The article is just restating US copyright law for people who don't know it.
The title indicates a recent change in policy, and there hasn't been one.
Now, this is US law. Much of the world does have "sweat of the brow" copyright.
But the US does not. See "Feist vs. Rural Telephone". There must be human creativity because the US constitution reads "(Congress shall have the power) To promote the Progress of Science and useful Arts, by securing for limited Times to Authors and Inventors the exclusive Right to their respective Writings and Discoveries." That limits the power of Congress to change this.
So a lot of things are uncopyrightable in the US. Feist is the big case, and it was about telephone directories. No copyright.
Useful objects are not copyrightable, which is why there are third party auto parts.
Facts are not copyrightable, nor are databases of facts. Nor are pictures of public domain art.
Nor is automotive sheet metal, despite lobbying by automakers.
That's what leads to AI-generated content not being copyrightable. Model weights don't seem to be copyrightable either.
It cannot be this clean cut or Google v Oracle would have been trivial to decide.
A telephone directory can be copyright protected, too. The facts within it cannot. Exactly like recipes. A particular presentation of a recipe is absolutely copyright protected. But this doesn’t mean no one can publish the same recipe in their own presentation
Hah, this triggers my pet peeve about IP for software: Copyright for code is a category error.
Copyright very explicitly does NOT cover functional aspects, only creative expression. But the value of software is not in the expressive aspects of the code that generates it, the value is in the functionality it provides. And this is the exact thing that copyright does NOT protect!
Like, literally, we're protecting the fact that we used a for-loop instead of a while-loop to do a useful thing, not the useful thing itself. This is why rewriting enough code is sufficient to bypass all copyright protection, as the recent AI rewrites have shown.
This was always a huge legislative hack, rightfully controversial at the time it happened, but legislators didn't want to bother figuring out a proper IP regime for software -- maybe something between copyright and patents -- so they shoehorned code into copyright.
Google v Oracle was trivial to decide, but it got decided the wrong way. Because it became a trial around laws based on an ill-founded legislative hack and the decades of flawed case-law it spawned. Stepping back, it is very clear that Sun spent billions making a very useful thing and Google co-opted it for its own uses without recompense because they bet (correctly) that they could legally get away with it.
It is somewhat analogous to what the AI companies are doing with all the code and data out there. Funny that people are unhappy about THAT but cheered when Google did it to Sun / Oracle. I wonder what changed ;-)
I think they're referring to more fundamental things like math, language, etc...
A recipe book can have copyright because it's for the specific combination of recipes, layout, etc... Not for a recipe or more generally a recipe book.
I think that's consistent with his the supreme court ruled in Google v Oracle.
> The CO has taken the (entirely correct) position that copyright law exists to incentivize human acts of creative expression.
The US Constitution grants Congress the power to grant copyrights "to promote the progress of Science and the useful Arts." Creative expression that is not useful therefore is not constitutionally copyrightable.
But that aside, I agree with the author that the Copyright Office refusing to further expand the already vastly over-expanded scope of copyrights is a rare positive ruling.
You have left out the second half of the sentence, "... by securing for limited Times to Authors and Inventors the exclusive Right to their respective Writings and Discoveries."
The position taken by the courts and the administration is that "authors and inventors" means "human". The monkey could not be the author of its own selfie photograph, for instance.
Maybe this doesn't reflect the intentions of the Framers, but it certainly should color any modern discussion surrounding the purpose of copyright law.
I also notice it doesn't say anything about their heirs. I wonder if there is a meaningful argument that copyright extending beyond the life of the original author is unconstitutional.
Are you just arguing that corporations should not be able to own copyrights but it is OK for them to own other things such as land and equipment, or are you arguing that corporations should not be allowed to own anything?
"Intellectual Property" is not, or at least should not, be the same as physical property. Copyrights and patents are temporary artificial, government enforced monopolies to incentivise creativity and publication of the results of that creativity. I don't think it's unreasonable to say that a corporation can't own the copyright itself. Although, they should still be able to hold a license to works from the actual human creators, and granting such a license could be a condition of employment (provided it was created as part of the employment, and not, say, on the employees own time).
"Useful Arts" is old timey talk for the mechanical trades. Engineering, manufacturing and that kind of thing. It applies to "Inventors" and their "Discoveries," and is the basis of patent law.
There is another clause later tha clarifies it a bit. It does NOT mean that an invention must be useful, and it's pretty clear that purely l utilitarian objects actually can not be copyrighted.
> Basically, we're about to recreate formalities, but managed by the private sector. To the extent that companies use AI in their work, they are going to have to track every pixel and every word and every note that is output by an AI system, and which parts were added or edited by humans after creation.
> That means that companies that want to protect their commercial interests aren't just going to get more protection based on how much work they get humans to do; they're also going to have to treat copyright's government-enforceable monopoly over expressive speech with the gravitas it deserves: not something you get automatically, but something you have to do real administrative work for.
