Key points (because I’d lost track of where this particular case was up to): A judge recently ruled in favour of Anthropic’s right to train on books that they’d legally acquired. Basically, they...
Exemplary
Key points (because I’d lost track of where this particular case was up to):
A judge recently ruled in favour of Anthropic’s right to train on books that they’d legally acquired. Basically, they ruled that training isn’t copying, it’s transformative use. That strikes me as a correct and sensible ruling.
In that same ruling, they said that the piracy allegations against Anthropic could go ahead. That also seems very reasonable, because Anthropic’s argument there basically seems to be “we really wanted to pirate every book ever written, because not doing so would have been hard”. My layman’s understanding is that from a legal perspective, this argument is categorised as fucking stupid.
The specific objection to the class action, which is the bit this article is covering, isn’t quite as absurd as it sounds. They’re saying two key things there:
The size of the class action means that any company hit with a suit like this is effectively forced to settle even if they believe they’re in the right, because the downside risk is just too great. This does indeed happen in many other areas of law and is a genuine problem. It seems like a much more justified argument than it initially sounds when it looks like they’re just saying “boo hoo we could be bankrupted if we lose”, but it also seems disingenuous given that their argument for why they were torrenting the books in the first place was laughably weak.
Proving ownership of copyrighted work is a mess, and identifying the members of the class is near impossible. This seems like a genuinely good argument for not allowing copyright class actions, and especially not ones at this scale. Seven million plaintiffs in a suit where the key criterion is proving ownership of each individual work just doesn’t seem like it’s plausible for the courts to handle in a fair and effective way. A decent number of those people won’t ever even know the suit is happening.
Honestly, things so far seem to be playing out more sensibly than I would have expected! I was worried about the training side of things, because that ruling going the other way would have created some absurd precedent even for data handling in non-ML computer systems.
On the piracy side, I kind of hope there’s a way for both sides to lose? Anthropic shouldn’t be allowed to just ignore the law because they feel like it, but copyright law is a fucking mess and I’ve been saying that since the RIAA was bankrupting Napster users a few decades ago. How about billing Anthropic some amount of punitive damages, but creating a precedent that those damages are vaguely sane? Five times the retail value of the work or whatever, rather than $150,000 each for every single $5 paperback or $0.99 mp3.
Class action minutiae, I honestly can’t bring myself to care that much. I think Anthropic’s procedural point is kinda fair, but whatever. System’s already fucked, this is too low on the list to worry about making it meaningfully worse.
I agree most with your last bullet -- copyright and class actions don't mix well. The previous bullet ... I mean, I understand the pragmatic issue here, I guess ... but if 100M people get lung...
I agree most with your last bullet -- copyright and class actions don't mix well.
The previous bullet ... I mean, I understand the pragmatic issue here, I guess ... but if 100M people get lung cancer, I don't see how Big Tobacco gets a pass because "that's just too many people". Prove your case or pay up.
In my eyes that penultimate one is more a problem with the $150,000 per item copyright infringement damages than with the number of counts of infringement here. Seven million counts of...
In my eyes that penultimate one is more a problem with the $150,000 per item copyright infringement damages than with the number of counts of infringement here.
Seven million counts of infringement? Yeah, if there’s a practical way to prosecute that then it’s totally justified for the punishment to be some big number. Hundreds of millions of dollars, sure, maybe even a couple of billion. But probably not a literal trillion dollars, any more than an individual downloading a single album back in the day deserved to be on the hook for a million or more.
It’s the class size interacting with fines that were already set at a ridiculous level to placate other corporate special interests in the past, rather than the class size in and of itself on this one.
I disagree, but I think the cat is out of the bag for older content. Future content will definitely have no-training clauses, if they don't already. Then again, this was a ruling for books....
Basically, they ruled that training isn’t copying, it’s transformative use. That strikes me as a correct and sensible ruling.
I disagree, but I think the cat is out of the bag for older content. Future content will definitely have no-training clauses, if they don't already.
Then again, this was a ruling for books. Disney's lawsuit over visual arts may have a different result. Synthesizing words through an LLM can probably erase any likeness of the writer in the process. Mickey mouse, less so.
Proving ownership of copyrighted work is a mess, and identifying the members of the class is near impossible.
I do sympathize with this. I'm sure anyone trying to trace down an IP's owner would. It's a mess.
But... They designed it like this. Maybe nit a tropic, but follow the money and I'm sure you'll find the same companies who spent decades rigging copyright in their favor. I don't exactly feel sympathetic now that these people are engangled in their own trap and now they want to remove copyright.
