Is it free software? [LWN.net]

Is it free software?

Posted Feb 20, 2026 15:36 UTC (Fri) by rfontana (subscriber, #52677)
Parent article: The Book of Remind

The README.md file says "Remind's source code may not be used to train an AI model,
including an LLM model, unless all of the output of said model is
released under the GNU General Public License, version 2. If you use
any of Remind's source code to train your model, then anything that
the model produces is a derived product of Remind and must be licensed
under the same terms as Remind."

I think that is probably a different license from vanilla GPLv2, and raises the issue whether that sort of declared interpretation makes the license non-FOSS (there's possibly also the argument that the interpretation is simply in conflict with GPLv2 even if it's otherwise FOSS). I'm initially inclined to think it's non-FOSS.

to post comments

Is it free software?

Posted Feb 20, 2026 15:40 UTC (Fri) by NightMonkey (subscriber, #23051) [Link] (30 responses)

Is there a FOSS license that prohibits AI ingestion of the source code and downstream usage?

Is it free software?

Posted Feb 20, 2026 16:28 UTC (Fri) by NYKevin (subscriber, #129325) [Link] (29 responses)

A blanket prohibition trivially violates freedoms 1 and 3 (because it prohibits e.g. training or operating a local model just on a single codebase, and redistributing its output under the same terms as the original). Current AI models aren't capable of that kind of exclusive single-project training (yet), but we must consider all future technologies that could plausibly exist, not just ones that have been demonstrated to date.

Whether you can craft a narrower prohibition that satisfies the four freedoms, but still imposes significant restrictions on current models, remains to be seen. It depends in part on various questions of law relating to AI models (i.e. if you can't use copyright law to restrict training in the first place, then that implies the use of a clickwrap license or EULA, and the FOSS community has historically tried to avoid such chicanery).

Is it free software?

Posted Feb 20, 2026 17:58 UTC (Fri) by rgmoore (✭ supporter ✭, #75) [Link] (26 responses)

I think any kind of prohibition on AI training goes against freedom 1: freedom to study how the program works. That seems like exactly what someone is doing by feeding source code into their AI training system. In this case, it's an AI that's studying it to understand how it works rather than a human, but I don't think that makes any real difference. It's still about freedom to study the program. Denying some ways of studying the program violates freedom 1 just as much as denying use of the program to some kinds of users violates freedom 0.

Is it free software?

Posted Feb 20, 2026 20:21 UTC (Fri) by NYKevin (subscriber, #129325) [Link] (25 responses)

I agree with this take, but I would clarify that freedom 1 applies to the human who is training the AI, not to the AI itself. Training implicates freedom 1 because it is ultimately a method for *humans* to distill information about some corpus, not because current-generation AIs are directly entitled to assert the four freedoms on their own behalf.

(Of course, AGI will greatly complicate this picture if and when it arises, but I don't want to get into that here.)

Is it free software?

Posted Feb 20, 2026 22:44 UTC (Fri) by rgmoore (✭ supporter ✭, #75) [Link] (24 responses)

I don't think it really matters whether we count it as the AI that's doing the training or a human who's doing it; either way, feeding source code into an AI model to train it is covered by Freedom 1. I feel like the discourse around AI and copyright could be simplified a lot if you ask the question "would this be copyright infringement if a human did it". It provides a conceptual framework for thinking about the problem that automatically includes everything we already know about copyright.

Is it copyright infringement if an AI produces a large fragment of a copyright work verbatim or nearly verbatim? It would be if a human did that, so yes, it is. Is it copyright infringement if an AI company gets bootleg copies of copyrighted works to train their AI? Sure thing, since getting bootleg copies is copyright infringement no matter what you intend to use them for. Is it copyright infringement is an AI is trained on a legitimately obtained book? No, since a human who reads a legitimately obtained book isn't committing copyright infringement.

There's still a bit of a fine point about who is actually responsible. I think you're correct that until an AI is legally recognized as a person, it's the people around the AI who are guilty of infringement, not the AI itself. It's also important to be able to assign blame when the trainer and prompter are different people. Is it the trainer's fault for producing a machine capable of infringement or the prompter's fault for eliciting infringing output? Even there, I think there are existing legal doctrines (e.g. contributory infringement) that can cover it without having to make up a whole new copyright system for AI.