Assuming this happens, wouldn't this mean that copyright becomes significantly less useful for people who aren't corporations and don't have the legal, financial, and administrative resources to produce this kind of bookkeeping?
Not if you take option 2, just don't use AI. The companies have to account for where they used AI in order to own the rights, if you didn't use AI, there's nothing to prove. If someone takes your work and you sue them for infringement, and there's no evidence you used AI, then there's no reason to dispute your right to the work
If you put something that is copyrighted through a mechanical transformation process it has the same copyright. For instance a video of a movie retains the copyright status of the movie (as well as probably being an illegal copy itself) even if you added camera shake and audience coughing sounds. So there's not an issue with things like this.
I think the point is to decide where to draw the line. An AI model is a stochastic process, you don't get the same results from the same prompt (even at temperature 0 because the GPU will do multiplications in a different order and errors compound). It's not really the same as a digital compositing pipeline and a lighting engine.
Also you have to account for the PB of training data which consists of copyrighted works that goes into the model and is interpolated to produce the result. Personally this tips the balance to AI output being new work that's not copyrightable for me because otherwise all of the people who own copyright on the training data should have joint copyright of the output too because in a sense all of the training data is part of the prompt too.
Your prompt doesn't really contain all of the decisions that were made in the final output most of the time. When you use a rendering engine it does contain only your decisions and you have a license to use the work of all the people who created the software and other assets you're using to make the film.
... then all elements of the prompt that come from your human creativity and that survive into the final result remain under your copyright. Which in most cases should be almost as good as owning the full copyright, although it might influence fair use defenses if even a full copy infringes on only a minor portion of the work.
If your prompt is a novel, it's probably copyright but you don't have copyright over someone else's thing they made using a different novel-length prompt, even if the thing is similar to yours, since they didn't copy the prompt. If your prompt is a sentence, it's probably too minimal to qualify for copyright.
> How will the USPTO draw a legal distinction between AI-generated and GPU-generated?
Patents and trademarks are not copyrights. The U.S. Copyright office handles copyrights, the USPTO handles the other two.
In the spirit of your question, though, the concern predates AI.
Programs that you can use a stylus to draw with - in a similar manner as pen/ink and paper - have been around a long time, and are getting better.
Let the definition of human art be human effort put into creating something. If I put "etymology of the word art" into Google I get "The word art comes from the Latin noun ars (meaning practical skill, craft, or work of art)."
If I spend an hour drawing something in a program - strokes in the same exact manner as physical paint and canvas, but digitally adjustable and editable afterward - this is still art. But it's hard to prove that from just the output file alone. It's even worse for 3d model creation - someone who painstakingly molds/creates a 3d model in Blender is definitely doing art, even if there is no physical artifact until it's printed or projected.
Perhaps art programs will need to start authenticating input devices and movement sensors and recording their stream during the artistic process, and including that as part of the output file. This will only be worse if you want to create art but not train AI. The only solution to that is not create digital art.
I’m surprised that Doctorow has such a poor understanding of US copyright law. “Sweat of the brow” does not exist here. It is not “fundamental to a correct legal understanding of copyright.” You can argue it should be, but it isn’t. All that matters is human authorship, and LLMs aren’t humans.
> The simplest, most reliable way to ensure the works you invest in are fully and entirely eligible for copyright is to just use humans for the whole project. Declare a "no-AI policy," and hire human artists to produce the works you hope to sell. Monitor their activity to ensure they're not using AI. Keep good records.
What happens with work that is derived from LLMs that were trained on copyrighted work? Does that mean whatever I generate is not subject to copyright laws because I generated with AI ?
> The simplest, most reliable way to ensure the works you invest in are fully and entirely eligible for copyright is to just use humans for the whole project. Declare a "no-AI policy," and hire human artists to produce the works you hope to sell. Monitor their activity to ensure they're not using AI. Keep good records.
That seems true.
But also: generate all you like and then have humans edit, and keep a record of it, seems plausibly defensible too.
The current state of a lot of AI generated content material, not all / not necessarily program code, seems to really benefit from human editing.
I'm confused by the framing that copyright protecting creative expression and not hard work is something that would come as a shock to media bosses. If you're in the business of buying up creative stories and paying for the hard work necessary to sell them to a large audience, then being able to automate the expensive part is good news. Imagine film credits listing just the director, screen writer and some assistants, with everyone whose job was never primarily about making creative decisions being eliminated.
Thinking of all the media execs, crying, because their plans to fire all their creatives, and replace them with slophaus tech, are rendered null and void, from this.
I'm having a real hard time, feeling sorry for them.
For media bosses, copyright is a pseudo-property† right
...it's not "pseudo". What does the author think "IP" stands for?