Copyright wasn't designed to protect billionaires anyway. It's precisely so a bigger richer company can't just copy an idea and scale faster than the creator. Not without proper compensation. I kind of want both to lose, but I also am not in the camp of "throw out all copyright" either. A reset back to 14+14 would be my dream scenario.
Yeah, it’s a fine balance of different interests to consider and it’s not clear cut from an ethical perspective. I do actually see it as morally and conceptually transformative, but I can entirely...
Yeah, it’s a fine balance of different interests to consider and it’s not clear cut from an ethical perspective. I do actually see it as morally and conceptually transformative, but I can entirely see how people would disagree.
My much bigger concern on that one is practical: I can’t see a way to enforce the ruling that training is copying in this case without interacting with existing laws and precedents to basically make computers illegal. Which they obviously can’t do, so it’d end up creating a selectively enforceable mess of a precedent that hinges on whether a given system does or doesn’t feel like it should be called an “AI” by non-expert judges.
I’d be interested if you think there’s a way to construct the opposite ruling to avoid that trap, actually, but I haven’t seen anything that manages to thread that needle yet.
And for complete clarity here, I think by far the biggest issue is that copyright law absolutely did not anticipate questions like this (how could it have?!) and it fundamentally needs reform. Has done for a long while, really, because it’s not even properly equipped to handle digital copying in general.
On the second point, yeah, we’re in this mess in the first place because of vested interests writing the laws. Kinda why I ended up at “they’ve probably got a point, but eh, the whole things a mess so fuck it”, to be honest!
Strong agree on returning copyright terms to something close to their original intent, that’s always sounded to me like one of the easiest and most effective fixes for the problem it’s become. The original US wording with the reminder about why it exists is great, too: it ain’t about a creator’s “right” to make money, that’s just a side effect of encouraging work that’ll benefit society.
The first point is laughable imo. It's like they're saying, "Our entire business relies on copyright infringement, but it's unfair that we should go bankrupt for those infringements!". Like, what?
The first point is laughable imo. It's like they're saying, "Our entire business relies on copyright infringement, but it's unfair that we should go bankrupt for those infringements!". Like, what?
First as in the training vs copying bit, or the first point about the class action? My concern with the latter is about the punishment fitting the crime - it's the same root problem as the...
First as in the training vs copying bit, or the first point about the class action?
My concern with the latter is about the punishment fitting the crime - it's the same root problem as the criminal courts basically coercing people to take plea bargains by threatening to absolutely ruin the defendant's life if they're found guilty. There's a tipping point where if the plea bargain is decent and the possible punishment otherwise is sufficiently draconian, even a lot of completely innocent people will fold because they just can't risk the consequences of losing.
A very large fine seems justified - maybe even one large enough to bankrupt Anthropic, in this case. But allowing the plaintiffs to speak unilaterally for every author in the class gives 7,000,000 * $150,000 = $1.05e+12, i.e. a trillion dollars, and when the calculator is returning the possible fine in scientific notation you know things are pretty wild.
It's basically an instakill for any company on the planet, and while I think Anthropic's piracy seems pretty clear here, I don't think it warrants giving copyright holders a financial doomsday device that they can use to coerce companies in more nuanced situations to settle.
You have a point, a flat 150k each is probably overkill. There's a strong argument that copyright law as a whole was already archaic even prior to AI, but I don't really see that as grounds for...
You have a point, a flat 150k each is probably overkill. There's a strong argument that copyright law as a whole was already archaic even prior to AI, but I don't really see that as grounds for special treatment in this case. At the end of the day, if you stole from practically every author out there, they're all owed compensation.
There should probably be tiers for payouts or something, but even if you dropped it down to 25k each across the board, that's still almost 200 billion. For anyone but the biggest players, that's automatic bankruptcy.
Yeah, I definitely don’t see grounds for special treatment here, and I’d much prefer broader reforms. That’s why my preferred scenario is both losing: Anthropic is found liable, but copyright...
Yeah, I definitely don’t see grounds for special treatment here, and I’d much prefer broader reforms. That’s why my preferred scenario is both losing: Anthropic is found liable, but copyright fines are defanged a bit. It can’t correct past wrongs that happened under the $150k per infringement setup, but it can at least prevent future wrongs.
And for what it’s worth, I do think $25,000 would still be vastly too high - it’s easy to be anchored by the current $150,000 and go down from there, but I honestly think it’s so far from reasonable that it just needs to be thrown out entirely.