Is it free software?

Posted Feb 21, 2026 12:59 UTC (Sat) by draco (subscriber, #1792) [Link] (23 responses)

> I feel like the discourse around AI and copyright could be simplified a lot if you ask the question "would this be copyright infringement if a human did it".

I'd agree completely if these AIs were actually intelligent enough to obey copyright law on their own. (I think there's a legitimate philosophical question whether it's ok to delete a true artificial intelligence once you've created it, since that's the equivalent of killing it.) Since LLMs aren't actually intelligent and aren't trained in a way that makes it possible to inherently obey copyright law, I feel it gets a lot murkier.

Is it free software?

Posted Feb 21, 2026 22:50 UTC (Sat) by Wol (subscriber, #4433) [Link] (3 responses)

> I'd agree completely if these AIs were actually intelligent enough to obey copyright law on their own.

Are humans intelligent enough to obey copyright law on their own? Take Bach and Beethoven for instance - how much of their work that is commercially available is protected by copyright? Most of it?

> Since LLMs aren't actually intelligent and aren't trained in a way that makes it possible to inherently obey copyright law, I feel it gets a lot murkier.

Well, Natural Intelligence is trained by bumping off versions that get it wrong ... a major part of the problem with Artificial Intelligence is there's no way correcting it when it gets it wrong?

Cheers,
Wol

Is it free software?

Posted Feb 23, 2026 17:00 UTC (Mon) by rgmoore (✭ supporter ✭, #75) [Link] (2 responses)

Take Bach and Beethoven for instance - how much of their work that is commercially available is protected by copyright?

Music is unusual because it has separate copyright for the score and the performance. None of Bach's or Beethoven's original scores are still under copyright, but most of the performances- basically all of the ones using remotely modern recording technology- are still under copyright. So if you want to use Beethoven or Bach for your movie, you could avoid any copyright problems by using the original score (or a very old arrangement) and having it newly performed, either yourself or as a work for hire.

Is it free software?

Posted Feb 23, 2026 22:14 UTC (Mon) by Wol (subscriber, #4433) [Link] (1 responses)

Don't forget that many modern scores are also under copyright.

So yes, if you find an old-enough score to copy you're safe, but if you go out and buy a score it's quite likely protected. It may not be the notes that are protected, but all the stuff around it, the typesetting, the font, the layout, etc etc.

If a human would have trouble getting all that right, I dread to think what a mess an AI would make of it ...

Cheers,
Wol

Is it free software?

Posted Mar 18, 2026 22:53 UTC (Wed) by sammythesnake (guest, #17693) [Link]

In those cases, it's the score as drawn, rather than the music it is a written form of. I need permission to make a photocopy of the score, but performing the music doesn't convey any element of the score that wasn't in the (public domain) music, so there's no copyright issue making a new recording of it.

Is it free software?

Posted Feb 21, 2026 23:24 UTC (Sat) by turistu (guest, #164830) [Link] (6 responses)

> I think there's a legitimate philosophical question whether it's ok to delete a true artificial intelligence once you've created it, since that's the equivalent of killing it.

There's nothing "philosophical" about that kind of nonsense.

Only people have rights, and they keep having all their rights no matter how stupid, feckless, machine-like repetitive, unimaginative and worthless they appear. Bots, corporations or other software or hardware contraptions have no rights.

Assuming that someone was ever able to build an intelligent machine which was not reproducible, you could have an argument for protecting it from deleting/"killing" it in the same way historical buildings, paintings, old books or pieces of tapestry are. Not in the same way people are.

Is it free software?

Posted Feb 23, 2026 13:01 UTC (Mon) by kleptog (subscriber, #1183) [Link] (5 responses)

> Only people have rights, ...

It's philosophical because that statement is a belief that you have. There is a school of thought called anthropocentrism that believes this. But there are alternatives, many people believe that animals also have rights (e.g. protection from cruelty by humans).

You may not agree with it, but denying the existence of other viewpoints is pretty extreme.