Regardless, 'I agree with these people I always disagree with are somehow right this time despite acting like they always do' should be a sign that some reflection is needed!
I guess “pseudo” based on an understanding of “ownership” that works only for physical objects; not for 1:1 replicas and not for “mind stuff”. You cannot “own” a thought. It’s difficult to own an earthworm, unless you keep it in a confined container. Once it slips out, you lose ownership.
Based on that perspective, intellectual property is a construct, an illusion, newspeak, and thus “pseudo”.
Oh for sure, I agree to some extent! All property rights are more illusory and arbitrary than we assume as people born into capitalism (propertyism), but property rights concerning brands, processes, and… basically any idea are particularly illusory.
In context of this paper though, the author is saying the opposite, and mocking the fatcat CEOs for corrupting the pure institution of copyright so.
> AI slop isn't copyrightable, but there can be copyrightable elements in an AI-generated image. The most obvious element is the prompt. So long as your prompt contains some quantum of creativity (e.g. "Make me a beautiful flower whose loveliness is reminiscent of a newborn baby", but not "make me a beautiful flower"), that prompt is copyrightable. That means you can use copyright to challenge others' rights to present the same prompt to a chatbot. But while the prompt can be copyrighted, the image it gives rise to isn't. So you can spend a month iteratively refining your prompt, consuming all your town's water along with 200,000 gallons of diesel and that prompt will be yours for your life plus 70 years, but the image it generates is in the public domain, and anyone can copy it, sell it or give it away
I don't find this line of thought very convincing. The more detailed the prompt is, the more control the author exercises over the result. A sufficiently detailed prompt may become indistingushable from source code, because it specifies the desired result so precisely that it acts like source code for a compiler. Compilers, too, are not necessarily deterministic. Compilers may have a lot of freedom in how they translate the source code into machine instructions, but this does not break the connection with the original human intellectual effort and the result remains copyrightable.
There should be a difference between asking AI to "draw a funny picture of a donkey", and spending a month iteratively refining your prompt.
How does this mesh with this, where the court decided to declare AIs as authors and originators of copyright, just to make sure billionaire companies like Microsoft avoid liability to programmers they copied from?
Obviously courts will cheat, or should I say "keep cheating", until the output of AI models is covered by copyright while the input is not, because if AI output isn't covered, then most tech companies have nothing of value in their entire AI investment. Anything to cheat individuals out of their rights into the deep pockets of billionaire companies.
Courts aren't forced to make decisions consistent with other courts and other decisions. It's entirely possible that one case says an AI was an author and another case says AIs can't be authors.
The court in that case didn't rule on AI authorship at all.
"The coders describe an AI tool that doesn’t look up and reproduce stored work, U.S. Circuit Judge Eric Miller determined, but rather creates new work. That new work, the Donald Trump appointee said, may or may not infringe plaintiffs’ copyrights in their code, but it cannot reasonably be described as a copy of that code from which the copyright management information has been removed or altered."
Yes they always give some sort of excuse to go for the megacorp, the billionaire. Remember the judgements that because 9 year olds in a basement "orchestrate" the copyright crime of copying buffers with copyrighted data in Telco's network and routers those 9 year olds should go to jail?
This was the court's excuse, because copyright doesn't technically say there's anything illegal about downloading, yet they still wanted the 9 year old in jail, so buffer copies counted as uploads. Of course, the party that actually did the vast majority of the copying, the Telco, being a megacorp, wasn't responsible (like any other party would be: if you get paid or even deceived into committing a crime you are very much guilty of that crime. It is not a defense, not even if you didn't realize, and will generally get your sentence further increased). This became the crime that the 9 year old was guilty of, nobody else involved (except of course the parents, who had to pay)
And now that megacorps, including Microsoft of all companies, got caught violating literally hundreds of millions of copyright licences ... we simply declare that to be OK (retroactively OK even, which is NOT something either Congress or the president even have the power to do) ... with another ridiculous excuse (okay a 2-part ridiculous exuse):
1) AI chatbots are the originators of copyright (despite copyright law saying only people can be the originators of the right in copyright)
(why? Because if they didn't originate copyright, it wouldn't be allowed in copyright law for the owners of the model to say who the copyright belongs to, ie. customers wouldn't be legally allowed to say they made the chatbot output. And that this violates the law? Oh well)
2) AI models, despite both the models themselves AND their output obviously being based on the works of humans, as described both in every paper on ML going back 70+ years AND that being the very goal, the very purpose of AI in the first place (to reproduce their inputs) ... aren't
Question: how do I get this sort of treatment from courts? After all, a billion dollar company idea is very easy if you just get to violate the law at such enormous scale. How about I build the Netflix everyone wants? How do I get courts to, say, that if I use the "I" ML model described in math literature going back a hundred years, the output is owned by me? I, being the identity of course, the model that just copies it's input verbatim to it's output. Then, with this court's exact reasoning applied here to avoid billionaire liability, I don't just own the copyright on the output (ie. Disney's movies, and everyone else's, and all books, and ...) (because the output is 100% identical to the input), but I get to sell all the rights on it, including derivative rights. So not only do I get to sell these movies, but I even get money if they make a sequel.
