USCO report/draft on copyright status of generative AI

[nice90sguy]

You're not reading in closely enough. As an outside observer (Australian), it's pretty easy to see with the current US administration there is no such thing as coincidence. That's all I'll say, so this thread doesn't get shifted to the Political Board.

So assuming it's not a coincidence, your post doesn't explain what the issue is. I assume it's something to do with AI companies bankrolling , threathening and bribing politicians and political institutions. I guess I'll just do my own research to get a clear answer to my question, I doubt if the political board here is a great place to learn facts.

 

[ElectricBlue]

So assuming it's not a coincidence, your post doesn't explain what the issue is. I assume it's something to do with AI companies bankrolling , threathening and bribing politicians and political institutions. I guess I'll just do my own research to get a clear answer to my question, I doubt if the political board here is a great place to learn facts.

Musk is spruiking his own AI, wants open slather for his training content. This US copyright position paper would limit his ability to get that. Given that, surely the "coincidence" doesn't need to be spelt out?

As I say, read in a bit more. And you're right, the PB is the last place you'd go for informed comment.

 

[SimonDoom]

Suppose you taught a standard English Literature course based on the Canon at a mid-west university, with a modern literature module (books in copyright); fair use?

Suppose you used the knowledge you acquired as a student and in academia to write a work of modern fiction; don't know? Really?

Take it from me, using your own knowledge and learning would amount to fair use.

In the two examples you've given (I'm basing this on US law) you wouldn't reach the question of a fair use defense, because there's no prima facie infringement. There's no reproduction. We're talking here about the use of AI, which, in the process of collecting and organizing and processing information, may well entail a great deal of "reproduction" of copyrighted works (digital copying of a work to include it in a database).

Fair use only becomes an issue once a prima facie case for infringement is made, because fair use is a defense. A prima facie case for infringement requires the plaintiff to prove a) ownership of a copyrighted work, b) registration of the work, c) access by the defendant to the work, and d) the defendant engaged in one of the exclusive rights of the copyright owner, e.g., reproduction, publication, creation of a derivative work, performance, etc. Once the plaintiff establishes a prima facie case, the defendant has the burden of proving its fair use defense under section 107.

I agree with you that in the first two cases you gave there's no actionable conduct, unless in the case of your first example the teacher makes unauthorized copies of the works being taught and distributes them to the class. That would not be fair use, even though it's done for instruction. Teachers must generally require students to purchase the works being read in class, or to get them from the library.

I recommend that anyone interested read the third report in Bramblethorn's link. Bramblethorn was correct that the report discusses how the mere use of AI in connection with a database of copyrighted works may be considered copyright infringement even if the use doesn't generate a text that itself appears infringing. This is still up in the air, but this view was supported by some of the people consulted for the report. My personal view--not the law, just my personal view--is that this is problematic for a lot of reasons and it may be difficult to sustain in light of the enormous pressure that is sure to come to enable the use of AI to process information and generate analysis.

I've always been a little unclear as to why the legal arguments for "human-produced" material and "human-using-AI" material should be any different.

There are two important differences. One is that an AI is not a person and therefore cannot be a copyright owner. Only actual persons can be authors within the meaning of copyright law. So, to the extent that a machine has created an original work, it is not a copyrightable work; only the creative input of the human involved should be considered in determining if it is subject to copyright protection. This doesn't mean, however, that the PRODUCT of the AI cannot be infringing if the human behind the scenes then publishes the story and profits from it. A work can be infringing even if it doesn't qualify for copyright protection. E.g., an exact copy of a copyrighted image made by scanning the image.

Second, humans and AI use source material in different ways, even if the purpose of the use is growing increasingly similar. When a human reads a set of books, takes notes, and files the material away in his brain for use later, there's no "reproduction" in the copyright sense. But when an AI scans a story database and saves the pages for processing later, there IS reproduction in the copyright sense.

 

[XerXesXu]

In the two examples you've given (I'm basing this on US law) you wouldn't reach the question of a fair use defense, because there's no prima facie infringement. There's no reproduction. We're talking here about the use of AI, which, in the process of collecting and organizing and processing information, may well entail a great deal of "reproduction" of copyrighted works (digital copying of a work to include it in a database).



I recommend that anyone interested read the third report in Bramblethorn's link.



There are two important differences. One is that an AI is not a person and therefore cannot be a copyright owner. Only actual persons can be authors within the meaning of copyright law.

‘In the two examples you've given (I'm basing this on US law) you wouldn't reach the question of a fair use defense, because there's no prima facie infringement.’

How do you know? There’s no rule of law to that effect. I could knock out an infringing paragraph or two in ten minutes. Look how many academics ‘accidentally’ plagiarise (copy). If you’ve read a work, you have the potential to copy it in an infringing manner, a sensitive copyright holder could sue.

“I recommend that anyone interested read the third report in Bramblethorn's link.’

I really couldn’t. In summary it says, some people say this, some say that, we’re sitting on the fence, we recommend that Congress do nothing and awaits events. I can see that DOGE may be coming for them.

‘I've always been a little unclear as to why the legal arguments for "human-produced" material and "human-using-AI" material should be any different.’

Nice90sGuy makes a valid point. In many jurisdictions, including the UK, AI produced material CAN be copyrighted by the operator of the AI and, absent a copyright infringement, won’t be actionable. It’s just another tool.

 

[SimonDoom]

‘In the two examples you've given (I'm basing this on US law) you wouldn't reach the question of a fair use defense, because there's no prima facie infringement.’

How do you know? There’s no rule of law to that effect. I could knock out an infringing paragraph or two in ten minutes. Look how many academics ‘accidentally’ plagiarise (copy). If you’ve read a work, you have the potential to copy it in an infringing manner, a sensitive copyright holder could sue.

