sex 15 mai 2026 às 04:28:18 (1778830098), [email protected] enviou: > Hugo Buddelmeijer via "Development of GNU Guix and the GNU System > distribution." <[email protected]> writes: > > > Every example of LLM's regurgitating copyrighted text starts with a > > prompt that is derived from that copyrighted text. > > > > Hugo > > This claim may be true at first sight, but copyright is about stealing > creativity and not just stealing code verbatim. I continue to believe > that we cannot ensure there will be no legal payment obligation to > authors guilds at odds with free software for commercial use. >
I keep seeing people using the adversary's language to talk about copyright. >From where I come from, stealing is a word related to taking something away from someone and GNU included it among the words to avoid when talking about copyright. When someone copies copyrighted works they are not taking anything away from copyright holders with or without license compliance. Long ago, I wrote an essay to FSF addressing Copilot, which unfortunately was not selected for publication. I'm attaching it here, as I still think it's applicable to the LLM's vs copyright debacle. Copyright was historicaly a means to control the freedom of press, of access to knowledge, not a means to enpower authors, creators nor of being an incentive to creativity. Loosing that perspective leads us to defend stances which are not conductive to a free software world. The goal of free software is to have a commons and user liberty by default, not to defend copyright holders' interests. Cheers ----
"Copy a lot!", says Microsoft When Microsoft acquired Github there was an outcry from many computing freedom activists who claimed this was but the first step of their traditional tactics of undermining free software, namely the well-known "embrace, extend and extinguish" routine. Therefore, free software concerned individuals should stear clear from Github reliance and embrace git's original distributed potential to avoid further harm to free software. Notwithstanding the warnings, time has passed and Github kept its status as a central place for coordinated software develpment and code hosting. So now again, when Microsoft announces "Copilot", a SaaSS designed to help coders in their programming efforts by means of suggesting common or good code solutions for common programming problems, there is a similar outcry. This time, however, computing freedom activists have been put on an unusual position: defending copyright and claiming that the corporation is making unlawfull usage of fair use exceptions to it. Both the positions above may be regarded by those who stand by them as clairvoyant, strategical or principled. On the following paragraphs, I'll argue that, to the contrary, they are actualy shortsighted and misguided criticism and come from an inferior, powerless regard of the freedom ethics we stand together to further, as if our ethics were at the mercy of Big Tech's efforts to undermine it. To sum up the arguments on the following paragraphs, I would say that, on both occasions, the headlines I wish we could have seen were on the lines of: "Microft pays $7.5 billion dollars for the privilege of hosting free software", "Microsoft acknowledges sharing code makes programmers better", "Microsoft develops AI to research and enhance free software"; "Microsoft acknowledges programmers should not have to rewrite code for problems which have already been solved"; "Microsoft takes the freedoms to study and modify source code to a new high"; or "Microsoft has been diagnosed with Linux-cancer, Doctors say there is no treatment to it". One could say that any or all of those are somewhat misleading, but it is also clear that all of them shed light on an important issue that got little to no attention at the time. When RMS wrote the GNU Manifesto, refused to break solidarity with other users and kick-started the free software movement, there was no publicly available free/libre Operating System and hardly any free software given that copyright had become the default and software was still very much of niche interest: the domain of governments, military, corporations and universities. Personal computing was on its dawn and not many believed that one day computers would be on the pockets of almost every human being on earth. Computer users then were literate and knowledgeable on programming. The default computer user interface was the console. We all know this by heart, but maybe we do not fully appreciate the consequences of a changing environment beyond the feeling of nostalgia for that long lost golden age that Steven Levy talks about and of which not many of us nowadays have had any first-hand experience. On those times, people who stood for computing freedom saw copyright as a threat to user's freedoms and the playground of corporations to divide and conquer users. In order to avoid that trap, those concerned made up various free and public licenses as a means of nullifying the restrictions as well as opposing the egotistic culture that copyright conveyed. Computing, they all knew, was a collective endeavour where each and every user can benefit tremendously when sharing is widespread. Thus we would all be better off if we shared our solutions to common problems and/or implementations of new ideas. Even more important for the scope of this article is to remember that the "share-alike clauses" of GNU/GPL and copyleft licenses were just a means of protecting those early freedom activists from unjust competition against unethical adversaries who would otherwise be able to claim their work for the purpose of advancing proprietary software while the reverse would not be true under the copyright restrictions applied to their software. Therefore, those licensing restrictions were a means to the end of advocating for users' freedoms, not the end goal to which software freedom was but a pretext. In that sense, now that Microsoft made available to the public