>The full case against LLMs — how they’re built on unconsented labor, sold through hype, and coercively pushed onto workers — deserves its own post, which I’ll link here once I publish it.
I don't dispute that the frontier LLM companies stole vast amounts of intellectual property ... like many orders of magnitude more IP than people who are spending years in jail for piracy. That's indisputable.
But when you start equating piracy (I'm reading your book without paying for it) to SLAVERY (enforced labor) ... that's just so completely ridiculous that I find it difficult to take anything this guy writes seriously.
Did you read what I quoted? It wasn't even from the earlier slavery part (that you're talking about).
It came much later ... but it nevertheless used the term "unconsented labor". When you use language like that ... after talking explicitly talking about slavery earlier ... it's really hard not to see it as the author contrasting piracy and slavery.
> but it nevertheless used the term "unconsented labor"
Which you quoted as "enforced labor". Now you could say that "unconsented labour" is not really a thing, but to be fair, "enforced labor" isn't either. And if we want to nitpick, "forced labour" (which implies coercion) is not "slavery" (which goes further and implies ownership).
Now back to genuinely trying to understand what the author meant, I would think that "unconsented labour" means that the result of the work was used without the consent of the person who owns it. Which is not forced labour and even less slavery.
It drives me crazy how people genuinely cannot understand what analogies are meant to do.
NO, they DID NOT equate piracy to slavery.
They made a thought experiment to convey a point. "Imagine that we could benefit from something that we all agree was produced by unethical means, would it be ethical to do it?".
When someone says that "DNS is a bit like the phonebook but for websites" and you say "this is UTTERLY wrong, DNS is NOT made of paper", then you are the one being completely ridiculous.
Author explicitly talks about slavery at the start of the article, then later (separately from the slavery stuff) uses language like "LLMs ... are built on unconsented labor" ...
... and you don't think he's drawing any parallels whatsoever between slaves and how LLMs are built (ie. from piracy)?
I don't think he is, I would not interpret "unconsented labor" as "forced labor." It's an awkward phrase but if he meant the latter he would have said it. I read it as meaning LLM training takes advantage of work that was done in the past and reuses it commercially with no attribution.
> Author explicitly talks about slavery at the start of the article
The word "slavery" appears 3 times in the article. Once in the quote (the analogy), and twice when clarifying that it was an analogy and nobody actually supports slavery (because, you know... people don't understand how analogies work).
> then later (separately from the slavery stuff) uses language like "LLMs ... are built on unconsented labor"
If the copyrighted work of an author serves as a building block for LLMs without the author agreeing to it, I think it's not completely weird to refer to that as "unconsented labour", is it?
> and you don't think he's drawing any parallels whatsoever
I don't.
Spoiler alert: when people talk about "piracy" in this context, there are no boats and no swords involved. Even if they happen to say "that ship has sailed" a few paragraphs after mentioning piracy.
> Freedom for users, not just programmers. Source code access means little if you can’t read it. Real software freedom has to include usability, accessibility, documentation, and support in the languages people actually speak.
Programming has gotten much more accessible with LLMs. This will eventually result in more user freedom in practice, once forking free software becomes the norm for end users. But custom software needs to be managed, so we will need hosting services and software installers that make finding, downloading, and customizing apps easy.
And running marathons has gotten much more accessible with automobiles.
I'm not really sure what your other point was. There's no universe in which a LLM can somehow correctly fork a project for a different purpose but somehow still needs help in hosting, publishing, building an installer or a ui to do so.
Relevant comment from Stefan Monnier (former emacs maintainer) on emacs-devel:
> The issue I have with this is that it moves the goal from "defend Free Software" to "defend against generative AI". The Free Software movement's goals are largely aligned with many other social movements, yet to avoid fragmenting our community, the FSF has been very careful to try and stick to software freedom (and let individual members combine this with any other issue they like). I'd rather leave the fight against slop to other groups, even if there might be a lot of overlap between that group and ours.
I don't have a strong opinion, but I wanted to say two things:
1. If Free Software must be ethical and therefore refuse contributions by LLMs, should Free Software refuse contributions made on hardware that was itself made of components that were not ethically-sourced? Should contributors prove that their laptop is made of ethically-sourced materials? Good luck with that, I guess?
2. Now that LLMs are there and are undeniably multipliers, isn't that an opportunity for Free Software? Reverse engineering proprietary firmwares used to take years, now LLMs make that a lot faster. In a way, is it bad to "use the weapon of the enemy in the fight against themselves"? I would think that this is debatable?
When (Theravada) Buddhist monks go out each morning to accept food donations (Dānasantuṭṭhi) from the laity, they do not exclude donations from anyone
In the Pali Canon, the Buddha accepted alms from murderers, thieves and courtesans
Monks follow this example in practising Upekkha, impartiality
(Of course, if the food is bad or inedible they will not consume it but the source of the donation, and how the giver was able to obtain the food donated, is not considered)
Unlike open source, Free Software (capitalized) was intended from its inception a political movement, so of course politics matters to it. Which way the politics should matter is of course a valid topic of discussion.