What would you really consider fair for torrenting a movie or downloading a song? $100 maybe? $500? I think stretching above $1,000 is already getting pretty out of hand. We should be in speeding ticket territory, not losing your home territory. It certainly shouldn’t be more than the fine for shoplifting the same creative work, that’s for sure!
Take a number somewhere in that range, throw a 2x or 3x multiplier on it (and/or a revenue-based modifier) because this was commercial infringement for profit rather than an individual doing it for personal use, and you get somewhere reasonable. Hundreds of millions to a few billion, in this case.
If you ask me, take the price of the item in question. Then multiply with a fudge factor to ensure it discourages piracy. In the case of public transport here, that factor is x3 or x10, roughly,...
What would you really consider fair for torrenting a movie or downloading a song? $100 maybe? $500? I think stretching above $1,000 is already getting pretty out of hand. We should be in speeding ticket territory, not losing your home territory. It certainly shouldn’t be more than the fine for shoplifting the same creative work, that’s for sure!
If you ask me, take the price of the item in question. Then multiply with a fudge factor to ensure it discourages piracy. In the case of public transport here, that factor is x3 or x10, roughly, just to give a real world example. That sounds a lot more reasonable than the x10000 implied by the 150k fine.
I also think we might not even have this problem if relevant quantities of these works could be licensed in bulk at reasonable rates.
I'm so much more pessimistic that any special treatment anthropic gets would set a precedent - which I'm not sure works anyway since I believe the penalties are written in these laws - that would...
I'm so much more pessimistic that any special treatment anthropic gets would set a precedent - which I'm not sure works anyway since I believe the penalties are written in these laws - that would apply to the average person instead of them just getting special treatment.
Individuals don't deserve those high fines but companies absolutely do.
Practically speaking you’re probably right. I’m getting caught up in the abstract legal debate when the reality is very much that billionaires are gonna billionaire to the detriment of the rest of...
Practically speaking you’re probably right. I’m getting caught up in the abstract legal debate when the reality is very much that billionaires are gonna billionaire to the detriment of the rest of us.
Only thing I really feel the need to stick by is the magnitude of the fines: using my suggestion of $500 per infringement and triple damages because it’s commercial, that still gives a fine larger than every penny Anthropic has ever made. Not just their profits, not even “just” all of their revenue, literally every single dollar that’s ever crossed their bank account - investments, customer payments, corporate deals, everything. And that’s using a fine set at 1% of what’s currently being applied to corporations and to individuals.
I would argue that they shouldn't have relied on that copyright infringement for their business then. You're basically arguing that the crime is too big to hold them accountable for. Especially...
I would argue that they shouldn't have relied on that copyright infringement for their business then. You're basically arguing that the crime is too big to hold them accountable for. Especially because it's not as if they don't believe in copyright and patent and trademark and all of those protections for themselves, they just don't believe in it for the people whose works they used without consent.
I do feel entirely differently about this when it involves individuals, But I'm a lot more empathetic to authors and other creators that I was as a teenager downloading music too. But don't believe that these companies have a right to (try to) profit off of the backs of everyone else who has ever done art in the history of the world. If they were operating on whatever the equivalent of a creative Commons license was for an llm, at least they would be closer to morally consistent
There are about three prongs to my hatred for AI such as it is and this is one of them.
Besides, I'm not convinced the money spigot turns off even with this kind of fine. Frankly, everyone just seems to keep giving these companies more. But if not, bankruptcy exists for a reason. I don't think they have a right to continue to exist just because that would be a really big fine. However, it probably would take a statutory fix, not just a judgment.
But if $500 infringement does the same thing sure. Only if it absolutely wipes them out though. Otherwise they're just getting special treatment or financed by their billionaires
I’m probably not making my point very well here. As far as I’m concerned, if this bankrupts Anthropic, that’s totally justified. Like you said, they shouldn’t have relied on copyright infringement...
I’m probably not making my point very well here. As far as I’m concerned, if this bankrupts Anthropic, that’s totally justified. Like you said, they shouldn’t have relied on copyright infringement and it’s not like there was anything preventing them from just buying the damn books in the first place.
I’m saying a $500 per charge fine is reasonable specifically because it’s a tethered to reality and it’s still more than large enough to wipe out Anthropic if they were liable on all counts.
Bankrupting them for infringement on this scale is arguably fair. A very very large fine, comprised of a sensible amount of a few hundred dollars per instance for an enormous number of instances is plenty to do that.
Copyright holders having the power to threaten what isn’t just a very very large fine but is, effectively, an infinity dollar fine is the bit that bothers me - because it destroys all proportionality. If the fine is so insanely large that literally 99% of the charges could be thrown out and the remaining 1% would still bankrupt a major company, that’s a system-breaking problem.