Yet "somehow", I won't get the billionaire's treatment at the courts, will I?
I mean, why do courts even bother? We should empty the buildings (great savings to be had there!), hammer the doors shut and just put up a plaque at the entrance "pay the billionaire, peasant". It'd be a lot more honest.
Again, the court didn't rule on AI authorship, so your interpretation "1) AI chatbots are the originators of copyright" is wrong. It only ruled that the AI output was not the plaintiffs' work, therefore they couldn't claim that their work was being distributed, although they could've claimed (but didn't) that a work infringing on their copyright over their work was being distributed.
You can get what you call the "billionaire's treatment" by hiring a lawyer who understands the distinction between those two situations, who can accordingly pick the right law for you to sue under.
You're totally misunderstanding. How can I get courts to find these sorts of ridiculous excuses to avoid I am held liable, as opposed to billionaires?
Actually your reasoning would still work: how do I get a court to say that AI outputs aren't copyrighted by what went into the AI? How do I get a court to say that if I use the simplest Transformer network possible, "I", in math terms, and get a court to judge that it's output is under my copyright, or at least not under Disney's? Obviously I is the "identity" network, which just reproduces it's input.
Oh and don't worry. There's a variant of I that reproduces it's training data too, the exact thing the copyright office approved? It's a lot larger, but it's not like you need to actually calculate it's output, so that doesn't even matter ...
How do I get courts to not say "heyhey, peasant, Disney's the billionaire company here, YOU will be paying THEM".
How do I get the AI lab deal for my Netflix-without-copyright-but-with-identity-ML project?
How do I get courts to make the same utterly ridiculous reasoning they made here, against MY liability towards a billionaire company? To be utterly unreasonable like they are here, except benefitting me?
I mean that the mere idea that the court will respect its reasoning here is so ridiculous that nobody's even willing to try it. Well, except Sam Altman and Amodei and other tech billionaires.
If you insist on the identity transformation, you'll fail even the basic "are these two different works or the same?" test that decided the court case you cited, so that's not going to work.
Saying 2 works are identical because the bytes are identical is the mathematical definition of identical. It's the very first axiom, generally referred to as the "reflexive property". "x=x". It says that, for mathematical purposes, a 3 made of concrete and a 3 made of candy are the same thing, because it doesn't differ in essential properties. In every mathematical equation you can always say that both x's are obviously different: they don't have the same position of the page! So this axiom really is needed and really does mean something.
But the point is: courts are absolutely famous for ignoring mathematical reality. Like in this case. The whole purpose of the entire field of ML is "universal approximation", which can be very accurately described as making 2 functions the same, using backprop with transformers being one way to do that. The whole point of making something like ChatGPT or Claude is that the outputs would NOT be different from the output of these programmers. And the judge ignored the entire field to make this judgement. The judge ignored one meaning of identical in favor of another, and what do you know, it just happens to be the one to give all the money to billionaires.
As you point out, obviously the judge will do the exact opposite reasoning to protect billionaires as well. They will protect Disney against these programmers making AI versions of Disney movies, based on their real movies, either through identity transformation (there is, of course, no "equality test", because even a dvdrip and a dvd aren't the same, not even when decoded the pixels don't match). We both know this to be true 99.999%. Hence "excuse". Exactness doesn't matter. And if I make 50 AI slop versions (all of which are different) of all movies in existence, and create a Netflix showing them, we all know that the judge will judge against me in the first 5 minutes to protect huge companies' revenues.
The only question that remains is what happens when the inevitable comes, and an AI company billionaire produces an obvious duplicate of a movie [2], and we see a billionaire vs billionaire case. Given that it's Elon Musk that might get sued and what he's done, I'm guessing it'll depend on which party gets elected.
But I repeat my question: how do I get the courts to make philosophically defensible choices where I get the money? Clearly the judge decided that this mathematically identical output of these programmers does not mean they hold copyright. And as you agree, an AI changed disney movie the judge will make the opposite judgement to protect Disney's billions. Well, how do I get the judge to take the exact same position on Disney movies, to throw Disney under the bus and protect me and give me money instead? I'll make the Netflix everyone wants with it. So how do I get the judge to have my back instead of some billionaire's back? Both viewpoints are obviously both valid, or at least philosophically defensible. How do I get the judge to pick the side where I get the money and not some billionaire.
The article is just restating US copyright law for people who don't know it. The title indicates a recent change in policy, and there hasn't been one.