“I recommend that anyone interested read the third report in Bramblethorn's link.’

I really couldn’t. In summary it says, some people say this, some say that, we’re sitting on the fence, we recommend that Congress do nothing and awaits events. I can see that DOGE may be coming for them.

‘I've always been a little unclear as to why the legal arguments for "human-produced" material and "human-using-AI" material should be any different.’

Nice90sGuy makes a valid point. In many jurisdictions, including the UK, AI produced material CAN be copyrighted by the operator of the AI and, absent a copyright infringement, won’t be actionable. It’s just another tool.

I "know" simply because your first example does not provide detail of anything that would constitute infringement. As my previous answer explained, you could add details to the hypothetical that would make it infringement, but you didn't do that. "Infringement" occurs when the infringer commits an act that is covered by section 106, which states what the copyright owner's exclusive rights are. You did not specifically identify anything in your first hypo that would constitute infringement.

As to your second point, copyright infringement and plagiarism are two distinct things, although people often use the terms interchangeably here. Infringement is the unauthorized intrusion upon a copyright owner's exclusive section 106 rights. Plagiarism is the use of another's ideas without attribution. Plagiarism is mostly confined to academia, because in the context of fiction, "ideas" are not protected, and you are free to borrow the "ideas" of another's copyrighted text without authorization or attribution without being ethically or legally concerned. What we're concerned with here is mostly infringement, not plagiarism. Proof of infringement requires evidence of a specific act that is the sole right of the author/owner under section 106.

Nice90sGuy makes a valid point. In many jurisdictions, including the UK, AI produced material CAN be copyrighted by the operator of the AI and, absent a copyright infringement, won’t be actionable. It’s just another tool.

My analysis is confined to US law. So far, US courts have said that only a human can be a copyright author. An example was the "monkey selfie" case, where the court held that a monkey that took selfies of itself with a phone could not be an author/owner of the photographs that resulted. Another example was a very recent federal court case that specifically held that AI-generated works are not entitled to copyright protection. US copyright law is based on an incentive, rather than moral desert, rationale, which makes it a bit different from European, especially continental, legal tradition. The purpose of copyright law is to incentivize authors and artists to make creative works. That rationale applies to humans, but not to monkeys or machines.

 

[SimonDoom]

Musk is spruiking his own AI, wants open slather for his training content. This US copyright position paper would limit his ability to get that. Given that, surely the "coincidence" doesn't need to be spelt out?

As I say, read in a bit more. And you're right, the PB is the last place you'd go for informed comment.

I think this is possible, but I'd like to see more evidence before jumping to conclusions. Musk, as I understand, is something of an IP skeptic, and someone who is likely to favor robust use of AI without too many IP constraints. So it's possible he might see this latest Copyright Office report as "anti-progress" (as well as hostile to his own business intentions). If I'm right about that I'm 60-40 inclined to think he's right about this issue. Not that it's a good reason to fire the Copyright Office director, whose authority is limited. Ultimately, the courts decide, not the Copyright Office.

 

[XerXesXu]

I "know" simply because your first example does not provide detail of anything that would constitute infringement. As my previous answer explained, you could add details to the hypothetical that would make it infringement, but you didn't do that. "Infringement" occurs when the infringer commits an act that is covered by section 106, which states what the copyright owner's exclusive rights are. You did not specifically identify anything in your first hypo that would constitute infringement.

As to your second point, copyright infringement and plagiarism are two distinct things, although people often use the terms interchangeably here. Infringement is the unauthorized intrusion upon a copyright owner's exclusive section 106 rights. Plagiarism is the use of another's ideas without attribution. Plagiarism is mostly confined to academia, because in the context of fiction, "ideas" are not protected, and you are free to borrow the "ideas" of another's copyrighted text without authorization or attribution without being ethically or legally concerned. What we're concerned with here is mostly infringement, not plagiarism. Proof of infringement requires evidence of a specific act that is the sole right of the author/owner under section 106.



My analysis is confined to US law. So far, US courts have said that only a human can be a copyright author. An example was the "monkey selfie" case, where the court held that a monkey that took selfies of itself with a phone could not be an author/owner of the photographs that resulted. Another example was a very recent federal court case that specifically held that AI-generated works are not entitled to copyright protection. US copyright law is based on an incentive, rather than moral desert, rationale, which makes it a bit different from European, especially continental, legal tradition. The purpose of copyright law is to incentivize authors and artists to make creative works. That rationale applies to humans, but not to monkeys or machines.

The whole point was to neither rule anything in, nor out. In both cases a claim for infringement could only arise if another person read the work and claimed infringement of their copyright. Absent an actual example, it's impossible to say, in either case, whether it may be infringing. Your AI-tuition example stands 4-square with my trad-tuition example.

Plagiarism and breach of copyright are not clean different things, they overlap, even in an academic context. In its ordinary meaning, as you and others use it on Lit, plagiarism by use of another's words, as I specified, is an infringement of copyright.'

You failed to quote Nice90sGuys in full. He was making a broad point about AI product versus non-AI product, not a point of law in any particular jurisdiction. In your initial example you suggested that because in generating the weightings for an LLM there may have been something you refer as "reproduction" (the inverted commas are yours) then the creative output of the AI-assisted teacher may be infringing. He makes the obvious point that there's no reason to put creativity which has deep history, involving a transformative use of reproduction in creating the weightings of an LLM, with any other creativity. In either case, the question of infringement can be determined by examining the creative work. Creative work should not be treated as presumptively an infringement simply because a particular tool has been used. I agree with him. In any real case, infringement can be determined by examination of the product.