And I care about this because I still remember what the fuckers did to the free flow of information on the internet. I remember when they made a 16 character string of hex digits illegal. I remember when they drove Aaron Swartz to suicide by throwing around exactly these kind of insane, life destroying, “throw out 99% and we don’t care, you’re still ruined if we can make even 1% stick” charges.
Nothing I’ve said anywhere in this thread is because I think Anthropic deserves to get off lightly. I just can’t in good conscience support the laws their opponents are using either, after all the harm they’ve done to people who absolutely did not deserve it, and all the harm they’ve done to broader society in the name of enforcement infrastructure.
Oh like I said, I feel entirely different about this applying to individuals, I just think it's annoying that it only gets highlighted as a problem anymore when it hits a popular company, not when...
Oh like I said, I feel entirely different about this applying to individuals, I just think it's annoying that it only gets highlighted as a problem anymore when it hits a popular company, not when it's impacting those individuals.
Like I agree the penalty is too high. I just don't support changing it in this instance because it'd be special treatment, not a statutory change. If I had a stronger principled stance on it I'd say it should change on that now and forever. But I think corporations being hit with the fines that devastated people since the Napster era is karma. And I just can bring myself to believe that the billionaires won't just give more money or pull corporate shenanigans to never repay the debt. So meh. I acknowledge the lack of principles on this one. (And it's not the same people who destroyed the internet info flow, books have never really before been a target and were far safer to torrent for that reason. Authors should get paid, they already get screwed. )
I'm not optimistic, but I do enjoy seeing the "AI" industry horrified.
AI industry groups are urging an appeals court to block what they say is the largest copyright class action ever certified. They've warned that a single lawsuit raised by three authors over Anthropic's AI training now threatens to "financially ruin" the entire AI industry if up to 7 million claimants end up joining the litigation and forcing a settlement.
Last week, Anthropic petitioned to appeal the class certification, urging the court to weigh questions that the district court judge, William Alsup, seemingly did not. Alsup allegedly failed to conduct a "rigorous analysis" of the potential class and instead based his judgment on his "50 years" of experience, Anthropic said.
If the appeals court denies the petition, Anthropic argued, the emerging company may be doomed. As Anthropic argued, it now "faces hundreds of billions of dollars in potential damages liability at trial in four months" based on a class certification rushed at "warp speed" that involves "up to seven million potential claimants, whose works span a century of publishing history," each possibly triggering a $150,000 fine.
I'm not optimistic, but I do enjoy seeing the "AI" industry horrified.
Dear Court We made our business model function on the exploitation of the entire internet's worth of labor for free. We ignored copyright and didn't compensate anyone and while we personally...
Dear Court
We made our business model function on the exploitation of the entire internet's worth of labor for free. We ignored copyright and didn't compensate anyone and while we personally haven't been found to be actively torrenting both porn and books illegally we probably did do that too.
If someone copied our LLM model(s) we would however be very upset about that. We worked so hard and that work shouldn't be stolen. (´;︵;`)
See it's fine because billionaires gave us a lot of money to do it and they're not giving the authors and writers and journalists a lot of money, so we're good though.
See it's fine because billionaires gave us a lot of money to do it and they're not giving the authors and writers and journalists a lot of money, so we're good though.
“Kill one man, and you are a murderer. Kill millions of men, and you are a conqueror. Kill them all, and you are a god." -- Jean Rostand Trying to find a way to bastardize this into stealing...
“Kill one man, and you are a murderer. Kill millions of men, and you are a conqueror. Kill them all, and you are a god."
Are they horrified? I feel like that’s just the journalist anthropomorphizing something that isn’t a human (“the AI industry”). This is an appeal by companies filed to court. These are written by...
Are they horrified? I feel like that’s just the journalist anthropomorphizing something that isn’t a human (“the AI industry”).
This is an appeal by companies filed to court. These are written by lawyers being paid hundreds of dollars an hour to be as convincing as possible. There is no horror or any such emotion, just arguments made by professional arguers.
The lawyers, who wrote this, could not give less of a shit about the inherent result of the case, just that it would both damage their reputation and lose a client if said client lost.
Horrified over losing money. Or worse, compensating the working class after ransacking them. It's click bait-Y, but I think the general idea is that this can have very costly consequences for...
Are they horrified
Horrified over losing money. Or worse, compensating the working class after ransacking them.
It's click bait-Y, but I think the general idea is that this can have very costly consequences for those trying to file an appeal.