Now, this is US law. Much of the world does have "sweat of the brow" copyright. But the US does not. See "Feist vs. Rural Telephone". There must be human creativity because the US constitution reads "(Congress shall have the power) To promote the Progress of Science and useful Arts, by securing for limited Times to Authors and Inventors the exclusive Right to their respective Writings and Discoveries." That limits the power of Congress to change this.
So a lot of things are uncopyrightable in the US. Feist is the big case, and it was about telephone directories. No copyright. Useful objects are not copyrightable, which is why there are third party auto parts. Facts are not copyrightable, nor are databases of facts. Nor are pictures of public domain art. Nor is automotive sheet metal, despite lobbying by automakers.
That's what leads to AI-generated content not being copyrightable. Model weights don't seem to be copyrightable either.
> Useful objects are not copyrightable
It cannot be this clean cut or Google v Oracle would have been trivial to decide.
A telephone directory can be copyright protected, too. The facts within it cannot. Exactly like recipes. A particular presentation of a recipe is absolutely copyright protected. But this doesn’t mean no one can publish the same recipe in their own presentation
Hah, this triggers my pet peeve about IP for software: Copyright for code is a category error.
Copyright very explicitly does NOT cover functional aspects, only creative expression. But the value of software is not in the expressive aspects of the code that generates it, the value is in the functionality it provides. And this is the exact thing that copyright does NOT protect!
Like, literally, we're protecting the fact that we used a for-loop instead of a while-loop to do a useful thing, not the useful thing itself. This is why rewriting enough code is sufficient to bypass all copyright protection, as the recent AI rewrites have shown.
This was always a huge legislative hack, rightfully controversial at the time it happened, but legislators didn't want to bother figuring out a proper IP regime for software -- maybe something between copyright and patents -- so they shoehorned code into copyright.
Google v Oracle was trivial to decide, but it got decided the wrong way. Because it became a trial around laws based on an ill-founded legislative hack and the decades of flawed case-law it spawned. Stepping back, it is very clear that Sun spent billions making a very useful thing and Google co-opted it for its own uses without recompense because they bet (correctly) that they could legally get away with it.
It is somewhat analogous to what the AI companies are doing with all the code and data out there. Funny that people are unhappy about THAT but cheered when Google did it to Sun / Oracle. I wonder what changed ;-)
I think they're referring to more fundamental things like math, language, etc...
A recipe book can have copyright because it's for the specific combination of recipes, layout, etc... Not for a recipe or more generally a recipe book.
I think that's consistent with his the supreme court ruled in Google v Oracle.
> The CO has taken the (entirely correct) position that copyright law exists to incentivize human acts of creative expression.
The US Constitution grants Congress the power to grant copyrights "to promote the progress of Science and the useful Arts." Creative expression that is not useful therefore is not constitutionally copyrightable.
But that aside, I agree with the author that the Copyright Office refusing to further expand the already vastly over-expanded scope of copyrights is a rare positive ruling.
You have left out the second half of the sentence, "... by securing for limited Times to Authors and Inventors the exclusive Right to their respective Writings and Discoveries."
The position taken by the courts and the administration is that "authors and inventors" means "human". The monkey could not be the author of its own selfie photograph, for instance.
Maybe this doesn't reflect the intentions of the Framers, but it certainly should color any modern discussion surrounding the purpose of copyright law.
I also notice it doesn't say anything about their heirs. I wonder if there is a meaningful argument that copyright extending beyond the life of the original author is unconstitutional.
Likewise corporations are not humans, so they shouldn't be able to own copyrights either.
Are you just arguing that corporations should not be able to own copyrights but it is OK for them to own other things such as land and equipment, or are you arguing that corporations should not be allowed to own anything?
"Intellectual Property" is not, or at least should not, be the same as physical property. Copyrights and patents are temporary artificial, government enforced monopolies to incentivise creativity and publication of the results of that creativity. I don't think it's unreasonable to say that a corporation can't own the copyright itself. Although, they should still be able to hold a license to works from the actual human creators, and granting such a license could be a condition of employment (provided it was created as part of the employment, and not, say, on the employees own time).
"Useful Arts" is old timey talk for the mechanical trades. Engineering, manufacturing and that kind of thing. It applies to "Inventors" and their "Discoveries," and is the basis of patent law.
There is another clause later tha clarifies it a bit. It does NOT mean that an invention must be useful, and it's pretty clear that purely l utilitarian objects actually can not be copyrighted.
> Basically, we're about to recreate formalities, but managed by the private sector. To the extent that companies use AI in their work, they are going to have to track every pixel and every word and every note that is output by an AI system, and which parts were added or edited by humans after creation.
