AI industry horrified to face largest copyright class action ever certified
- Thread starter JournalBot
- Start date
Sure but that isn't what was said. They said when teachers must have a license to use anything when they are "training human students". That is just plain wrong.
Upvote
4
(9
/
-5)
Edit: Ok, the quote is blocked which is fine. Regardless, saying this is theft whether by "AI" companies or individuals is such bullshit. Copyright infringement has been explicitly stated by SCOTUS and multiple other federal courts to not be equivalent to theft.
Upvote
-16
(4
/
-20)
Another distinct possibility is they're paid shills who are prohibited from discussing the case publicly because they're agents of the company in reality, even if not openly so. I don't necessarily think so but it'd also fit the facts so far.
Upvote
20
(21
/
-1)
What's interesting about the appeal argument is it boils down to "we couldn't possibly identify the owners of the training materials and nobody else can either (including lots of creators who have no idea we used their work) therefore it was OK for us to use it". It's circular reasoning.
Assuming the court concludes that AI training is covered by copyright law and that each unauthorised use constitutes a breach, then the task for the court is quite simple. How many distinct works was your AI trained on? That's N. How many licences did you obtain? That's M. Your fine is $150K*(N-M) which you will pay into a court-administered trust fund, from which rights holders will be paid as and when they come forward over time AND you will destroy your existing AI and recreate it using only those M works, plus such other materials for which you obtain licences.
I also haven't seen any discussion of "terms of service" breaches where robots.txt directives are simply ignored by AI crawlers. There has got to be scope in that for a separate class action.
Assuming the court concludes that AI training is covered by copyright law and that each unauthorised use constitutes a breach, then the task for the court is quite simple. How many distinct works was your AI trained on? That's N. How many licences did you obtain? That's M. Your fine is $150K*(N-M) which you will pay into a court-administered trust fund, from which rights holders will be paid as and when they come forward over time AND you will destroy your existing AI and recreate it using only those M works, plus such other materials for which you obtain licences.
I also haven't seen any discussion of "terms of service" breaches where robots.txt directives are simply ignored by AI crawlers. There has got to be scope in that for a separate class action.
Upvote
29
(30
/
-1)
The irony of libertarians who want to disrupt things asking the government to keep them from being distupted.
Upvote
29
(29
/
0)
Upvote
10
(10
/
0)
I don't think AI is going to work well enough to do anything to the labor market, but it does provide a way to steal lots of intellectual property, and unfortunately I think SCOTUS is eventually going to back this massive theft.
Upvote
19
(19
/
0)
Upvote
14
(14
/
0)
Interesting defence argument: because they’ve ripped off every author of anything ever written, sung or drawn, figuring out how to pay them all off is just too hard. So just ignore them while they keep doing that.
Upvote
23
(23
/
0)
Missing Minute
Wise, Aged Ars Veteran
Right, because courts are the exclusive arbiter of the linguistic meaning of theft. You can absolutely call someone a thief for stealing your idea to wear a blue dress to prom.
Upvote
-8
(4
/
-12)
Missing Minute
Wise, Aged Ars Veteran
Where did you get the idea "that not a single user was sued that downloaded CR items without DRM"?
Upvote
9
(12
/
-3)
You know, they pay for engineers, they pay hundreds of billions for compute and energy, but they can't pay anything for the work of authors, musicians, artists, journalists. Put a little energy and capital into tech for licensing and problem solved. ASCAP does a pretty good job of compensating composers despite the volume of music.
Upvote
39
(39
/
0)
In fact, AI models are degenerating, and getting worse, and there does not currently exist a solution to that problem.
https://www.ibm.com/think/topics/model-collapse
https://www.ibm.com/think/topics/catastrophic-forgetting
Among other resources.
There is is not really any logical path to AI improvement going forward. All the extant human-created data has already been stolen and trained on. We're already generating AI slop faster than humans can generate useful data, and those humans will not have any motivation to continue doing so anyway, due to the aforementioned theft of their work product. So from here on, AI models will be trained on their own slop, with predictably terrible results.
Upvote
33
(36
/
-3)
Upvote
-5
(2
/
-7)
Missing Minute
Wise, Aged Ars Veteran
I would argue that making jokes about morals not existing is bad for society.
Upvote
16
(19
/
-3)
All current AI implementations are just enormous averaging machines. They are trained on a set and when queried, they return the average answer from their training. They are useless for creating anything new, they only "create" an average of whatever already exists, that was the work of humans. Unless their input is well curated (by humans), their output will just be more mediocrity, and nothing novel nor useful.
Upvote
19
(21
/
-2)
And I would reply that the disintegration of the concept of morality is what actually harms society, and observing this disintegration -- with humor or without -- is necessary if that decline is ever to be reversed.
Upvote
5
(8
/
-3)
Every statement I read in this article saying how this will doom AI comes off as a tech bro sociopath. OMG think about the lost revenue!
