AI industry horrified to face largest copyright class action ever certified
- Thread starter JournalBot
- Start date
Not all AI. Primarily just the useless LLM shit. And frankly, if that needs to steal literally everything from everyone without compensation to be viable, then it deserves to die.
Artificial Intelligence was a thing that existed well before the LLM fad perverted the term and will continue well after it dies its most deserved death.
Upvote
30
(32
/
-2)
According to the people who do LLM training, you are wrong:
Language Modeling Is Compression, Delétang et al., 2023
LLMs are secretly lossless text compressor and how to use it like one, Liu, 2024
And then, there is the evidence: If all that the LLMs knew was rules about how words interact, then the odds of reproducing a given paragraph from a specific text would be very, very close to zero; you could try a million times and not have it happen. And yet, LLMs are notorious for reproducing large chunks of text exactly as it was written (the "memorization" problem).
Specifically what creativity does a random number machine bring to the table?
Except that it does.
Djiré, Albérick Euraste, et al. "Memorization or Interpolation? Detecting LLM Memorization through Input Perturbation Analysis." arXiv preprint arXiv:2505.03019 (2025).
Hartmann, Valentin, et al. "Sok: Memorization in general-purpose large language models." arXiv preprint arXiv:2310.18362 (2023).
Kassem, Aly M., et al. "Alpaca against vicuna: Using llms to uncover memorization of llms." arXiv preprint arXiv:2403.04801 (2024).
Kiyomaru, Hirokazu, et al. "A comprehensive analysis of memorization in large language models." Proceedings of the 17th International Natural Language Generation Conference. 2024.
Schwarzschild, Avi, et al. "Rethinking llm memorization through the lens of adversarial compression." Advances in Neural Information Processing Systems 37 (2024): 56244-56267.
Satvaty, Ali, Suzan Verberne, and Fatih Turkmen. "Undesirable memorization in large language models: A survey." arXiv preprint arXiv:2410.02650 (2024).
Xiong, Alexander, et al. "The Landscape of Memorization in LLMs: Mechanisms, Measurement, and Mitigation." arXiv preprint arXiv:2507.05578 (2025).
Upvote
41
(42
/
-1)
Love to see this go through. Copyright holders need fair compensation for their work and if AI companies tank ... so be it.
Upvote
21
(22
/
-1)
I don't think this suit would ruin the development of LLMs as we know it. It may tank the segment of the AI market that centers exclusively on producing AI slop from derived works, but AI has far better applications than that.
Upvote
12
(13
/
-1)
The Orange Guy would get involved if AI companies stole HIS original ideas …. Oh, never mind.
Upvote
10
(10
/
0)
Not to be a jerk, but no teachers do not. That is literally the first exception under US Fair Use: https://www.law.cornell.edu/uscode/text/17/107
Upvote
1
(9
/
-8)
Darn it. Sorry, you got that first before my comment. What I get for commenting so late.
Upvote
4
(7
/
-3)
Upvote
2
(3
/
-1)
As has been pointed out before, that is not exactly right.
If the material being copied is strictly for in-class use and is pure research, then it is almost certainly fair use.
But if the material being copied is for public use by the class (e.g., a play or song), then it is not fair use.
And if the material being copied is from an existing text book, then it is not fair use.
Both teachers and school districts have been sued for copyright violations.
Upvote
15
(16
/
-1)
which should wake people up to whether they themselves should give respect to any concept of copyright,because this is the repeating story of forever
Upvote
5
(6
/
-1)
I think Anthropic is fully aware that they can "win" this litigation by getting the class action status dismissed. Instead of one case representing seven million claimants, they force the matter to be settled by seven million individual cases. These don't happen all at once - they are dribbled through the court system for decades, limiting how quickly Anthropic has to pay penalties, increasing the individual cost to bring them to court, and depriving all of us of any judgements broader than the specific case.
In other words (if my memory serves) they are taking a page from DuPont's playbook shirking responsibility for PFOAs in teflon manufacturing (see the career of Robert Billott trying to hold them accountable), which followed a similar strategy involving bad-faith settlements.
In other words (if my memory serves) they are taking a page from DuPont's playbook shirking responsibility for PFOAs in teflon manufacturing (see the career of Robert Billott trying to hold them accountable), which followed a similar strategy involving bad-faith settlements.
Upvote
14
(14
/
0)
Upvote
10
(10
/
0)
I've never heard of too big to fail bef... Oh.. oh no, well hopefully the new TARP might be called FART, so at least it might be entertaining edit 4 spolling
Upvote
4
(4
/
0)
“We are too rich to face consequences or lose any of our wealth. That’s only for the little people.”