> That means that companies that want to protect their commercial interests aren't just going to get more protection based on how much work they get humans to do; they're also going to have to treat copyright's government-enforceable monopoly over expressive speech with the gravitas it deserves: not something you get automatically, but something you have to do real administrative work for.
Assuming this happens, wouldn't this mean that copyright becomes significantly less useful for people who aren't corporations and don't have the legal, financial, and administrative resources to produce this kind of bookkeeping?
Not if you take option 2, just don't use AI. The companies have to account for where they used AI in order to own the rights, if you didn't use AI, there's nothing to prove. If someone takes your work and you sue them for infringement, and there's no evidence you used AI, then there's no reason to dispute your right to the work
I don't quite get it. Can Pixar copyright its GPU-rendered films or can it only copyright the human-created models?
How will the USPTO draw a legal distinction between AI-generated and GPU-generated?
If you put something that is copyrighted through a mechanical transformation process it has the same copyright. For instance a video of a movie retains the copyright status of the movie (as well as probably being an illegal copy itself) even if you added camera shake and audience coughing sounds. So there's not an issue with things like this.
Well OK, so what if I put my copyrighted prompt through the mechanical transformation process of the AI model?
I think the point is to decide where to draw the line. An AI model is a stochastic process, you don't get the same results from the same prompt (even at temperature 0 because the GPU will do multiplications in a different order and errors compound). It's not really the same as a digital compositing pipeline and a lighting engine.
Also you have to account for the PB of training data which consists of copyrighted works that goes into the model and is interpolated to produce the result. Personally this tips the balance to AI output being new work that's not copyrightable for me because otherwise all of the people who own copyright on the training data should have joint copyright of the output too because in a sense all of the training data is part of the prompt too.
Your prompt doesn't really contain all of the decisions that were made in the final output most of the time. When you use a rendering engine it does contain only your decisions and you have a license to use the work of all the people who created the software and other assets you're using to make the film.
... then all elements of the prompt that come from your human creativity and that survive into the final result remain under your copyright. Which in most cases should be almost as good as owning the full copyright, although it might influence fair use defenses if even a full copy infringes on only a minor portion of the work.
If your prompt is a novel, it's probably copyright but you don't have copyright over someone else's thing they made using a different novel-length prompt, even if the thing is similar to yours, since they didn't copy the prompt. If your prompt is a sentence, it's probably too minimal to qualify for copyright.
> How will the USPTO draw a legal distinction between AI-generated and GPU-generated?
Patents and trademarks are not copyrights. The U.S. Copyright office handles copyrights, the USPTO handles the other two.
In the spirit of your question, though, the concern predates AI.
Programs that you can use a stylus to draw with - in a similar manner as pen/ink and paper - have been around a long time, and are getting better.
Let the definition of human art be human effort put into creating something. If I put "etymology of the word art" into Google I get "The word art comes from the Latin noun ars (meaning practical skill, craft, or work of art)."
If I spend an hour drawing something in a program - strokes in the same exact manner as physical paint and canvas, but digitally adjustable and editable afterward - this is still art. But it's hard to prove that from just the output file alone. It's even worse for 3d model creation - someone who painstakingly molds/creates a 3d model in Blender is definitely doing art, even if there is no physical artifact until it's printed or projected.
Perhaps art programs will need to start authenticating input devices and movement sensors and recording their stream during the artistic process, and including that as part of the output file. This will only be worse if you want to create art but not train AI. The only solution to that is not create digital art.
Makes me think of Worth Enough, by Radoxist[0]. I think he spent about a year on it.
[0] https://www.deviantart.com/radoxist/art/Worth-enough-7324787...
I’m surprised that Doctorow has such a poor understanding of US copyright law. “Sweat of the brow” does not exist here. It is not “fundamental to a correct legal understanding of copyright.” You can argue it should be, but it isn’t. All that matters is human authorship, and LLMs aren’t humans.
Which is exactly what the article explains through the 10 year compilation effort Vs the 10 second haiku.
Did you read the whole thing? I'm not seeing where you're disagreeing with the article
> The simplest, most reliable way to ensure the works you invest in are fully and entirely eligible for copyright is to just use humans for the whole project. Declare a "no-AI policy," and hire human artists to produce the works you hope to sell. Monitor their activity to ensure they're not using AI. Keep good records.
What happens with work that is derived from LLMs that were trained on copyrighted work? Does that mean whatever I generate is not subject to copyright laws because I generated with AI ?
> The simplest, most reliable way to ensure the works you invest in are fully and entirely eligible for copyright is to just use humans for the whole project. Declare a "no-AI policy," and hire human artists to produce the works you hope to sell. Monitor their activity to ensure they're not using AI. Keep good records.
That seems true.
But also: generate all you like and then have humans edit, and keep a record of it, seems plausibly defensible too.
The current state of a lot of AI generated content material, not all / not necessarily program code, seems to really benefit from human editing.