a software which, in order to be legal, requires that fair use exceptions to copyright be in some cases broader than what is currently undertood to be, they are effectively changing the socio-cultural background in which copyright restrictions stand. They are implying that copyright does not warrant its holders all the powers to restrict everyone else from using their works to the extent they might have previously thought to be the case. In other words, they are implicitly claiming that the default public freedoms are much broader in the face of copyright regardless of any licensing. That being the case, we can safely say that they are creating a good social precedent for user freedom even if, as a side effect, they might be hitting a blow on those GNU/GPL's powers which are grounded on copyright. However, the GPL itself was a clever hack conceived as a means to bypass copyright's default restrictions on the general public. If those restrictions were to start to crumble would we work towards strenghtening them in order to keep the usefullness of the hack or would we hit them even harder and be hopeful once again that the hack may one day not be needed anymore? Do we still hope to see the reversal of current culture in such a way that freedom becomes the default and proprietary is seen as oppressive and unjust? As the con it has always been? To tackle these questions, I propose that we take a serious look at where we stand right now, keeping in mind the path that led us here. From those early times when GNU was but a dream in the minds RMS and a handful of hackers to the present times, we have come a long way to a world where, whilst proprietary software is still widespread and a threat to user freedom, there is a huge, almost unbeliavable trove of free software scattered all over the place, countless GNU based distros relying heavily on free software, even if not many with strict adherence to FSDG, and there is free software being used on billions of computing devices. Yes, we cannot forget that not many of those rely solely on free software. Yes, we should not overlook that the overwhelming majority of computer users on the globe have never even heard of software freedom, even when many unknowingly use or interact with some free software component on a daily basis. However, instead of focusing on these saddening facts, I would like to highlight the obvious, truly amazing fact that we are now able to run free software for virtualy any tasks one might want, need or imagine. We have more free software readily available to us than we probably could even be interested in on an individual level. How many of those dozens of thousands of free software packages do you run on a daily, weekly or monthly basis? How many have you run ever? How many of those publicly hosted free software have never been packaged to any distros? Let that sink for a moment. It is way too easy to get lost on our pursuit to the complete, undeniable, irreversible and universal acclaim of our ethics and just forget to take a time to contemplate and celebrate our many smaller feats, our organical growth along the years and rest assured that we have been excedingly good following the path. Anxiety may be one of the symptoms of our pressing times. So let us remember once again that we are not on the same weak position that RMS and others were when they had to literaly build things from scratch, ask for permission from other people to use their code, be socially well positioned to even have access to a computer. We are not facing the exact same issues they had ahead of them. Thanks for them and many others, we have free software available to us for almost any task for quite some years now. One may have never ever even seen source code in their lifetime and still be able to install some free software distro with just a few keystrokes and a negligible amount of reading or asking around. So, with our minds in that state of ease and awe, let us get back to the challenges we have been facing, which by all means are not the same as 40 years ago. The javascript trap, nonfree firmware, nonfree hardware designs, SaaSS, locked up app stores, peer presure to remain using nonfree software and centralized services for communication on the grounds of "everybody uses that already". All of these issues share one common pattern: none of them have a clear programmable solution. It would be unfeasible for us to replace alone every nonfree javascript routine on every single website. It has been impossible to reverse engineer every single firmware blob for every single device available on the market. Free hardware designs are almost inexistent. We cannot code people out of SaaSS and app stores. Even though there are good free software replacements for communication, many people are not even willing to try them, since their family, friends and coworkers do not use them. All of them have a clear, however cumbersome social solution: if users were to strongly value their freedoms - think a society of people half as stubborn as Stallman on their demand for freedom, maybe even one tenth as stubborn -, corporations would have no chance when trying to push users to accept their swindles. Alas, this is our end goal as an ethico-political movement: to touch people's minds and beliefs in such a way that our society, culture and legislation may change accordingly. More than ever, we now need to raise awareness on the importance of freedom, not only for the programming enthusiast - most of them are bound to hear something about GNU and free software -, but for the lay person. We can only hope to effect the change we envision once we get a critical mass of freedom concerned users who would rather refuse to being subjugated to third party's interests in the name of a little gain in convenience, old time inertia, processing speed or what have you. Before that time, we will be facing enourmous hardships overcomming the chains and barriers that corporations silently impose on