For the time being whether an AGI would have rights is philosophical because the actual existence of an AGI hasn't happened yet.

Is it free software?

Posted Feb 23, 2026 20:32 UTC (Mon) by draco (subscriber, #1792) [Link] (4 responses)

I wasn't even thinking that far in my comment (regarding animals having some rights).

If intelligent aliens came to visit us in spaceships, would they not be [non-human] people too? They wouldn't share our genes, but they would be intelligent.

Is that enough? I'm pretty sure a lot of people would automatically assume so.

So...if we make machines as intelligent as that (or us), what about them?

Like you said, this is all hypothetical right now, since intelligent machines don't exist yet, no matter what the LLM boosters say. But eventually we'll figure out how to make them, and if we don't have good answers for these questions when that happens, we're frankly idiots considering the considerable media discussing the possibility (and consequences for ignoring the questions).

From what I've seen the last few years...I don't have high hopes. :-/

Is it free software?

Posted Feb 23, 2026 22:17 UTC (Mon) by Wol (subscriber, #4433) [Link]

> From what I've seen the last few years...I don't have high hopes. :-/

From what I've seen in the last few years, I have very "low" hopes of our ability to create anything remotely intelligent :-)

I don't think we'll have any problems with philosophical problems of deleting AIs at the moment :-)

Cheers,
Wol

Is it free software?

Posted Feb 23, 2026 23:46 UTC (Mon) by rgmoore (✭ supporter ✭, #75) [Link]

We already recognize the idea of legal persons who aren't natural human beings, like corporations, so it wouldn't be a stretch to grant personhood to other things. Of course this is making the huge category mistake of behaving as if there is only one legal system with a single set of well established rights. Each country has its own legal system, so it's completely possible that different places would recognize AGI (or extraterrestrial beings) differently.

Is it free software?

Posted Feb 24, 2026 14:52 UTC (Tue) by paulj (subscriber, #341) [Link] (1 responses)

> They wouldn't share our genes,

.... or would they?

;)

Asimov's robots series tackles robot sentience and equality in at least 1 story - "Bicentennial Man". The story won a SF writing award. Also made into a film. Some love it, some don't - bit sickly sweat at the end, but it's good IMO.

Is it free software?

Posted Feb 24, 2026 15:12 UTC (Tue) by Wol (subscriber, #4433) [Link]

Or E.E. (Doc) Smith and the Lensman series. The Arisians, whether by design or accident, seeded the galaxy with "human-ness", so when the galactic collision occurred that filled both galaxies with planets happened, they mostly got seeded with Arisian-like life.

(Yes I know much of the science in the Lensman series is now known to be rubbish, but as Doc Smith said, he did his best not to break any of the scientific laws that were known/understood at the time. Much of his imagination has not withstood the march of Science :-)

Cheers,
Wol

Is it free software?

Posted Feb 22, 2026 14:22 UTC (Sun) by ptime (subscriber, #168171) [Link] (1 responses)

It’s not murder to kill mildew in my shower with bleach, even if I’m the one who let the mildew grow to begin with. Deleting a model file is no different than scrubbing away some nasty mildew, and we should celebrate those who do.

Is it free software?

Posted Feb 23, 2026 20:14 UTC (Mon) by draco (subscriber, #1792) [Link]

Last I heard, no one ever claimed mildew was intelligent ("truly" or otherwise).

I never claimed LLMs were intellegent, or anything near it, so I don't see how your comment is relevant.

Is it free software?

Posted Feb 23, 2026 11:31 UTC (Mon) by farnz (subscriber, #17727) [Link] (9 responses)

There's also a philosophical question about whether the current round of AIs should be responsible for obeying copyright law on their own, or whether they should be treated as tools like text editors and Napster.

If they're responsible for obeying copyright law, then it becomes a requirement on AI vendors to make them intelligent enough to obey copyright law on their own. That is obviously something that AI vendors will push back on.

If they're a tool, like a text editor or Napster, then it comes down to how it's used; if it's "clearly" intended to permit copyright infringement (as the courts in the USA found that Napster was), then the vendor is liable. Otherwise, the user is liable for the resulting infringement - and the first time there's an AI user who loses a large sum of money to a court judgement of infringement, the AI vendors face scary press.