> Free Software (capitalized) was intended from its inception a political movement
I completely disagree. Free Software is an ideological movement based on a deeply pragmatic viewpoint of software use (not using software that does not respect your freedom). It has nothing to do with politics and is open to everyone of any political persuasion to use.
> Please avoid using the term “open” or “open source” as a substitute for “free software.” Those terms refer to a different position23 based on different values. Free software is a political movement; open source is a development model.
I think you are discussing a related but quite different topic here (worth discussing, for sure, but I'm taking issue with your phrasing "I wrote about this paradoxical approach of the free software movement").
The post discusses about how the code gets written collectively, yours discusses what the code can be used for. Yours discusses potentially desirable limitations on freedom 0 (and indeed, contradicts the accepted definition of free software - that's not a judgement, it's a fact). The other post discusses being more demanding on how free software is made in a way that doesn't contradict the free software definition, and (that's where both are indeed related I think) whether we as maintainers should have our shoulders to the software freedom wheel a bit less and consider more the bigger human freedom picture (and I guess my phrasing gives away what I think about this).
Restricting freedom 0 is tempting but perilous and can have consequences you have not anticipated and might not like. One can totally agree with your goal (avoiding an uber-like situation for workers) but not with how you do it. It's not the same ethical / philosophical stance as one who considers other ethical matters than software freedom out of scope (and both stances are actually not incompatible).
The author is clearly against AI models that are trained on data produced by people without their consent. Yet the whole free software movement is built on the notion that proprietary software is unethical and any attempts to protect it (and other content like movies and music) using DRM or other mechanisms are wrong.
> [...] trained on data produced by people without their consent. Yet the whole free software movement is built on [...]
consent. The whole free software movement is built on consent. For the most part, that consent is conditional - like requiring that you pay it forward (GPL), or at least mention who the author is (MIT).
I think I see what you're saying, but no, I don't think it's a contradiction.
I think the worry from the "free software movement" side is that someone consentually shares some code, then a company takes it, profits from it, but "locks it down" so their additions are not likewise shared. So basically the company is freeloading on what others produced, taking but not giving back — that's what people frown upon.
Likewise, training on data without people's consent is also a type of freeloading, and is frowned upon in a similar fashion.
Note that this happens with humans, too: if an artist takes too-direct "inspiration" from another without getting credit, they are called out.
I suppose there are some more absolutist stances, like "abolish copyright" (rather than the milder "abolish freeloading" described above) which would argue against DRM, etc. of any kind, full stop. And, if such a world ever came to pass, and those people felt that their data (artwork, code, etc) should not be available for AI training, that would perhaps be a contradiction, yes.
Well, it's not necessarily a contradiction, as the copyright protection and licensing aspects are separate. They don't support what they see as copyright infringement, but they also don't think that copy-protection schemes are good for the user, since they also prevent legitimate uses under the law. Also note that the position of the FSF et al. is that there's no reason that art (I mean this in the traditional sense of art: books, music, movies, etc.) can't be proprietary, but that programs are different.
Not necessarily, becase the FSM is not about "freeing" software. Its about "free" software. They're not cracking DRM. They're not trying to pirate other software.
Rather they're arguing that software should be free, they work on free software, and then promote their free software over non-free software.
Basically, everyone involved in the FSM is consenting.
With AI IP scraping, that's not so much necessarily the case.
> Their argument was that in the context of Emacs development, the only thing that matters is the quality of the code being contributed
This is something I've never agreed with. One part of open-source is to have some principle that is rooted in ethics, but how can you be true to that when you ignore other ethical dimension that can't be separated from the traditional open-source one?
I don't think the only think that matters is the quality of the code. It's a myopic viewpoint. The reasoning behind having open code in the first place matters as well, which was that one person was trying to take advantage of another in a way that seemed intuitively unfair. LLMs do the same thing. If you fight for only one ideal but reject another parallel one for convenience, then I believe the first stands for nothing at all.
Over 20,000 people have contributed to the Linux kernel and there are around 2,000 contributors per release. How can it be viable to judge whether each contributor is overall a good person?
One way to make moderation practical is by limiting the moderators’ jurisdiction to the things they have direct evidence for. For example, presumably the moderators of Hacker News limit their scope to what people do in Hacker News, rarely going beyond that to looking at behavior in other forums or what people do in their personal lives.
Similarly, moderators of an open source project could mostly limit their scope to what people do when interacting with the project’s systems and community.
You can't judge whether everyone is a good person. But you can put in basic rules that obviously go against immoral behaviour. Not everyone obeys the "no shirt in the shopping mall" rule but it's there and it's a good one.
Many open source projects have a Code of Conduct, for a reason. And people who break it are called out. Many open source projects are anti-AI as well like YT-DLP and Asahi Linux. It can be done and easily without much fuss.