I'm confused by the framing that copyright protecting creative expression and not hard work is something that would come as a shock to media bosses. If you're in the business of buying up creative stories and paying for the hard work necessary to sell them to a large audience, then being able to automate the expensive part is good news. Imagine film credits listing just the director, screen writer and some assistants, with everyone whose job was never primarily about making creative decisions being eliminated.
Wait, but what if you worked hard for 10 months iterating with a chatbot, which used an LLM to write the code behind the scenes?
You didn't write the actual code by hand, but you did iterate a lot and fix a lot of bugs, etc.
Thinking of all the media execs, crying, because their plans to fire all their creatives, and replace them with slophaus tech, are rendered null and void, from this.
I'm having a real hard time, feeling sorry for them.
Regardless, 'I agree with these people I always disagree with are somehow right this time despite acting like they always do' should be a sign that some reflection is needed!
>it's not "pseudo". What does the author think "IP" stands for?
https://www.gnu.org/philosophy/words-to-avoid.html#Intellect...
I guess “pseudo” based on an understanding of “ownership” that works only for physical objects; not for 1:1 replicas and not for “mind stuff”. You cannot “own” a thought. It’s difficult to own an earthworm, unless you keep it in a confined container. Once it slips out, you lose ownership.
Based on that perspective, intellectual property is a construct, an illusion, newspeak, and thus “pseudo”.
> Once it slips out, you lose ownership
You lose possession and control. Ownership is fuzzier.
Oh for sure, I agree to some extent! All property rights are more illusory and arbitrary than we assume as people born into capitalism (propertyism), but property rights concerning brands, processes, and… basically any idea are particularly illusory.
In context of this paper though, the author is saying the opposite, and mocking the fatcat CEOs for corrupting the pure institution of copyright so.
There should be a difference between asking AI to "draw a funny picture of a donkey", and spending a month iteratively refining your prompt.
How does this mesh with this, where the court decided to declare AIs as authors and originators of copyright, just to make sure billionaire companies like Microsoft avoid liability to programmers they copied from?
https://www.courthousenews.com/coders-lose-appeal-in-copyrig...
Obviously courts will cheat, or should I say "keep cheating", until the output of AI models is covered by copyright while the input is not, because if AI output isn't covered, then most tech companies have nothing of value in their entire AI investment. Anything to cheat individuals out of their rights into the deep pockets of billionaire companies.
Courts aren't forced to make decisions consistent with other courts and other decisions. It's entirely possible that one case says an AI was an author and another case says AIs can't be authors.
The court in that case didn't rule on AI authorship at all.
"The coders describe an AI tool that doesn’t look up and reproduce stored work, U.S. Circuit Judge Eric Miller determined, but rather creates new work. That new work, the Donald Trump appointee said, may or may not infringe plaintiffs’ copyrights in their code, but it cannot reasonably be described as a copy of that code from which the copyright management information has been removed or altered."
Yes they always give some sort of excuse to go for the megacorp, the billionaire. Remember the judgements that because 9 year olds in a basement "orchestrate" the copyright crime of copying buffers with copyrighted data in Telco's network and routers those 9 year olds should go to jail?
This was the court's excuse, because copyright doesn't technically say there's anything illegal about downloading, yet they still wanted the 9 year old in jail, so buffer copies counted as uploads. Of course, the party that actually did the vast majority of the copying, the Telco, being a megacorp, wasn't responsible (like any other party would be: if you get paid or even deceived into committing a crime you are very much guilty of that crime. It is not a defense, not even if you didn't realize, and will generally get your sentence further increased). This became the crime that the 9 year old was guilty of, nobody else involved (except of course the parents, who had to pay)
And now that megacorps, including Microsoft of all companies, got caught violating literally hundreds of millions of copyright licences ... we simply declare that to be OK (retroactively OK even, which is NOT something either Congress or the president even have the power to do) ... with another ridiculous excuse (okay a 2-part ridiculous exuse):
1) AI chatbots are the originators of copyright (despite copyright law saying only people can be the originators of the right in copyright)
(why? Because if they didn't originate copyright, it wouldn't be allowed in copyright law for the owners of the model to say who the copyright belongs to, ie. customers wouldn't be legally allowed to say they made the chatbot output. And that this violates the law? Oh well)
2) AI models, despite both the models themselves AND their output obviously being based on the works of humans, as described both in every paper on ML going back 70+ years AND that being the very goal, the very purpose of AI in the first place (to reproduce their inputs) ... aren't
Question: how do I get this sort of treatment from courts? After all, a billion dollar company idea is very easy if you just get to violate the law at such enormous scale. How about I build the Netflix everyone wants? How do I get courts to, say, that if I use the "I" ML model described in math literature going back a hundred years, the output is owned by me? I, being the identity of course, the model that just copies it's input verbatim to it's output. Then, with this court's exact reasoning applied here to avoid billionaire liability, I don't just own the copyright on the output (ie. Disney's movies, and everyone else's, and all books, and ...) (because the output is 100% identical to the input), but I get to sell all the rights on it, including derivative rights. So not only do I get to sell these movies, but I even get money if they make a sequel.