One of the largest AI companies going poof because of a copyright class action lawsuit should deeply shake an industry. While I'm sure there'd be whining about a fine of even $10/infringement, the...
One of the largest AI companies going poof because of a copyright class action lawsuit should deeply shake an industry.
While I'm sure there'd be whining about a fine of even $10/infringement, the current risk stretches into the hundreds of billions which is an absolutely insane sum of money. No company has that kind of cash! A fine that large would devastate even the largest companies in human history.
Well, yeah. That should be a good lesson. Trying to rip off your labor is a billion dollar mistake. Trying to rip off a country is a trillion dollar mistake. Even the richest aren't rich enough to...
Well, yeah. That should be a good lesson. Trying to rip off your labor is a billion dollar mistake. Trying to rip off a country is a trillion dollar mistake. Even the richest aren't rich enough to do that as of now.
If a business can't sustain itself, then it either needs to optimize or not exist. No one but themselves compelled them to resort to rampant mass pillages of data with no regard to IP.
Copyright only works becsuse a lot of countries agreed to comply with it, and punish across international lines. China was not one of those countries. The best we could do is sanction infringing...
Copyright only works becsuse a lot of countries agreed to comply with it, and punish across international lines. China was not one of those countries.
The best we could do is sanction infringing rights in other countries. But no one seems willing to really do that. Technically the TikTok ban is still on board (not that it's infringing on anything), but it keeps getting kicked down the road.
I mean, I can buy an Android-based game piracy machine loaded with 50,000 copyrighted works from AliExpress and have it sent right to my door. You're right - nobody is doing anything.
I mean, I can buy an Android-based game piracy machine loaded with 50,000 copyrighted works from AliExpress and have it sent right to my door.
You can. Here is GoPro suing Insta360 for alleged patent violation, for instance. https://petapixel.com/2025/07/11/insta360-infringed-on-gopros-patents-us-trade-commission-judge-finds/
You can. Here is GoPro suing Insta360 for alleged patent violation, for instance.
China and the US both (rightly) see that AI is a potential new frontier. One large enough that falling behind in it can have implications for world superpower status. If American AI companies are...
China and the US both (rightly) see that AI is a potential new frontier. One large enough that falling behind in it can have implications for world superpower status. If American AI companies are not allowed to train on the works out there, but Chinese ones continue to do so, Chinese companies will outpace American ones.
They can be sued, but I somehow doubt the Chinese government would allow for any negative rulings to be enforced against their companies when so much is at stake.
I mean, Open AI can just cancel that $1.5M bonus they gave to each employee and use that to pay back all the people whose content they stole to train their models.
I mean, Open AI can just cancel that $1.5M bonus they gave to each employee and use that to pay back all the people whose content they stole to train their models.
Key points (because I’d lost track of where this particular case was up to):
A judge recently ruled in favour of Anthropic’s right to train on books that they’d legally acquired. Basically, they ruled that training isn’t copying, it’s transformative use. That strikes me as a correct and sensible ruling.
In that same ruling, they said that the piracy allegations against Anthropic could go ahead. That also seems very reasonable, because Anthropic’s argument there basically seems to be “we really wanted to pirate every book ever written, because not doing so would have been hard”. My layman’s understanding is that from a legal perspective, this argument is categorised as fucking stupid.
The specific objection to the class action, which is the bit this article is covering, isn’t quite as absurd as it sounds. They’re saying two key things there:
Honestly, things so far seem to be playing out more sensibly than I would have expected! I was worried about the training side of things, because that ruling going the other way would have created some absurd precedent even for data handling in non-ML computer systems.
On the piracy side, I kind of hope there’s a way for both sides to lose? Anthropic shouldn’t be allowed to just ignore the law because they feel like it, but copyright law is a fucking mess and I’ve been saying that since the RIAA was bankrupting Napster users a few decades ago. How about billing Anthropic some amount of punitive damages, but creating a precedent that those damages are vaguely sane? Five times the retail value of the work or whatever, rather than $150,000 each for every single $5 paperback or $0.99 mp3.
Class action minutiae, I honestly can’t bring myself to care that much. I think Anthropic’s procedural point is kinda fair, but whatever. System’s already fucked, this is too low on the list to worry about making it meaningfully worse.
I agree most with your last bullet -- copyright and class actions don't mix well.
The previous bullet ... I mean, I understand the pragmatic issue here, I guess ... but if 100M people get lung cancer, I don't see how Big Tobacco gets a pass because "that's just too many people". Prove your case or pay up.
In my eyes that penultimate one is more a problem with the $150,000 per item copyright infringement damages than with the number of counts of infringement here.