Contributors who use AI for some other unrelated project wouldn't be called out, though? The code of conduct only applies to what people do as contributors.
So what is the "no shirt" rule for OSS development? It seems like it's basically Google's "do no evil", right?
Now, Google itself stopped following that years ago, which offers a hint as to the problem. We can say "OSS contributors can't be evil" ... but we can't agree on what "evil" should be defined as.
I am a software, systems, and security engineer with two decades of experience so I could already build just about anything by hand, so all I require from an LLM at this point is speed of execution on my specific instructions.
I could run larger models with fewer sessions, but with my cluster of 1x strix halo and 8x r9700s I get 13 sessions of qwen 3.8 flash next 135b at > 100tps each and that is the sweet spot for me to solve any problem in front of me from kernel and compiler bugs to writing hardware drivers to building specialized operating systems. I cannot think of any systems engineering problems I have seen in my career I cannot readily solve at home now on my own.
With radiance and/or strata a single strix halo or r9700 is enough for one session though for someone that wants to start on a smaller budget with less concurrency.
Stallman's reply to this all is really good, especially as this who discussion appears to be a contest of Who Is The Most Principled And Consistent and Stallman is the OG there and still very hard to beat. Some excerpts:
"The example that you're responding to is about where to draw the line
-- it was designed to be a borderline case."
"We don't reject contributions from companies such as Google and Amazon
which are intentionally subjugating users. These practices are not as
evil as outright enslavement, but they are plenty nasty nonetheless."
"Rejecting their useful contributions would not help us fight to stop
their unjust practices. We may as well accept the former, as long as
we don't mute our condemnation of the latter."
This does not make sense to me. Does the author use money? Is the author remotely aware of the way money is generated and multiplied? Yet once in circulation it can be used for totally honest transactions. Same with code, with the important distinction that code is not inherently a scam. So code should be cut more slack than money. Code is not evil, its use may be. If bad person curiously donates correct code, I keep it.
After the code of conduct , do we want a moral certification for code contributors? You do know where this mindset is leading, don't you?
Well, it's highly unusual for me to defend the mac, but I guess we'd be all cooked really, because i think horrible working conditions in mines concerns any computer hardware, and that's true for many other things as well. It's not specifically Apple. IIUC, your link seems to be about lawsuits against false claims from Apple (which is extra bad, of course, but others are not saints).
So, it's everyone's devices really.
Here's how I would resolve it: it is a real concern. We should fight against it, and maybe also work on reducing our systemic dependency on computers, use hardware as long as possible, limit the amount of hardware, and encourage reuse (and recycling when everything else fails).
But for now computers are more omnipresent than ever anyway so users having control over their computing and computers is also more important than ever. It's still a fight worth fighting despite software basically depending on horrible labor conditions. (And btw, software already depending on horrible labor conditions is not an excuse to get even worse)
Do you want to be right or do you want to be effective? As much as the article tries to claim it's not setting up a "moral purity test", that's what these sorts of hypothetical conundrums do. This is the exact type of privileged navel gazing that cause Democrats to lose the plot on what working class people in America had actually been experiencing, opening the door to much more blatantly obvious corruption and grift from Trump.
It's important for social and political movements to scope their ambitions to not alienate potential allies with overcooked positions. Build coalitions as well, but do so opportunistically where you have a chance to actually win something and move the needle. Don't craft your charter to already be so exclusive out of the gate that you only appeal to the like-minded fringe. You want to do something for software freedom? Talk about John Deere, not imaginary enslaved open source coders.
I honestly think the author of this article doesn't really consider the value of ideological interoperability. In real life, sometimes maintaining a narrow mandate is itself an ethical decision ESPECIALLY when that mandate allows people with incompatible worldviews to cooperate.
>Software has effects beyond the machine. Systems that decide who gets a loan, a job, or bail are freedom issues, whether or not the license is permissive.
Hang on to that a second, I'll come around to it.
Consider the single responsibility principle... OSS projects are organized around software freedoms as a well defined set of commitments, allowing contributors to disagree about almost everything else while still cooperating toward those SPECIFIC founding commitments. In daily life, when collaboratively making decisions other people have to live with, I think the narrower the shared commitment, the broader the range of people who can cooperate under it.
The moment you decide to step outside your founding ideology and take on more ethical commitments, you're now attempting to make broader decisions about which other ideologies are morally|ethically correct, creating new reasons for disagreement and potentially new reasons to exclude contributors. Today it's AI-assisted code, tomorrow it's defense employees and then maybe whether corporate sponsorship matches the project's values.
Really, it's easy to see how this causes fragmentation because there isn't a principled stopping point once the project's mandate becomes doing what is ethically right rather than protecting a particular set of freedoms.
I posit that it's COMPLETELY okay to refuse to take a stance on subjects that seem to sit outside the ethics of a project's purpose. There should be a clear delineation between an individual's ethical convictions and making those convictions binding on the group. It's a really blurry line especially since additional ethical positions can seem perfectly reasonable in isolation but collectively just add more friction to cooperation.