The third option is the one the AI vendors want, because it's good for them: AI can't obey copyright law by itself, so isn't responsible for doing so, but also is sufficiently transformative of the input that the output of an AI tool is a new work, not a derivative of any past work and thus the AI user cannot lose a large sum of money to a court judgement.

Note, too, that if you're not able to pursue copyright infringement claims, you effectively don't have copyright protection - there's no practical difference between "farnz copied your work in full, and you didn't pursue" and "farnz copied you work in full, and you lost when you pursued".

Is it free software?

Posted Feb 23, 2026 11:41 UTC (Mon) by paulj (subscriber, #341) [Link] (6 responses)

This may be one of the biggest impacts ABNI has on society - making the carefully constructed walled-gardens of copyright cartels (yes, copyright also gave some protection to "the little guys" - though, ability to access justice was limited for those without deep pockets, and overall the large and powerful built systems around copyright that protected themselves and punished others) obsolete, by using the slice-and-dice ABNI machines to get around copyright. (Whether ABNI in principle /really/ does get copyright or does not may not be relevant - what matters is there are very large, well resourced, well connected entities who /want/ ABNI to get around copyright, and so it probably will).

Is it free software?

Posted Feb 23, 2026 15:01 UTC (Mon) by mathstuf (subscriber, #69389) [Link] (3 responses)

Could you expand "ABNI" please? I'm not finding anything it stands for that matches the conversation here.

Is it free software?

Posted Feb 23, 2026 16:05 UTC (Mon) by paulj (subscriber, #341) [Link] (2 responses)

ABNI -> "Artificial But Not Intelligent".

ABI -> "Anything But Intelligent" another contender, but has a predefined meaning, least amongst the kinds of technical people who read LWN.

Is it free software?

Posted Feb 23, 2026 16:28 UTC (Mon) by Wol (subscriber, #4433) [Link]

My favourite - Artificial Idiocy :-)

Cheers,
Wol

Is it free software?

Posted Feb 23, 2026 17:14 UTC (Mon) by rgmoore (✭ supporter ✭, #75) [Link]

I've also seen PI ("Pseudo Intelligence") used in science fiction. It seems like a good term because it gets the idea that it's an attempt to create intelligence but doesn't respect it as actually intelligent. Unfortunately, those initials have other established meanings and might trip people up.

Possible impacts of AI vendors

Posted Feb 23, 2026 15:38 UTC (Mon) by farnz (subscriber, #17727) [Link] (1 responses)

There is a much more likely outcome on that path - AI is deemed a tool capable of being used for infringement, but not automatically infringing (so it's case-by-case "did this output infringe"), and the big AI vendors get round this by paying the cartels a "permission fee" for "minor" infringement, while the big companies continue to sue over blatant infringement, and "the little guys" get steamrollered because they aren't big enough to threaten the AI vendors' income streams.

As long as the AI vendors can buy their way out of trouble for reasonable money, they will do so. And there's a route to that where the AI vendors pay to cover "accidental" infringement (e.g. a very Star Wars looking "robot flying a spacecraft"), but not infringement where it's "obvious" from the prompt that the user wanted the output to infringe (e.g. "an R2 series astromech droid flying an X-wing starfighter"), where the AI vendor instead forwards prompt, user details, and output to the big copyright holders so that they can pursue the users for infringement.

And that's a situation where the "little guys" get steamrollered. If you can prove that something is infringing on your copyright, then the fact it was made with Google Gemini using Nano Banana doesn't protect the person who used it from legal consequences, but you first have to find out that this is happening, and then take action. Meanwhile, the owners of the Star Wars copyrights (currently Disney) get a fast track to discovering that people are creating infringing material that's derived from their copyrighted material, and all the information they need to sue people on a platter, plus an income stream to cover cases where the AI vendors didn't notice that they were infringing, where the AI vendor is indemnified.

Possible impacts of AI vendors

Posted Feb 23, 2026 16:16 UTC (Mon) by paulj (subscriber, #341) [Link]

Yes, that seems likely.