every user, unknowingly to most. That being clear to us, we may now face the issues and possible paths that Copilot brought about. By developing it and making it available to the public the once self-declared adversary of free software and champion of copyrights, economic incentives, and scaremonger of hobbyists is signaling very clearly that it sees imense value on public hosted source code and it is also encouraging people to shamelessly make use, copy, modify and share source code. More than that, it is giving way to socio-cultural background for a new take on fair use exceptions to copyright restrictions. That stands in high contrast with the low value that most of their useds assume that free software has. The preconceptions on the lines of "if it is given for free, then it is not worth much" permeates the imaginary of many. Some call this "ideology", but I would like to avoid that word. However appropriate to describe the situation at hand, this word has lost an important aspect of its meaning and nowadays people use it even to describe ethics and doctrines they adhere to and support. Nonetheless, we need to have it clear that the majority of people we need to reach hold as undoubtly true some set of beliefs that are not grounded on hards facts or personal experience on the subject matter, they were taught to believe them as facts. No amount of "free software is not free beer" will address this since free software is often available gratis. Other ingrained beliefs that comes to mind are the strongly held assumptions that economic success is highly correlated with and indicative of technical superiority; that only by restricting user freedom it is possible to develop software; that monetary gain is the only motivation which can bring about innovation and development; that user freedom would make developers/authors miserable; that tech companies only rely on proprietary code; that there is no other way of funding the development of software. These discoursive battles need to be in the forefront of our preocupations. If we are to convience large swaths of not technically inclined people that freedom is at the same time ethicaly superior, technically and economically feasible and desirable, we need to be facing these beliefs head on. One way of doing it is by engaging with things like Copilot not with the intent of protecting our current ground but with the intent of conquering newer, larger ones. So instead of being afraid that our current free software commons might be used here or there without proper compliance to our copyleft licenses, we may be better off focusing on the venues it opens up to us to once again talk about user freedom and attack the prejudices that might hinder its reception by a broader public. Instead of trying to condemn Copilot and its useds, we may use them as an example of things that only became possible because there was a free and open source software environment. We may use them to educate the broader public on the collective capabilities that only exist when people share their works instead of employing techniques to hide them. We may teach them that 'git' itself is a free software used by a whole lot of companies and that there is no proprietary solution that trumps it. Notice then that these are not mutually exclusive paths. Either way we can make it a victory to computing freedom: if their AI usage of GPL code does not fall into a fair use exception, their customers found to be copying GPL source code will thereby be liable for licensing all of their code that relies on it by the GPL's terms and conditions and thus FSF or the copyright holders may approach them in order to seek the desired compliance, to educate them on the principles we stand by and hopefully increase the amount of free software available to us and of free software enthusiasts; on the other hand, if it is indeed deemed to be a fair use of a copyright protected work, then the same reach of fair use should be applicable to any free software which came to rely on partial copies of otherwise nonfree code. In this later case, it is worth remembering two common happenstances. The first is proprietary source code which is released to the public without license - a situation which is very common even on Github itself - or proprietary source code which was leaked to the public. The other one is the employees who are coerced to assign all their copyrights to their employers. With a broader fair use exception, they may start to use some functions, routines or maybe even small libraries to help the development of software elsewhere without infringing the signed agreements. To conclude, I would like to remind you that making something unlawful does not prevent it from happening. Even before Copilot came about, programmers were able to use copyleft source code without acknowledging it or being compliant with their licenses. Copilot may make it even more common and may induce people on doing it unknowningly. However, if instead of using this to the advantage of furthering our ethics we were to take the path of trying to prevent what has already happened from happening, we will be directed towards a path very much at odds with our ethics: trying to argue for stronger copyright protections, for fewer or narrower exceptions to it, developing techniques which try to obfuscate code meaning in order fool machine learning or to make it less general or less useful in other contexts, at great cost to code reusability among free software projects and creating an artificial barrier to newcommers to free software projects. Moreover, we would probably be misidentifying the difficulties we face today. Our greatest obstacle right now lies in making our ethics known to broder audience not in the threat that some copyleft code might be used without compliance. For this we already have the means to defend ourselves. For that we are in need of taking any and every opportunity to make a case for user freedom. Omnia sunt communia