Seven million counts of infringement? Yeah, if there’s a practical way to prosecute that then it’s totally justified for the punishment to be some big number. Hundreds of millions of dollars, sure, maybe even a couple of billion. But probably not a literal trillion dollars, any more than an individual downloading a single album back in the day deserved to be on the hook for a million or more.
It’s the class size interacting with fines that were already set at a ridiculous level to placate other corporate special interests in the past, rather than the class size in and of itself on this one.
I disagree, but I think the cat is out of the bag for older content. Future content will definitely have no-training clauses, if they don't already.
Then again, this was a ruling for books. Disney's lawsuit over visual arts may have a different result. Synthesizing words through an LLM can probably erase any likeness of the writer in the process. Mickey mouse, less so.
I do sympathize with this. I'm sure anyone trying to trace down an IP's owner would. It's a mess.
But... They designed it like this. Maybe nit a tropic, but follow the money and I'm sure you'll find the same companies who spent decades rigging copyright in their favor. I don't exactly feel sympathetic now that these people are engangled in their own trap and now they want to remove copyright.
Copyright wasn't designed to protect billionaires anyway. It's precisely so a bigger richer company can't just copy an idea and scale faster than the creator. Not without proper compensation. I kind of want both to lose, but I also am not in the camp of "throw out all copyright" either. A reset back to 14+14 would be my dream scenario.
Yeah, it’s a fine balance of different interests to consider and it’s not clear cut from an ethical perspective. I do actually see it as morally and conceptually transformative, but I can entirely see how people would disagree.
My much bigger concern on that one is practical: I can’t see a way to enforce the ruling that training is copying in this case without interacting with existing laws and precedents to basically make computers illegal. Which they obviously can’t do, so it’d end up creating a selectively enforceable mess of a precedent that hinges on whether a given system does or doesn’t feel like it should be called an “AI” by non-expert judges.
I’d be interested if you think there’s a way to construct the opposite ruling to avoid that trap, actually, but I haven’t seen anything that manages to thread that needle yet.
And for complete clarity here, I think by far the biggest issue is that copyright law absolutely did not anticipate questions like this (how could it have?!) and it fundamentally needs reform. Has done for a long while, really, because it’s not even properly equipped to handle digital copying in general.
On the second point, yeah, we’re in this mess in the first place because of vested interests writing the laws. Kinda why I ended up at “they’ve probably got a point, but eh, the whole things a mess so fuck it”, to be honest!
Strong agree on returning copyright terms to something close to their original intent, that’s always sounded to me like one of the easiest and most effective fixes for the problem it’s become. The original US wording with the reminder about why it exists is great, too: it ain’t about a creator’s “right” to make money, that’s just a side effect of encouraging work that’ll benefit society.
The first point is laughable imo. It's like they're saying, "Our entire business relies on copyright infringement, but it's unfair that we should go bankrupt for those infringements!". Like, what?
First as in the training vs copying bit, or the first point about the class action?
My concern with the latter is about the punishment fitting the crime - it's the same root problem as the criminal courts basically coercing people to take plea bargains by threatening to absolutely ruin the defendant's life if they're found guilty. There's a tipping point where if the plea bargain is decent and the possible punishment otherwise is sufficiently draconian, even a lot of completely innocent people will fold because they just can't risk the consequences of losing.
A very large fine seems justified - maybe even one large enough to bankrupt Anthropic, in this case. But allowing the plaintiffs to speak unilaterally for every author in the class gives 7,000,000 * $150,000 = $1.05e+12, i.e. a trillion dollars, and when the calculator is returning the possible fine in scientific notation you know things are pretty wild.
It's basically an instakill for any company on the planet, and while I think Anthropic's piracy seems pretty clear here, I don't think it warrants giving copyright holders a financial doomsday device that they can use to coerce companies in more nuanced situations to settle.
You have a point, a flat 150k each is probably overkill. There's a strong argument that copyright law as a whole was already archaic even prior to AI, but I don't really see that as grounds for special treatment in this case. At the end of the day, if you stole from practically every author out there, they're all owed compensation.
There should probably be tiers for payouts or something, but even if you dropped it down to 25k each across the board, that's still almost 200 billion. For anyone but the biggest players, that's automatic bankruptcy.
Yeah, I definitely don’t see grounds for special treatment here, and I’d much prefer broader reforms. That’s why my preferred scenario is both losing: Anthropic is found liable, but copyright fines are defanged a bit. It can’t correct past wrongs that happened under the $150k per infringement setup, but it can at least prevent future wrongs.
