraph in a review or a 30s sample is the limit of fair use. If your model can work with the latter then its fair, if it can't then you owe the copyright owners royalties.
Re: So for clarity....
Royalty owners seem to have caved in for nothing (or more likely their useless lawyers caved on their behalf, since said lawyers got a fat payday). A fair settlement would have been that the thieves of Anthropic pay this puny compensation, but also nhave to revert their model to the state it existed before stealing the entire world's collection of literary works.
Anthropic also agreed to destroy the pirated works.
But said nothing about the fact that the bloody statistics are already extracted from the books. What good does deleting the copy do if its effect is not subtracted from the total? Something probably akin to unscrambling an egg...
"a distinguished VP analyst"
That was a very impressive title, until the words "at Gartner" were added . . .
to authors whose work it knowingly pirated
Or just to to authors who know it pirated their work?
Re: to authors whose work it knowingly pirated
To the publishers.
Where the heck are these AI companies going to get the money?