Upvote
20
(20
/
0)
I think everything needs to be from a physical book, a copy of which being distributed to every student for free. Licensing should be attached to the physical item. Honestly the time for allowing individuals control over curricula is over. I've seen teachers using handouts for different grade levels from who they were teaching, and who knows what else. That was waaay before the present mess we're in.
Upvote
-15
(1
/
-16)
Upvote
16
(16
/
0)
Upvote
2
(2
/
0)
Most AI cannot exist without breaking copyright/DMCA-like laws. If we're lucky, it might be the first time copyright laws actually help society,
Upvote
5
(7
/
-2)
I've said from the beginning, anything created with AI needs to be public domain, and it can't simply be "I made this with AI, and tweaked it after, now it's mine". If AI touched whatever you created, you lose the benefits of copyright protection.
There might be some nuance in there, but it's hard to inject any nuance into something like this without it being taken advantage of.
There might be some nuance in there, but it's hard to inject any nuance into something like this without it being taken advantage of.
Upvote
21
(22
/
-1)
It's been like this for a part of the population for most of the history of the USA.
The Warren court was the anomaly.
Upvote
13
(13
/
0)
These applications don't require any novels / movies / paintings / whatever to train the algorithm, so I really don't see what point you're trying to make.
Upvote
21
(21
/
0)
Did the plaintiffs use copyright material without consent?
Yes.
Case closed.
I really don't care if the AI companies go bankrupt, it serves them right.
I really don't care if it leads to millions of tiny copyright actions in court, that's a result of the plaintiff's stupid actions, so they should take ownership of it, not suggest it's someone else's problem.
I really do hope that the AI companies lose bigglies. Only then will a robust law protecting authors' works be able to move forward.
The same with any other medium AI companies believe are ripe for exploitation.
Would it be a loss? Yes, there are lots of uses for AI dessimination of all kinds of things, but those must clearly remain tools, not replace that which they've disseminated.
For example, as a would-be writer, i use an online tool to correct my grammar and spelling and make critical comments on aspects of my writing that need improvement. It might even suggest rewording paragraphs to improve readability (it often gets it hopelessly wrong btw). It also compares my writing to the works of published authors, to see if there any comparisons. It might say Stephen King's work uses 54% of whatever my work uses, making the suggestion that perhaps emulation might improve the chances of my book being accepted.
It's a silly comparison in many ways, but it does give you an idea about your work if it veers far outside what is generally accepted popularist writing. It's a relative mark of your work's saleability.
That's a great use of AI that doesn't impinge on an author's copyright.
But wholesale lifting other works to create "new" works is just plain theft.
Someone mentioned the rich getting richer as if it's a new thing.
Nothing much had changed has it? The rich get greedier and richer, the poor get poorer, until every drop of blood has been drained from their bodies and the whole lot collapses. Because stupid people rule and even stupider people let them.
Let's hope stupid doesn't win this time, but be prepared for the worst.
Yes.
Case closed.
I really don't care if the AI companies go bankrupt, it serves them right.
I really don't care if it leads to millions of tiny copyright actions in court, that's a result of the plaintiff's stupid actions, so they should take ownership of it, not suggest it's someone else's problem.
I really do hope that the AI companies lose bigglies. Only then will a robust law protecting authors' works be able to move forward.
The same with any other medium AI companies believe are ripe for exploitation.
Would it be a loss? Yes, there are lots of uses for AI dessimination of all kinds of things, but those must clearly remain tools, not replace that which they've disseminated.
For example, as a would-be writer, i use an online tool to correct my grammar and spelling and make critical comments on aspects of my writing that need improvement. It might even suggest rewording paragraphs to improve readability (it often gets it hopelessly wrong btw). It also compares my writing to the works of published authors, to see if there any comparisons. It might say Stephen King's work uses 54% of whatever my work uses, making the suggestion that perhaps emulation might improve the chances of my book being accepted.
It's a silly comparison in many ways, but it does give you an idea about your work if it veers far outside what is generally accepted popularist writing. It's a relative mark of your work's saleability.
That's a great use of AI that doesn't impinge on an author's copyright.
But wholesale lifting other works to create "new" works is just plain theft.
Someone mentioned the rich getting richer as if it's a new thing.
Nothing much had changed has it? The rich get greedier and richer, the poor get poorer, until every drop of blood has been drained from their bodies and the whole lot collapses. Because stupid people rule and even stupider people let them.
Let's hope stupid doesn't win this time, but be prepared for the worst.
Last edited:
Upvote
16
(18
/
-2)
Upvote
6
(6
/
0)
Somewhat more importantly, they've also openly discussed their "copyright problem" before. This isn't something they did by mistake thinking it'd be fair use. They violated these copyrights knowingly and need to pay the legal penalty for that. This whole idea that something new should get a pass on following the rules because it's potentially a big deal needs to die in a fire. Follow the rules or GTFO.