It's not the responsibility of the group to take a stance on every moral dilemma separate from its founding ethical stance unless it's existential to that ideology or necessary to fulfill, protect, or practically sustain the original mandate. In this case, I don't think it is.
And tbh, I'd expect leaders of OSS projects to respect the moral agency of contributors by declining to exercise moral authority over questions the group never collectively agreed to delegate to them.
Following that, I sincerely believe that code is absolutely all that matters but this is real life and abstractions can absolutely be leaky. Software really does have an effect beyond the machine. That might seem like an "aha, you see?" scenario but not really...
All that matters is code but code is also, at its most fundamental level, intention. The decision of what to build is actually the most tangible expression of a project's ethical purpose, even more so than whatever its founding principles say. The purpose of a system, after all, is what it does, not what it claims to do.
The code is the vehicle through which a project's particular vision of how the world ought to be is brought into existence. The software's effects beyond the machine are ergo not some separate ethical dimension the project has failed to consider, they're precisely why the project exists in the first place. The movement has already chosen the change it wants to effect in the world, and the software it produces is how it pursues that change.
So the project's ethical commitments are already embedded in the work itself, in what it chooses to build, what it chooses not to build, and the freedoms its software is designed to protect. Its moral direction is expressed through action.
I sincerely don't think that every moral issue connected to software must become part of the project's mandate. The real question is whether that issue materially bears on the particular change the project exists to bring about.
Even for individuals, it's never a good idea to attempt to have an opinion on everything. Sometimes, the most meaningful expression of value is to remain relentlessly focused on the one thing you're trying to accomplish.
I agree with the premise that the crux of the issue is consent, the movement is political, and the concerns largely of moral and political philosophy.
From that frame and on its own terms, there is a specific mechanism to context the conclusions and their moral authority: the dimension of where individual liberty (and hence consent) collides with the interests of the body politic and the society at large.
We can imagine the arguments that follow without needing to advocate for them: if you are a good faith accelerationist, the problem with this position is not that it is wrong on its own terms and given its assumed priors,
it's one of the moral foundation. The best interests of society may reasonably be argued to supersede the interests of the individual.
Whether one agrees with that premise generally, or with respect to AI accelerationism, or the ethics of building and using LLMs as we know them, or not, is a different question than the author addresses.
IMO the author's concern is premature. Or, alternately, assumes that this foundational question has an obvious or settled answer.
It does not.
Like it, or not, modern societies do in every dimension—legal, political, ethical—presume that the interests of the commons, of society, of the body politic itself, of the state, do in many places supersede those of individuals.
Libertarians disagree of course: and when they formulate and inhabit their own state, well, they shall be free to operate it on their own terms.
Until then and for the rest of us: the prior art here is against this position, at least the degree you believe one takes seriously that, variously, we are headed into a singularity from which various relatively utopian or post-scarcity outcomes may emerge; or, alternately, that the dystopian reality of a not just dead but actively hostile internet requires one deploy whatever tools are available and capable to survive it. (And the only tool we have to understand and defend against bad AI is good AI.)
>The full case against LLMs — how they’re built on unconsented labor, sold through hype, and coercively pushed onto workers — deserves its own post, which I’ll link here once I publish it.
I don't dispute that the frontier LLM companies stole vast amounts of intellectual property ... like many orders of magnitude more IP than people who are spending years in jail for piracy. That's indisputable.
But when you start equating piracy (I'm reading your book without paying for it) to SLAVERY (enforced labor) ... that's just so completely ridiculous that I find it difficult to take anything this guy writes seriously.
> But when you start equating piracy (...) to SLAVERY
They never equate it. Not even comparing it.
They just give an example of something presumably almost everyone agrees is immoral as a thought experiment.
There are two things that the blog author is conflating
1) The enslavement of ppl and the monopolisation of their intellectual work
2) The donation of the monopolised work to a project where all donations become available to everyone, under a Free license
Item 1 is bad. Item 2 is good (provided only that the work is good)
He's practising a rhetorical sleight of hand and his argument should be rejected as sophistry
Did you read what I quoted? It wasn't even from the earlier slavery part (that you're talking about).
It came much later ... but it nevertheless used the term "unconsented labor". When you use language like that ... after talking explicitly talking about slavery earlier ... it's really hard not to see it as the author contrasting piracy and slavery.
> but it nevertheless used the term "unconsented labor"
Which you quoted as "enforced labor". Now you could say that "unconsented labour" is not really a thing, but to be fair, "enforced labor" isn't either. And if we want to nitpick, "forced labour" (which implies coercion) is not "slavery" (which goes further and implies ownership).
Now back to genuinely trying to understand what the author meant, I would think that "unconsented labour" means that the result of the work was used without the consent of the person who owns it. Which is not forced labour and even less slavery.
Piracy is OK as it's purely consumption. This is not piracy. It's something much worse. It's producing derivative work. It's statistical plagiarism.