The "big guys" in the AI and content fields will continue to find ways to have their interests protected. I do think the AI players will have more influence, and there will be some tilting of how copyright works in favour of the interests of the pushers of the ABNI slice-and-dice machines. There may will be a private arrangement of a truce between the major players in those 2 arenas - by blanket licenses (which will of course indenmify the ABNI slice-and-dicers fully, but not necessarily the users - agreed) - agreed.

The "little guys" will be screwed by both though.

The major content cartels will use ABNI to minimise need for digital artists, re-using all the little bits of stuff that was fed in to the slice-and-dicers and recombined - some of which came from smaller content creators. The major ABNI pushers will make money selling their slice-and-dice machines, and their recombination of humanities prior work to all.

It will be interesting to see how the next generation navigate this world.

Is it free software?

Posted Feb 23, 2026 16:42 UTC (Mon) by rgmoore (✭ supporter ✭, #75) [Link] (1 responses)

The third option is the one the AI vendors want, because it's good for them: AI can't obey copyright law by itself, so isn't responsible for doing so, but also is sufficiently transformative of the input that the output of an AI tool is a new work, not a derivative of any past work and thus the AI user cannot lose a large sum of money to a court judgement.

I don't think that's going to work out. There are really two separate question: is the AI itself a derivative work, and is the output of the AI a derivative work. This is how considering the parallel situation of a human being doing the same thing is instructive. A human being is allowed to learn from existing copyrighted works without problems, but there are standard legal tests to determine if their work is infringing. The same thing should be true of AI. The AI itself is almost certainly sufficiently transformative that it isn't a derivative of its training corpus, but individual outputs can infringe copyright depending on any of the standard tests. One minor wrinkle is that some of the standard tests are easier to prove if the alleged copier had access to the work they're accused of infringing, and that's probably going to be easier to prove with AI than a human. I wouldn't be surprised if AI companies just stipulate their AI had access to avoid having to divulge all their training material.

Is it free software?

Posted Feb 23, 2026 16:49 UTC (Mon) by farnz (subscriber, #17727) [Link]

I agree that it's the least likely to work out - that's why it's a non-answer to the question I posed of "whether the current round of AIs should be responsible for obeying copyright law on their own, or whether they should be treated as tools like text editors and Napster".

It's instead an option that they'd like, but that basically says "the question doesn't need answering, because it's neither a tool nor responsible for obeying the law". In practice, I think the models and weights will be deemed non-infringing (by legislative fiat if needed), but the outputs will be potentially infringing, and the law will have to find an equilibrium.

Is it free software?

Posted Feb 23, 2026 15:02 UTC (Mon) by kleptog (subscriber, #1183) [Link] (1 responses)

> Current AI models aren't capable of that kind of exclusive single-project training (yet),

Depends on what you mean. It's fairly straight-forward (and much cheaper) to take an existing model and then fine-tune it. That's how you make domain specific LLMs. It would seem perfectly feasible to make an LLM that was designed to answer questions about a single software project by tuning it with the source for that single project. Whether that would actually give you better results than a generic LLM I don't know. Such an LLM could probably reproduce the original project with higher probability than a generic LLM.

Ideally you'd want a model that understood language, but didn't have any facts other than the ones you specifically give it. But I don't think you can understand a language without knowing something about the real world. I think only the language of math is sufficiently abstract that it could be understood without knowledge of the real world.

IMO the question of whether LLM output is copyrighted will never be a blanket yes/no but depend on the situation. We need to keep an eye on the goals of copyright law (e.g. allowing creators to live of their creations) rather than the machinery (copyright law itself).

Is it free software?

Posted Feb 23, 2026 20:57 UTC (Mon) by NYKevin (subscriber, #129325) [Link]

I was referring to an LLM that is trained entirely and solely on the GPL'd project and no other corpus. Such LLMs generally don't exist, and may not be feasible to create at all.

But, of course, there is more to machine learning than LLMs. You could train something dumber on a single project, and perhaps it would be useful for some narrow task such as correcting indentation and other stylistic problems (when used in combination with a language server and other non-AI supporting infrastructure).