And for what it’s worth, I do think $25,000 would still be vastly too high - it’s easy to be anchored by the current $150,000 and go down from there, but I honestly think it’s so far from reasonable that it just needs to be thrown out entirely.
What would you really consider fair for torrenting a movie or downloading a song? $100 maybe? $500? I think stretching above $1,000 is already getting pretty out of hand. We should be in speeding ticket territory, not losing your home territory. It certainly shouldn’t be more than the fine for shoplifting the same creative work, that’s for sure!
Take a number somewhere in that range, throw a 2x or 3x multiplier on it (and/or a revenue-based modifier) because this was commercial infringement for profit rather than an individual doing it for personal use, and you get somewhere reasonable. Hundreds of millions to a few billion, in this case.
If you ask me, take the price of the item in question. Then multiply with a fudge factor to ensure it discourages piracy. In the case of public transport here, that factor is x3 or x10, roughly, just to give a real world example. That sounds a lot more reasonable than the x10000 implied by the 150k fine.
I also think we might not even have this problem if relevant quantities of these works could be licensed in bulk at reasonable rates.
I'm so much more pessimistic that any special treatment anthropic gets would set a precedent - which I'm not sure works anyway since I believe the penalties are written in these laws - that would apply to the average person instead of them just getting special treatment.
Individuals don't deserve those high fines but companies absolutely do.
Practically speaking you’re probably right. I’m getting caught up in the abstract legal debate when the reality is very much that billionaires are gonna billionaire to the detriment of the rest of us.
Only thing I really feel the need to stick by is the magnitude of the fines: using my suggestion of $500 per infringement and triple damages because it’s commercial, that still gives a fine larger than every penny Anthropic has ever made. Not just their profits, not even “just” all of their revenue, literally every single dollar that’s ever crossed their bank account - investments, customer payments, corporate deals, everything. And that’s using a fine set at 1% of what’s currently being applied to corporations and to individuals.
I would argue that they shouldn't have relied on that copyright infringement for their business then. You're basically arguing that the crime is too big to hold them accountable for. Especially because it's not as if they don't believe in copyright and patent and trademark and all of those protections for themselves, they just don't believe in it for the people whose works they used without consent.
I do feel entirely differently about this when it involves individuals, But I'm a lot more empathetic to authors and other creators that I was as a teenager downloading music too. But don't believe that these companies have a right to (try to) profit off of the backs of everyone else who has ever done art in the history of the world. If they were operating on whatever the equivalent of a creative Commons license was for an llm, at least they would be closer to morally consistent
There are about three prongs to my hatred for AI such as it is and this is one of them.
Besides, I'm not convinced the money spigot turns off even with this kind of fine. Frankly, everyone just seems to keep giving these companies more. But if not, bankruptcy exists for a reason. I don't think they have a right to continue to exist just because that would be a really big fine. However, it probably would take a statutory fix, not just a judgment.
But if $500 infringement does the same thing sure. Only if it absolutely wipes them out though. Otherwise they're just getting special treatment or financed by their billionaires
I’m probably not making my point very well here. As far as I’m concerned, if this bankrupts Anthropic, that’s totally justified. Like you said, they shouldn’t have relied on copyright infringement and it’s not like there was anything preventing them from just buying the damn books in the first place.
I’m saying a $500 per charge fine is reasonable specifically because it’s a tethered to reality and it’s still more than large enough to wipe out Anthropic if they were liable on all counts.
Bankrupting them for infringement on this scale is arguably fair. A very very large fine, comprised of a sensible amount of a few hundred dollars per instance for an enormous number of instances is plenty to do that.
Copyright holders having the power to threaten what isn’t just a very very large fine but is, effectively, an infinity dollar fine is the bit that bothers me - because it destroys all proportionality. If the fine is so insanely large that literally 99% of the charges could be thrown out and the remaining 1% would still bankrupt a major company, that’s a system-breaking problem.
And I care about this because I still remember what the fuckers did to the free flow of information on the internet. I remember when they made a 16 character string of hex digits illegal. I remember when they drove Aaron Swartz to suicide by throwing around exactly these kind of insane, life destroying, “throw out 99% and we don’t care, you’re still ruined if we can make even 1% stick” charges.
Nothing I’ve said anywhere in this thread is because I think Anthropic deserves to get off lightly. I just can’t in good conscience support the laws their opponents are using either, after all the harm they’ve done to people who absolutely did not deserve it, and all the harm they’ve done to broader society in the name of enforcement infrastructure.
Oh like I said, I feel entirely different about this applying to individuals, I just think it's annoying that it only gets highlighted as a problem anymore when it hits a popular company, not when it's impacting those individuals.