It's not stealing work that is done, it is stealing work that is to be done.
In some cases it’s not even probability, it just _is_ someone else’s work with the copyright removed.
It drives me crazy how people genuinely cannot understand what analogies are meant to do.
NO, they DID NOT equate piracy to slavery.
They made a thought experiment to convey a point. "Imagine that we could benefit from something that we all agree was produced by unethical means, would it be ethical to do it?".
When someone says that "DNS is a bit like the phonebook but for websites" and you say "this is UTTERLY wrong, DNS is NOT made of paper", then you are the one being completely ridiculous.
Author explicitly talks about slavery at the start of the article, then later (separately from the slavery stuff) uses language like "LLMs ... are built on unconsented labor" ...
... and you don't think he's drawing any parallels whatsoever between slaves and how LLMs are built (ie. from piracy)?
I don't think he is, I would not interpret "unconsented labor" as "forced labor." It's an awkward phrase but if he meant the latter he would have said it. I read it as meaning LLM training takes advantage of work that was done in the past and reuses it commercially with no attribution.
> Author explicitly talks about slavery at the start of the article
The word "slavery" appears 3 times in the article. Once in the quote (the analogy), and twice when clarifying that it was an analogy and nobody actually supports slavery (because, you know... people don't understand how analogies work).
> then later (separately from the slavery stuff) uses language like "LLMs ... are built on unconsented labor"
If the copyrighted work of an author serves as a building block for LLMs without the author agreeing to it, I think it's not completely weird to refer to that as "unconsented labour", is it?
> and you don't think he's drawing any parallels whatsoever
I don't.
Spoiler alert: when people talk about "piracy" in this context, there are no boats and no swords involved. Even if they happen to say "that ship has sailed" a few paragraphs after mentioning piracy.
Well they are asking us to extrapolate.
When we build something for our livelihood, they take it, tell people they made it, and that we only deserve the table scraps.
They say eventually we will thank them for their handouts and stop desiring freedom or respect or rights
> Freedom for users, not just programmers. Source code access means little if you can’t read it. Real software freedom has to include usability, accessibility, documentation, and support in the languages people actually speak.
Programming has gotten much more accessible with LLMs. This will eventually result in more user freedom in practice, once forking free software becomes the norm for end users. But custom software needs to be managed, so we will need hosting services and software installers that make finding, downloading, and customizing apps easy.
And running marathons has gotten much more accessible with automobiles.
I'm not really sure what your other point was. There's no universe in which a LLM can somehow correctly fork a project for a different purpose but somehow still needs help in hosting, publishing, building an installer or a ui to do so.
Sometimes the software already works well for you, but you want to make a small change, perhaps adding a feature.
Here is some software I use that supports this:
https://blog.exe.dev/customizing-shelley
It’s a coding agent, but the same thing could work for other software.
Relevant comment from Stefan Monnier (former emacs maintainer) on emacs-devel:
> The issue I have with this is that it moves the goal from "defend Free Software" to "defend against generative AI". The Free Software movement's goals are largely aligned with many other social movements, yet to avoid fragmenting our community, the FSF has been very careful to try and stick to software freedom (and let individual members combine this with any other issue they like). I'd rather leave the fight against slop to other groups, even if there might be a lot of overlap between that group and ours.
https://lists.gnu.org/archive/html/emacs-devel/2026-10/msg00...
I don't have a strong opinion, but I wanted to say two things:
1. If Free Software must be ethical and therefore refuse contributions by LLMs, should Free Software refuse contributions made on hardware that was itself made of components that were not ethically-sourced? Should contributors prove that their laptop is made of ethically-sourced materials? Good luck with that, I guess?
2. Now that LLMs are there and are undeniably multipliers, isn't that an opportunity for Free Software? Reverse engineering proprietary firmwares used to take years, now LLMs make that a lot faster. In a way, is it bad to "use the weapon of the enemy in the fight against themselves"? I would think that this is debatable?
When (Theravada) Buddhist monks go out each morning to accept food donations (Dānasantuṭṭhi) from the laity, they do not exclude donations from anyone
In the Pali Canon, the Buddha accepted alms from murderers, thieves and courtesans
Monks follow this example in practising Upekkha, impartiality
(Of course, if the food is bad or inedible they will not consume it but the source of the donation, and how the giver was able to obtain the food donated, is not considered)
Do you really want to die on that hill, or do you want to take a valid contribution that fits in with the existing project?
I've never understood why political beliefs should matter for code, unless it actually affects the code in a destructive way.
Unlike open source, Free Software (capitalized) was intended from its inception a political movement, so of course politics matters to it. Which way the politics should matter is of course a valid topic of discussion.
Yes, but there is wisdom in knowing which battles to fight, and which issues to let go. Ultimately, this up to the owner of the project, of course.
And this is, to my understanding, the point of the article.