Upvote
21
(21
/
0)
The will replace Social Security with crypto if we let them. Who cares if millions of elderly people are left destitute? It's not like they've been paying into the system their entire lives.
Upvote
11
(11
/
0)
Upvote
13
(14
/
-1)
Much more nuanced than that in academic libraries. The source of the content that is being shared is very important. Licensing agreements for that content determine how or even if you can share it. Who is doing the copying is another factor. A student for personal use? OK. A teacher for classroom use as a handout? Check the license.
Don't even think about an audio or video recording for in-class use. Bottom line is you always get permission from somebody in the ownership chain prior to use. No exceptions. Same with e-books. Licensing terms determine what you can do with them.
When I was working in an undergraduate library (early 2000's) we had two 4-drawer filing cabinets full of file folders labeled by course number containing photocopies of magazine articles and book chapters that were on reserve for those classes. Every piece had a barcode and was tracked. They could only be checked out by students in those classes, and often had a 2-3 hour time limit (less chance a student walks out with it). The copies were grainy for a reason- copies made from copies suck.
Those cabinets were retired around 2020 (finally). Despite it being a pita to maintain all that content (from finding the article, getting permission, copying, assembling, cataloging and barcoding, and distributing) I imagine it is a lot more difficult to manage that content in digital formats. Glad I retired before I had to learn that.
Upvote
8
(9
/
-1)
ITT: Tons of people who are absolutely mistaken that this case is about whether LLMs by their very nature violate copyright. Except that this same court case has already adjudicated that it is not, and that the model's use of training data is transformative and fair use (you may disagree, but regards to this case it's a settled issue). Not even the plaintiffs, per the judge in a previous decision a couple months back, alleged otherwise. What Anthropic is being found liable for is a thoroughly conventional copyright issue. They have used the strategy of buying, digitizing, then destroying books--and if they stuck to that, they'd also have been in the clear. Their problem is that they just started outright stealing shit and retaining it. And they are going to lose, because despite any pleading about them being special AI snowflakes, it is, again, a thoroughly conventional copyright issue. This article is purely about how bad/chaotic it will be when they lose because of the issues surrounding class certification. That's all.
Upvote
-18
(5
/
-23)
The case is about who can be a plaintiff so whilst I (and 90% of Ars) agree with the rest of your post, the problem is this just isn't how litigation works. The objection to the class definition here is that in practice, the case will not get to this question because they'll get mired in what "the plaintiffs" means.
You need standing to sue, and the argument is that the definition of the "class" here is so terrible it's really not going to be valid to assume anyone the class says has standing should actually have standing.
The answer is to argue for a much narrower class, and if that's aimed aimed to catch thousands of the $150k awards, it'll do just as much damage. I guess whoever argued for this class in the first place was trying to make a political point and one I agree with, but this is why you don't use litigation and courts to do politics.
We can console ourselves with the idea this will only buy Anthropic time. Once this has been narrowed down a bit then someone can go after OpenAI instead with a stronger case, which is really what needs to happen. The rest of the circus revolves around Altman.
It does not cause a problem for plaintiff or dependent. It causes a problem for the court and anyone else who wants to use that court on an unrelated matter.
Upvote
9
(9
/
0)
If this is your sincere belief, you have been lead astray by a cadre of stupid racists who can’t even spell AI.
If you don’t enforce copyright law, how are you any different from “the Chinese” that you fear? The ones who “don’t value” copyright? What makes your copyright violations ok? Is it maybe, perhaps, just that you’re also a racist who’s afraid of $current_targeted_racial_group because deep down you know how mediocre you and your country folk are? I hope not! I think you’re better than that.
For some additional perspective, consider that you’ve spent the last few months ruining any perceived advantage the US ever had over any country, let alone China (the country that singlehandedly facilitates your day-to-day). America is a failed state that finds itself careening into full throated fascism, which nobody respects. Your military hasn’t had success on the battlefield since 1945, no one is afraid of your hard power anymore. Soft power? Literally evaporated within a week of Trump taking over. Your entire economy is just… fraud. Who cares how many trump bucks the plagiarism machine is worth? The Chinese don’t, I’ll tell you that much.
Upvote
28
(31
/
-3)
Anthropic isn't exactly a company that has shown that it plays nice on the Internet ...
In other words, they have shown many times that they have ill intent. Why should anyone believe otherwise in this case?
In other words, they have shown many times that they have ill intent. Why should anyone believe otherwise in this case?
Upvote
13
(13
/
0)
Upvote
14
(14
/
0)