Upvote
25
(25
/
0)
The other side of the coin:
https://www.techdirt.com/2025/08/08...-after-lawyers-panic-about-the-public-domain/
AI got no copyright rights.
https://www.techdirt.com/2025/08/08...-after-lawyers-panic-about-the-public-domain/
AI got no copyright rights.
Upvote
11
(11
/
0)
God forbid a company is sued out of existence for systematic crime. That would hurt all the rich people holding onto the stock! I predict the billionaire bubble boys will get their special IP theft exemption from corrupt American courts.
Upvote
27
(27
/
0)
SubWoofer2
Ars Tribunus Militum
It's quite easy to find copyright holders who could be plaintiffs. Throw a stone, if it hits a writer, ask them. An example is here in Melbourne, Australia, where turns out over 90% of the authors speaking at or members of a SF book discussion group have had their works assimilated by the meta AI borg. . List below. Some have very low sales, mere hundreds of copies. But still they were borged. Australian Society of Authors has asked all affected writers to contact them.
Upvote
24
(24
/
0)
Missing Minute
Wise, Aged Ars Veteran
Okay, let me rephrase my question. Where did you read that the group of people you are claiming did not get sued, did not get sued?
Upvote
11
(11
/
0)
Perhaps if asked to use material instead of stealing it which is what they have already done. I remember working in a college library and copying more then 10 or 15 pages was a violation and that all the professors "reserved class materials" where limited unless they had special premission. I understand what AI wants but if that is the case then you need to pay to use it just like anyone else. They are basically arguing that they broke the law but we needed to break the law because otherwise we could not financially develop our product but now that we broke the law if you hold us responsible we cannot financially recover. That is a crazy, wild and highly non legal argument.
Upvote
20
(20
/
0)
So every day we hear tech-titans auger the end of mankind with this technology - a technology they feed with stolen material and we're expected to get behind the argument that this lawsuit is bad for their business?
Upvote
16
(16
/
0)
Is it really financial gain if they've been bleeding money since day one with no end in sight? /s
Upvote
1
(5
/
-4)
"we'll go bankrupt if we have to pay for all the content we stole"
It's a bold strategy, Cotton. Let's see if it pays off for 'em.
It's a bold strategy, Cotton. Let's see if it pays off for 'em.
Upvote
23
(23
/
0)
I wouldn’t. Children are people. AI is not a person. AI is a part of machines, machines which are built and run by corporations and adults who are culpable for their actions.
It’s not different at all. Both things involve computer data encoding source information. They just happen to involve encoding it in different ways.
No, that does not follow in any way, because humans are not machines and exist in nature. Neural nets are artificial, digital constructs that exist entirely in machines as a way to simulate a facsimile of how a brain works.
It’s not even remotely a stretch. The model is built from the data fed into it. It creates a mass statistical model of all the data fed into it to distill that information down into a smaller form which can then later be decoded by prompting. Once created, the model itself is in effect simply a lossily compressed copy of the training data. Applying compression - lossy or otherwise - does not wash away the copyright. There’s plenty of more conventional lossy compression that uses statistical methods to encode the data. This is not a novel or controversial area, the difference is really just the scale and breadth.
Even if we accept that premise, which is more debatable than your framing would suggest, then it is still incumbent upon them to prove it. If they can’t, then the fact they fed the data into the model at all is all anyone has to go on.
Upvote
25
(26
/
-1)
AI industry should pay a reasonable license fee for the material used witch is still protected by copyright.
Like human beings should buy the books they read and the music that listen. AI needs to pay for the content they use and profit.
Probably a % of their sales should go for this. And it is actually even good for the industry in the long term, as there is a need for humans to continue to produce quality, original content.
Like human beings should buy the books they read and the music that listen. AI needs to pay for the content they use and profit.
Probably a % of their sales should go for this. And it is actually even good for the industry in the long term, as there is a need for humans to continue to produce quality, original content.
Upvote
10
(11
/
-1)
How dare people try to hold them responsible for their actions when they're trying to make money.
Upvote
19
(19
/
0)
AI is very much going to disrupt the labor market, just not the way vendors and boosters think. How do we know? We've been here before with offshoring.
Much like hiring Indian companies to offshore US office work, CEOs across the country are looking at AI as a way to slash their payroll and cement their next bonus check. Also like hiring Indian companies to replace US workers, it's not going to go well. In some cases it will be a spectacular failure. But the CEOs driving this won't care because a) They've already nailed down their gargantuan bonuses and b) The failure will be the problem for the next CEO. Even if they stick around long enough for it to be their problem, they have a golden parachute to make sure that they don't actually suffer from their fuck-up.
How many CEOs did you see actually pay a price for screwing up offshoring?
Upvote
36
(36
/
0)
As the Robber Barons demonstrated, there are lots of ways of having a financial gain even when the company is bleeding money.
Upvote
23
(23
/
0)
Is anyone actually supposed to care that these companies face financial ruin? For how long has it been okay for copyright claims against individuals? Why on earth should these corporations be immune?
Upvote
21
(21
/
0)