> Free Software (capitalized) was intended from its inception a political movement
I completely disagree. Free Software is an ideological movement based on a deeply pragmatic viewpoint of software use (not using software that does not respect your freedom). It has nothing to do with politics and is open to everyone of any political persuasion to use.
You might be thinking of open source. From https://www.gnu.org/doc/Press-use/fsfs3.pdf (p. 99):
> Please avoid using the term “open” or “open source” as a substitute for “free software.” Those terms refer to a different position23 based on different values. Free software is a political movement; open source is a development model.
No, I very much mean Free Software. I am a donor to the FSF and well read on its motivations.
Free Software is apolitical
No, we take the hill and live on it while ANT will crash tomorrow.
If you read French, or can use a translator, I wrote about this paradoxical approach of the free software movement a few years back: https://p4bl0.net/post/2022/11/Paradoxe-libriste-et-logiciel...
I think you are discussing a related but quite different topic here (worth discussing, for sure, but I'm taking issue with your phrasing "I wrote about this paradoxical approach of the free software movement").
The post discusses about how the code gets written collectively, yours discusses what the code can be used for. Yours discusses potentially desirable limitations on freedom 0 (and indeed, contradicts the accepted definition of free software - that's not a judgement, it's a fact). The other post discusses being more demanding on how free software is made in a way that doesn't contradict the free software definition, and (that's where both are indeed related I think) whether we as maintainers should have our shoulders to the software freedom wheel a bit less and consider more the bigger human freedom picture (and I guess my phrasing gives away what I think about this).
Restricting freedom 0 is tempting but perilous and can have consequences you have not anticipated and might not like. One can totally agree with your goal (avoiding an uber-like situation for workers) but not with how you do it. It's not the same ethical / philosophical stance as one who considers other ethical matters than software freedom out of scope (and both stances are actually not incompatible).
The author is clearly against AI models that are trained on data produced by people without their consent. Yet the whole free software movement is built on the notion that proprietary software is unethical and any attempts to protect it (and other content like movies and music) using DRM or other mechanisms are wrong.
Does anyone else see this as a contradiction?
> [...] trained on data produced by people without their consent. Yet the whole free software movement is built on [...]
consent. The whole free software movement is built on consent. For the most part, that consent is conditional - like requiring that you pay it forward (GPL), or at least mention who the author is (MIT).
I think I see what you're saying, but no, I don't think it's a contradiction.
I think the worry from the "free software movement" side is that someone consentually shares some code, then a company takes it, profits from it, but "locks it down" so their additions are not likewise shared. So basically the company is freeloading on what others produced, taking but not giving back — that's what people frown upon.
Likewise, training on data without people's consent is also a type of freeloading, and is frowned upon in a similar fashion. Note that this happens with humans, too: if an artist takes too-direct "inspiration" from another without getting credit, they are called out.
I suppose there are some more absolutist stances, like "abolish copyright" (rather than the milder "abolish freeloading" described above) which would argue against DRM, etc. of any kind, full stop. And, if such a world ever came to pass, and those people felt that their data (artwork, code, etc) should not be available for AI training, that would perhaps be a contradiction, yes.
Well, it's not necessarily a contradiction, as the copyright protection and licensing aspects are separate. They don't support what they see as copyright infringement, but they also don't think that copy-protection schemes are good for the user, since they also prevent legitimate uses under the law. Also note that the position of the FSF et al. is that there's no reason that art (I mean this in the traditional sense of art: books, music, movies, etc.) can't be proprietary, but that programs are different.
Not necessarily, becase the FSM is not about "freeing" software. Its about "free" software. They're not cracking DRM. They're not trying to pirate other software.
Rather they're arguing that software should be free, they work on free software, and then promote their free software over non-free software.
Basically, everyone involved in the FSM is consenting.
With AI IP scraping, that's not so much necessarily the case.
> Their argument was that in the context of Emacs development, the only thing that matters is the quality of the code being contributed
This is something I've never agreed with. One part of open-source is to have some principle that is rooted in ethics, but how can you be true to that when you ignore other ethical dimension that can't be separated from the traditional open-source one?
I don't think the only think that matters is the quality of the code. It's a myopic viewpoint. The reasoning behind having open code in the first place matters as well, which was that one person was trying to take advantage of another in a way that seemed intuitively unfair. LLMs do the same thing. If you fight for only one ideal but reject another parallel one for convenience, then I believe the first stands for nothing at all.
Over 20,000 people have contributed to the Linux kernel and there are around 2,000 contributors per release. How can it be viable to judge whether each contributor is overall a good person?
One way to make moderation practical is by limiting the moderators’ jurisdiction to the things they have direct evidence for. For example, presumably the moderators of Hacker News limit their scope to what people do in Hacker News, rarely going beyond that to looking at behavior in other forums or what people do in their personal lives.
Similarly, moderators of an open source project could mostly limit their scope to what people do when interacting with the project’s systems and community.