Yet "somehow", I won't get the billionaire's treatment at the courts, will I?
I mean, why do courts even bother? We should empty the buildings (great savings to be had there!), hammer the doors shut and just put up a plaque at the entrance "pay the billionaire, peasant". It'd be a lot more honest.
Again, the court didn't rule on AI authorship, so your interpretation "1) AI chatbots are the originators of copyright" is wrong. It only ruled that the AI output was not the plaintiffs' work, therefore they couldn't claim that their work was being distributed, although they could've claimed (but didn't) that a work infringing on their copyright over their work was being distributed.
You can get what you call the "billionaire's treatment" by hiring a lawyer who understands the distinction between those two situations, who can accordingly pick the right law for you to sue under.
You're totally misunderstanding. How can I get courts to find these sorts of ridiculous excuses to avoid I am held liable, as opposed to billionaires?
Actually your reasoning would still work: how do I get a court to say that AI outputs aren't copyrighted by what went into the AI? How do I get a court to say that if I use the simplest Transformer network possible, "I", in math terms, and get a court to judge that it's output is under my copyright, or at least not under Disney's? Obviously I is the "identity" network, which just reproduces it's input.
Oh and don't worry. There's a variant of I that reproduces it's training data too, the exact thing the copyright office approved? It's a lot larger, but it's not like you need to actually calculate it's output, so that doesn't even matter ...
How do I get courts to not say "heyhey, peasant, Disney's the billionaire company here, YOU will be paying THEM".
How do I get the AI lab deal for my Netflix-without-copyright-but-with-identity-ML project?
How do I get courts to make the same utterly ridiculous reasoning they made here, against MY liability towards a billionaire company? To be utterly unreasonable like they are here, except benefitting me?
I mean that the mere idea that the court will respect its reasoning here is so ridiculous that nobody's even willing to try it. Well, except Sam Altman and Amodei and other tech billionaires.
If you insist on the identity transformation, you'll fail even the basic "are these two different works or the same?" test that decided the court case you cited, so that's not going to work.
Saying 2 works are identical because the bytes are identical is the mathematical definition of identical. It's the very first axiom, generally referred to as the "reflexive property". "x=x". It says that, for mathematical purposes, a 3 made of concrete and a 3 made of candy are the same thing, because it doesn't differ in essential properties. In every mathematical equation you can always say that both x's are obviously different: they don't have the same position of the page! So this axiom really is needed and really does mean something.
But the point is: courts are absolutely famous for ignoring mathematical reality. Like in this case. The whole purpose of the entire field of ML is "universal approximation", which can be very accurately described as making 2 functions the same, using backprop with transformers being one way to do that. The whole point of making something like ChatGPT or Claude is that the outputs would NOT be different from the output of these programmers. And the judge ignored the entire field to make this judgement. The judge ignored one meaning of identical in favor of another, and what do you know, it just happens to be the one to give all the money to billionaires.
As you point out, obviously the judge will do the exact opposite reasoning to protect billionaires as well. They will protect Disney against these programmers making AI versions of Disney movies, based on their real movies, either through identity transformation (there is, of course, no "equality test", because even a dvdrip and a dvd aren't the same, not even when decoded the pixels don't match). We both know this to be true 99.999%. Hence "excuse". Exactness doesn't matter. And if I make 50 AI slop versions (all of which are different) of all movies in existence, and create a Netflix showing them, we all know that the judge will judge against me in the first 5 minutes to protect huge companies' revenues.
The only question that remains is what happens when the inevitable comes, and an AI company billionaire produces an obvious duplicate of a movie [2], and we see a billionaire vs billionaire case. Given that it's Elon Musk that might get sued and what he's done, I'm guessing it'll depend on which party gets elected.
But I repeat my question: how do I get the courts to make philosophically defensible choices where I get the money? Clearly the judge decided that this mathematically identical output of these programmers does not mean they hold copyright. And as you agree, an AI changed disney movie the judge will make the opposite judgement to protect Disney's billions. Well, how do I get the judge to take the exact same position on Disney movies, to throw Disney under the bus and protect me and give me money instead? I'll make the Netflix everyone wants with it. So how do I get the judge to have my back instead of some billionaire's back? Both viewpoints are obviously both valid, or at least philosophically defensible. How do I get the judge to pick the side where I get the money and not some billionaire.
[1] https://www.france24.com/en/live-news/20251111-german-court-...
[2] https://variety.com/2026/film/news/elon-musks-grok-ai-genera...