Is it free software?

Posted Feb 20, 2026 15:41 UTC (Fri) by epa (subscriber, #39769) [Link] (18 responses)

Yes, if it tries to impose additional restrictions beyond copyright law then it’s non-free in my view. A licence should be a grant of permission to do something that’s otherwise reserved to the copyright holder. I think it would be safer to say

“It is not yet settled whether, if you train an AI model on this source code, the resulting model is a derivative work of the code. But if it is, and does not fall under ‘fair use’ in your jurisdiction, then as with any other derivative work you may only distribute it under GPLv2.”

Is it free software?

Posted Feb 20, 2026 16:13 UTC (Fri) by dskoll (subscriber, #1630) [Link] (16 responses)

Thank you for that comment. I have updated the README.md file to reflect it.

I don't intend for Remind to be non-Free, but I still want to prevent it from being used to train LLMs whose outputs is not GPLv2, to the greatest extent permitted under copyright law.

Is it free software?

Posted Feb 20, 2026 18:35 UTC (Fri) by rfontana (subscriber, #52677) [Link] (14 responses)

So the new language is:

> It is not yet settled whether, if you train an AI model on this
source code, the resulting model is a derivative work of the code. But
if it is, and does not fall under "fair use" or equivalent in your
jurisdiction, then as with any other derivative work you may only
distribute it under the terms of the GNU General Public License,
version 2.

That seems OK to me, but I think it's a different kind of assertion than what it replaces since here you're talking about the "resulting model" possibly being a GPLv2 derivative work, while before you talked about the "output of the model" or "anything the model produces". Not sure if that change in focus was intentional or not. Even if a model trained on GPLv2 stuff is a derivative work of that GPLv2 stuff, it doesn't mean the output would also be a derivative work (and vice versa).

Is it free software?

Posted Feb 20, 2026 20:09 UTC (Fri) by epa (subscriber, #39769) [Link] (1 responses)

Yes, that’s another open question. I think it would be hard to argue that the output of ChatGPT is a derivative work of the model definition, without at the same time admitting that the model is derivative of its training data.

Is it free software?

Posted Feb 23, 2026 8:49 UTC (Mon) by taladar (subscriber, #68407) [Link]

I don't think anyone ever contested that it is derivative of the training data, just to which extent it is derivative of any particular piece in the training data and if that meets the threshold for copyright to become an issue.

Technically speaking you could e.g. make a model that is a single bit and reflects if the input has an odd or an even number of characters and produces output with the same characteristics. Most people would agree that that would not violate copyright.

On the other end of the spectrum you could just make a "model" that is a filesystem folder of all the input data and the algorithm just selects a random section from the input verbatim. Most people would agree that that should violate copyright.

The question is where in between those two extremes do we draw the line where copyright no longer applies.

Is it free software?

Posted Feb 20, 2026 21:31 UTC (Fri) by dskoll (subscriber, #1630) [Link] (11 responses)

I might change "the resulting model" to "the resulting model and the output it produces", though I think there's already American precedence stating that the output of an AI model is not copyrighted.

As I said, I don't want Remind to be non-Free. But it seems to me that LLMs can be used for "copyleft washing" whereby you pass copyleft code thought an LLM and then the results, even if very similar to the training input, are somehow magically no longer copyleft. This seems like an underhanded trick and I'm trying to express my disapproval of it.

Is it free software?

Posted Feb 23, 2026 7:28 UTC (Mon) by gfernandes (subscriber, #119910) [Link]

I think it's perfectly fair for you, the author, to express your intent that the license covers human use and not AI/LLM training.

It has been OK to dial license the GPL since roughly forever. That is exactly what you have done. You should not have to justify this.

Is it free software?

Posted Feb 24, 2026 0:31 UTC (Tue) by rgmoore (✭ supporter ✭, #75) [Link] (9 responses)

though I think there's already American precedence stating that the output of an AI model is not copyrighted.