You can't judge whether everyone is a good person. But you can put in basic rules that obviously go against immoral behaviour. Not everyone obeys the "no shirt in the shopping mall" rule but it's there and it's a good one.
Many open source projects have a Code of Conduct, for a reason. And people who break it are called out. Many open source projects are anti-AI as well like YT-DLP and Asahi Linux. It can be done and easily without much fuss.
Contributors who use AI for some other unrelated project wouldn't be called out, though? The code of conduct only applies to what people do as contributors.
So what is the "no shirt" rule for OSS development? It seems like it's basically Google's "do no evil", right?
Now, Google itself stopped following that years ago, which offers a hint as to the problem. We can say "OSS contributors can't be evil" ... but we can't agree on what "evil" should be defined as.
> "no shirt in the shopping mall" rule but it's there and it's a good one.
Hey man you put it off the beach and I just want an Orange Julius what’s the problem man?
Hahaha. Good one.
Some interesting overlap here with the "What Color Are Your Bits"[1] article. Check it out if you haven't already (:
It describes some of the communication mismatches when people talk about IP issues. "Lawyers see color" but computer scientists don't.
[1] https://ansuz.sooke.bc.ca/entry/23
AI(local) is the feature of OS, when it's powerful enough to create your Windows and Word your Photoshop and Chrome then software will be truly free.
Local AI can already do all of those things. I build operating systems and compilers using only consumer GPUs in my garage.
Which models do you recommend running? What amount of RAM do your GPUs have?
I am a software, systems, and security engineer with two decades of experience so I could already build just about anything by hand, so all I require from an LLM at this point is speed of execution on my specific instructions.
I could run larger models with fewer sessions, but with my cluster of 1x strix halo and 8x r9700s I get 13 sessions of qwen 3.8 flash next 135b at > 100tps each and that is the sweet spot for me to solve any problem in front of me from kernel and compiler bugs to writing hardware drivers to building specialized operating systems. I cannot think of any systems engineering problems I have seen in my career I cannot readily solve at home now on my own.
With radiance and/or strata a single strix halo or r9700 is enough for one session though for someone that wants to start on a smaller budget with less concurrency.
Stallman's reply to this all is really good, especially as this who discussion appears to be a contest of Who Is The Most Principled And Consistent and Stallman is the OG there and still very hard to beat. Some excerpts:
"The example that you're responding to is about where to draw the line -- it was designed to be a borderline case."
"We don't reject contributions from companies such as Google and Amazon which are intentionally subjugating users. These practices are not as evil as outright enslavement, but they are plenty nasty nonetheless."
"Rejecting their useful contributions would not help us fight to stop their unjust practices. We may as well accept the former, as long as we don't mute our condemnation of the latter."
https://lists.gnu.org/archive/html/emacs-devel/2026-11/msg00...
This does not make sense to me. Does the author use money? Is the author remotely aware of the way money is generated and multiplied? Yet once in circulation it can be used for totally honest transactions. Same with code, with the important distinction that code is not inherently a scam. So code should be cut more slack than money. Code is not evil, its use may be. If bad person curiously donates correct code, I keep it.
After the code of conduct , do we want a moral certification for code contributors? You do know where this mindset is leading, don't you?
Should GNU not accept contributions from Mac users, since Macs have been built from child slavery? https://www.freedomunited.org/news/apple-child-labor-lawsuit...
Well, it's highly unusual for me to defend the mac, but I guess we'd be all cooked really, because i think horrible working conditions in mines concerns any computer hardware, and that's true for many other things as well. It's not specifically Apple. IIUC, your link seems to be about lawsuits against false claims from Apple (which is extra bad, of course, but others are not saints).
So, it's everyone's devices really.
Here's how I would resolve it: it is a real concern. We should fight against it, and maybe also work on reducing our systemic dependency on computers, use hardware as long as possible, limit the amount of hardware, and encourage reuse (and recycling when everything else fails).
But for now computers are more omnipresent than ever anyway so users having control over their computing and computers is also more important than ever. It's still a fight worth fighting despite software basically depending on horrible labor conditions. (And btw, software already depending on horrible labor conditions is not an excuse to get even worse)
Honestly? It would be funny for a popular project to do this. Code quality would likely improve as well.
Wouldn't most, if not all, those issues go away if AI generated code is always AGPL?
They also go away if capitalism is abolished, or if gravity ceases to function. What's your point?
Do you want to be right or do you want to be effective? As much as the article tries to claim it's not setting up a "moral purity test", that's what these sorts of hypothetical conundrums do. This is the exact type of privileged navel gazing that cause Democrats to lose the plot on what working class people in America had actually been experiencing, opening the door to much more blatantly obvious corruption and grift from Trump.
It's important for social and political movements to scope their ambitions to not alienate potential allies with overcooked positions. Build coalitions as well, but do so opportunistically where you have a chance to actually win something and move the needle. Don't craft your charter to already be so exclusive out of the gate that you only appeal to the like-minded fringe. You want to do something for software freedom? Talk about John Deere, not imaginary enslaved open source coders.