Like I agree the penalty is too high. I just don't support changing it in this instance because it'd be special treatment, not a statutory change. If I had a stronger principled stance on it I'd say it should change on that now and forever. But I think corporations being hit with the fines that devastated people since the Napster era is karma. And I just can bring myself to believe that the billionaires won't just give more money or pull corporate shenanigans to never repay the debt. So meh. I acknowledge the lack of principles on this one. (And it's not the same people who destroyed the internet info flow, books have never really before been a target and were far safer to torrent for that reason. Authors should get paid, they already get screwed. )
I'm not optimistic, but I do enjoy seeing the "AI" industry horrified.
Dear Court
We made our business model function on the exploitation of the entire internet's worth of labor for free. We ignored copyright and didn't compensate anyone and while we personally haven't been found to be actively torrenting both porn and books illegally we probably did do that too.
If someone copied our LLM model(s) we would however be very upset about that. We worked so hard and that work shouldn't be stolen. (´;︵;`)
Did we mention we are rich?
Thank you.
"you wouldn't steal a book" - that's illegal
"No but we're gonna steal ALL the books" - this is an emerging sector
See it's fine because billionaires gave us a lot of money to do it and they're not giving the authors and writers and journalists a lot of money, so we're good though.
Pffft, giving money to journalists and writers seems like a commie thing and hence unpatriotic /s
Their inherent poverty proves they're wrong...Default judgement please
〈(•ˇ‿ˇ•)-→We win
“Kill one man, and you are a murderer. Kill millions of men, and you are a conqueror. Kill them all, and you are a god."
-- Jean Rostand
Trying to find a way to bastardize this into stealing books, but it's just not coming to me.
In other news ... TIL this isn't just a Megadeth song.
Don’t you mean Metallica? They’re the ones who sued their fans at one point afaik.
Captive Honor
No I literally mean Megadeth. It's in one of their songs.
Edit: Oh, yeah ... That one.👆
Oh you meant the quote
Granted, and here's $100m for your troubles
Are they horrified? I feel like that’s just the journalist anthropomorphizing something that isn’t a human (“the AI industry”).
This is an appeal by companies filed to court. These are written by lawyers being paid hundreds of dollars an hour to be as convincing as possible. There is no horror or any such emotion, just arguments made by professional arguers.
The lawyers, who wrote this, could not give less of a shit about the inherent result of the case, just that it would both damage their reputation and lose a client if said client lost.
Horrified over losing money. Or worse, compensating the working class after ransacking them.
It's click bait-Y, but I think the general idea is that this can have very costly consequences for those trying to file an appeal.
One of the largest AI companies going poof because of a copyright class action lawsuit should deeply shake an industry.
While I'm sure there'd be whining about a fine of even $10/infringement, the current risk stretches into the hundreds of billions which is an absolutely insane sum of money. No company has that kind of cash! A fine that large would devastate even the largest companies in human history.
Well, yeah. That should be a good lesson. Trying to rip off your labor is a billion dollar mistake. Trying to rip off a country is a trillion dollar mistake. Even the richest aren't rich enough to do that as of now.
If a business can't sustain itself, then it either needs to optimize or not exist. No one but themselves compelled them to resort to rampant mass pillages of data with no regard to IP.
Hey now, don't threaten me with a good time
If only we could sue (and win against) Chinese companies as well.
Copyright only works becsuse a lot of countries agreed to comply with it, and punish across international lines. China was not one of those countries.
The best we could do is sanction infringing rights in other countries. But no one seems willing to really do that. Technically the TikTok ban is still on board (not that it's infringing on anything), but it keeps getting kicked down the road.
I mean, I can buy an Android-based game piracy machine loaded with 50,000 copyrighted works from AliExpress and have it sent right to my door.
You're right - nobody is doing anything.
If only all these countries that subscribe so deeply to copyright law could do the same for taxing the rich.
You can. Here is GoPro suing Insta360 for alleged patent violation, for instance.
https://petapixel.com/2025/07/11/insta360-infringed-on-gopros-patents-us-trade-commission-judge-finds/
China and the US both (rightly) see that AI is a potential new frontier. One large enough that falling behind in it can have implications for world superpower status. If American AI companies are not allowed to train on the works out there, but Chinese ones continue to do so, Chinese companies will outpace American ones.
They can be sued, but I somehow doubt the Chinese government would allow for any negative rulings to be enforced against their companies when so much is at stake.
I mean, Open AI can just cancel that $1.5M bonus they gave to each employee and use that to pay back all the people whose content they stole to train their models.