That's right. Under US law, a work must show a degree of creativity to qualify for copyright. Under current precedent, machines and non-human animals are assumed not to have any creativity, so their works are inherently unable to be copyrighted. Just because something can't be copyrighted, though, doesn't mean it can't violate a copyright. For example, the official work of US government employees is automatically in the public domain, but there have been cases where artists have successfully sued the US government for copyright infringement.

Is it free software?

Posted Feb 24, 2026 14:50 UTC (Tue) by anselm (subscriber, #2796) [Link]

Here in Germany, copyright law stipulates that any works must be the “personal mental creation” of a human in order to qualify for protection. Hence, at least for now, no copyright protection for generative-AI output in the Land der Dichter and Denker (country of poets and thinkers).

Is it free software?

Posted Feb 24, 2026 15:08 UTC (Tue) by Wol (subscriber, #4433) [Link] (7 responses)

The obvious way to think of it is that an AI cannot claim creation rights over its work. Aka as commented "the ouput of an AI is not copyrightable".

But that clearly does not preclude the AI *copying* a copyrighted work, and hence infringing copyright. Going back to that "Bach and Beethoven" example, here it's inside out, but the principle is the same, with the music the music itself is copyright-free but has a wrapper of copyrightable arrangements, fonts, layout etc. With the AI, the AI wrapper is copyright-free, but the contents may well be copyrighted.

Cheers,
Wol

Is it free software?

Posted Feb 24, 2026 21:47 UTC (Tue) by rgmoore (✭ supporter ✭, #75) [Link] (6 responses)

It would be very strange if the output of an AI were somehow incapable of infringing copyright. The output of a machine everyone can agree is not intelligent, like a film camera or a photocopier, can clearly be infringing. The output of a human being, something everyone agrees is at least capable of being intelligent, can be infringing. Why should the output of something in the middle suddenly be incapable of infringing? The real question isn't whether the AI's output can be infringing but whether the AI itself is responsible for infringing output or if the humans around it are. Right now, it's pretty clear "AI" isn't actually intelligent enough to be capable of infringement on its own, so the liability should rest on the humans.

In practice, even if the courts decided and AI was capable of infringement, the humans around it would still wind up liable. The AI would probably be treated the same way as an employee is, and businesses can be held liable for their employees' misdeeds when those employees are doing their official duties. You could wind up with several levels of liability. The company operating the AI would be primarily liable, but the company that created it could be secondarily liable for contributory infringement. Unless/until we have free existing AIs that do things on their own without any kind of human prompting, liability will eventually wind up with the people who keep the AI running and the ones who provide it with the prompts that elicit infringing outputs.

Is it free software?

Posted Feb 25, 2026 9:51 UTC (Wed) by farnz (subscriber, #17727) [Link] (5 responses)

The case of concern is more around how the legalities wind up shaking out.

It's easy to imagine a scenario in which the AI model is not itself responsible for infringement, nor is it inherently a derived work, and where liability therefore falls on a mix of the AI vendor and the user.

In that world, the big copyright holders are in a strong position to negotiate a deal with the AI vendor, where the AI vendor is paying for licensing for "minor" infringement, and passes details of "deliberately" infringing prompts and outputs to the big copyright holders for legal enforcement.

That's a worst case for FOSS. We're not big copyright holders, so we don't get into the deals for "minor" infringement and "deliberate" infringement. But our output is still used to train the models, and to make the AIs profitable - meaning that it's now on us to find the people infringing our copyright, and pursue them in court (giving them the option of bringing the AI vendor in as a contributor to their infringement) to stop AI becoming just another way to infringe FOSS copyrights without significant risk.

The history of FOSS licence enforcement (or rather, the lack thereof in most cases) does not fill me with hope here - the cost of blanket enforcement is high, and that's what's needed to make the AI vendors worry about the risk of infringing our copyrights.

Is it free software?

Posted Feb 26, 2026 11:15 UTC (Thu) by kleptog (subscriber, #1183) [Link] (4 responses)

I'm a bit confused about what kind of infringement you're worried about?

Are you worried that somebody could use an LLM to "vibe-code" a competitor to PostgreSQL while it not looking like PostgreSQL, purely because the source is in the training set? What FOSS is out there that people would like to copy, don't feel like they can just get away with it but figure they can use an LLM to get around it?