I honestly think the author of this article doesn't really consider the value of ideological interoperability. In real life, sometimes maintaining a narrow mandate is itself an ethical decision ESPECIALLY when that mandate allows people with incompatible worldviews to cooperate.
>Software has effects beyond the machine. Systems that decide who gets a loan, a job, or bail are freedom issues, whether or not the license is permissive.
Hang on to that a second, I'll come around to it.
Consider the single responsibility principle... OSS projects are organized around software freedoms as a well defined set of commitments, allowing contributors to disagree about almost everything else while still cooperating toward those SPECIFIC founding commitments. In daily life, when collaboratively making decisions other people have to live with, I think the narrower the shared commitment, the broader the range of people who can cooperate under it.
The moment you decide to step outside your founding ideology and take on more ethical commitments, you're now attempting to make broader decisions about which other ideologies are morally|ethically correct, creating new reasons for disagreement and potentially new reasons to exclude contributors. Today it's AI-assisted code, tomorrow it's defense employees and then maybe whether corporate sponsorship matches the project's values.
Really, it's easy to see how this causes fragmentation because there isn't a principled stopping point once the project's mandate becomes doing what is ethically right rather than protecting a particular set of freedoms.
I posit that it's COMPLETELY okay to refuse to take a stance on subjects that seem to sit outside the ethics of a project's purpose. There should be a clear delineation between an individual's ethical convictions and making those convictions binding on the group. It's a really blurry line especially since additional ethical positions can seem perfectly reasonable in isolation but collectively just add more friction to cooperation.
It's not the responsibility of the group to take a stance on every moral dilemma separate from its founding ethical stance unless it's existential to that ideology or necessary to fulfill, protect, or practically sustain the original mandate. In this case, I don't think it is.
And tbh, I'd expect leaders of OSS projects to respect the moral agency of contributors by declining to exercise moral authority over questions the group never collectively agreed to delegate to them.
Following that, I sincerely believe that code is absolutely all that matters but this is real life and abstractions can absolutely be leaky. Software really does have an effect beyond the machine. That might seem like an "aha, you see?" scenario but not really...
All that matters is code but code is also, at its most fundamental level, intention. The decision of what to build is actually the most tangible expression of a project's ethical purpose, even more so than whatever its founding principles say. The purpose of a system, after all, is what it does, not what it claims to do.
The code is the vehicle through which a project's particular vision of how the world ought to be is brought into existence. The software's effects beyond the machine are ergo not some separate ethical dimension the project has failed to consider, they're precisely why the project exists in the first place. The movement has already chosen the change it wants to effect in the world, and the software it produces is how it pursues that change.
So the project's ethical commitments are already embedded in the work itself, in what it chooses to build, what it chooses not to build, and the freedoms its software is designed to protect. Its moral direction is expressed through action.
I sincerely don't think that every moral issue connected to software must become part of the project's mandate. The real question is whether that issue materially bears on the particular change the project exists to bring about.
Even for individuals, it's never a good idea to attempt to have an opinion on everything. Sometimes, the most meaningful expression of value is to remain relentlessly focused on the one thing you're trying to accomplish.
I agree with the premise that the crux of the issue is consent, the movement is political, and the concerns largely of moral and political philosophy.
From that frame and on its own terms, there is a specific mechanism to context the conclusions and their moral authority: the dimension of where individual liberty (and hence consent) collides with the interests of the body politic and the society at large.
We can imagine the arguments that follow without needing to advocate for them: if you are a good faith accelerationist, the problem with this position is not that it is wrong on its own terms and given its assumed priors,
it's one of the moral foundation. The best interests of society may reasonably be argued to supersede the interests of the individual.
Whether one agrees with that premise generally, or with respect to AI accelerationism, or the ethics of building and using LLMs as we know them, or not, is a different question than the author addresses.
IMO the author's concern is premature. Or, alternately, assumes that this foundational question has an obvious or settled answer.
It does not.
Like it, or not, modern societies do in every dimension—legal, political, ethical—presume that the interests of the commons, of society, of the body politic itself, of the state, do in many places supersede those of individuals.
Libertarians disagree of course: and when they formulate and inhabit their own state, well, they shall be free to operate it on their own terms.
Until then and for the rest of us: the prior art here is against this position, at least the degree you believe one takes seriously that, variously, we are headed into a singularity from which various relatively utopian or post-scarcity outcomes may emerge; or, alternately, that the dystopian reality of a not just dead but actively hostile internet requires one deploy whatever tools are available and capable to survive it. (And the only tool we have to understand and defend against bad AI is good AI.)
The number of business shills in the comments here herding their open source slaves back into the enclosure is funny.
FLOSS is about human rights pertaining to software, and not some mythical 'software rights'.
The fact the FSF does not understand this (or chooses to be oblivious) is why I quit donations to them.
Edit: The people who downmodded this are part of the problem. FLOSS is all about human rights, as pertaining to software.