What you are referring to with big companies I do see as a thing for purely creative works. To prevent LLMs and Stable Diffusion from producing video clips with X-Wing fighters and Mickey Mouse. Though to me it this feels more trademark than copyright related. ISTM non-commerical stuff should be allowed anyway.

But the value of copyright on source code has always been a bit weird. Code is mostly functional. For any given problem there really are only a few good solutions. Sure, you can split the functions up in different ways, the variables can have different names, you can use a different language. If you want to write a new program, in almost always you are not doing anything new, just doing what other people have done before, but framed in a different way.

The GPL is mostly a political tool. We don't care if the FreeBSD guys want to take the source of some driver the make a better driver for FreeBSD. Most of the code in the kernel is useless in any other context. We care that someone takes the Linux kernel as a whole. The tool we are given is copyright law, so we have the GPL.

If someone takes an algorithm in the Linux kernel and uses it in their own product, that's not copyright infringement. ISTM LLMs will mostly take the functional structure and ignore the literal text. There where their value is after all.

But maybe there is some risk to FOSS from LLMs that I'm missing. I'm just not seeing it.

Is it free software?

Posted Feb 26, 2026 11:19 UTC (Thu) by farnz (subscriber, #17727) [Link] (3 responses)

If all free software is in the training set, and the output of the LLM is effectively not copyright protected (because it's too expensive to prove), then things like OpenWRT would never have happened, because you could avoid the problem with GPL licensing by having the LLM "vibe-code" an entire router OS, instead of copying Linux.

It's not a new risk. But if you're looking at LLMs as "they will obliterate copyright walls", I think you're barking up the wrong tree; I expect them to be bad for small copyright holders (since proving copying becomes harder), but not big ones.

Is it free software?

Posted Feb 27, 2026 12:27 UTC (Fri) by taladar (subscriber, #68407) [Link] (2 responses)

At least right now it is hard to get an LLM to move a function from the file foo/bar to the file foo/baz without modification but a hypothetical future AI based on a different fundamental concept might be usable that way.

Is it free software?

Posted Feb 27, 2026 12:32 UTC (Fri) by farnz (subscriber, #17727) [Link] (1 responses)

For someone abusing LLMs to "copyright-wash" infringement, that modification is a feature, not a bug. The copying is no longer literal, so it makes it harder to show that copying took place, while as long as the modifications don't introduce too many new bugs, you can refer back to the original and fix them (or ignore them if they're irrelevant to your product - if you're building a WiFi router, and the bugs relate to DCCP NAT, "just" don't support that).

Is it free software?

Posted Feb 27, 2026 12:58 UTC (Fri) by taladar (subscriber, #68407) [Link]

In the specific case I had in mind the LLM just kept changing APIs back from the version of a library I was using to the version of the same library that was current when it was trained (bevy 0.17 -> 0.14) which was incredibly annoying.

Is it free software?

Posted Feb 20, 2026 20:10 UTC (Fri) by epa (subscriber, #39769) [Link]

Thank you for changing that. I believe it makes explicit what is the legal position for any GPL-covered work.

Is it free software?

Posted Feb 20, 2026 16:26 UTC (Fri) by Wol (subscriber, #4433) [Link]

Or even, in LICENCE.TXT or COPYING.TXT something as simple as

"It the opinion of the authors of this code, the output of any AI model trained on this code is a derivative work of this code. As such, we believe any and all output of such an AI model must be licenced under the GPL".

That would then make life somewhat risky for any bot-operator just slurping the web :-) EU/UK law says using slurped data to train your model is perfectly legal (imho they're explicitly treating it exactly the same as a human reading copyrighted books). But the law says nothing about the output (again imho treating it exactly the same as the output from a human who has read copyrighted books).

And that gives you as the author of the code extremely good grounds for saying "Well if you copy the code you should have checked the COPYING file. What! You didn't! More fool you!".

It's all very well for a defendant to say "there wasn't a copyright notice", but a lot harder to argue "innocent infringement" if it's a case of "I didn't look for a copyright notice"!

Cheers,
Wol