An image with Rousseau in a laptop which has a sticker with the logo of the open source software foundation.

Abstract

The ethical and legal questions around the use of someone’s code by others led to a revolution in the 1980s around the laws of intellectual property, and the freedom of software. In this article I briefly retrace that history, and then I analyze the related problems from the perspective of Rousseau, his views on property, and the politics of societies.


Background

Software Licenses

Software that is publicly available is generally not released with a “blank check”; we cannot do whatever we want with it.1 There are specific rules which are covered by software licenses. A software license is a legal contract between the creator and the user of a software which grants permissions (use, modify, distribute) or imposes conditions (e.g., can only be used with attribution). Software licenses are the primary legal tool for protecting any intellectual property related to software. They brought a revolution in the legal circles because they introduced a model of protecting intellectual property that moved away from copyright (Moglen, 1999).

Software licenses constitute serious technocratic business reserved for lawyers. Software developers cannot and do not develop the licenses of their software. In fact, creating a license is in general such a tricky endeavor that there are basically only a handful of licenses which get reused.2 Rarely does a new piece of software ship with its own custom license.

Finally, licenses are not “irrelevant bureaucracy” for a software project; they can determine a good chunk of its fate. For example, LLVM (Lattner and Adve, 2004)—a project created at UIUC and literally one of the most important and used pieces of software ever—owes much of its success (according to its creators) to its permissive license, which made it easier—compared to the then most popular alternative, GCC (GCC, 2026)—for companies to use all or parts of it (we will return to this point later).


Free Software and Open-Source Development

In the software world there are several characterizations of software with to what one can do with the source code.3 At minimum, we need a term that denotes software whose source code is publicly available. Perhaps surprisingly, the software community has not reached consensus for such a term. The one that the U.S. Department of Defense uses, and which Wikipedia copied, is “source-available.”4

The term “open-source” is closely related—in fact, it appears to be a synonym and that is what many software developers think. However, the de facto definition of “open-source” (accepted by several international governments (Open Source Initiative, 2015)) is the one by the Open Source initiative (Open Source Initiative, 2007). According to that definition, open-source software is almost identical to free software (which is why the two are usually composed to form Free and Open-Source Software a.k.a. FOSS).

In “free software” the term “free” should not be understood as related to price—not “free of charge”—but related to freedom. Because of the polysemy of the term “free” in English, sometimes people refer to such software by borrowing terms from other languages. For example, in French and Spanish “free” as in “a free state” is “libre” (whereas “free of charge” is “gratuit” in French and “gratis” in Spanish). So, free software is also referred to as “libre software,” “libreware,” etc.

A simplistic definition of free software is that the fewer restrictions a piece of software imposes (which is regulated by the software license), the freer it is. Until recently, the focal point has almost always been what people are allowed to do with the source code (the concerns extended to data in the age of AI, something outside the scope of this article), which is why free software is closely related to open software: if the source code is not available, then no one can do anything with it anyway. So, usually free software is source-available software shipped with a license which allows users to study, change, modify, and (re)distribute it.

An important implication of that is that multiple people (from different affiliations, countries, and backgrounds) can co-develop the software. Thus, by extension FOSS has come to also indicate a development process—a distributed process where everyone can contribute from anywhere (affiliation, country, etc.). It is this model that has allowed the unimaginable size, complexity, but also success of projects such as LLVM and Linux (one of the most popular operating systems and clearly the most successful FOSS ever).


Rousseau and Free Software

In this section I discuss free software using the Rousseauian framework. To the best of my knowledge, this is a novel analysis whose central thread starts from Rousseau’s “Real Property” to end up in unpropertied societies. The only prior work that explicitly connects Rousseau and free software glosses over that connection (Deibel, 2022, p. 74).

The most obvious path to discuss software rights is to view software as a form of property. To make a direct connection to Rousseau’s work, we can compare software with land, a form of property Rousseau explicitly talks about. Apart from the Discourse on Inequality, we may also consult The Social Contract, Chapter 9, Du Domaine Réel5 where Rousseau talks about property in general, but more specifically about who gets to have a legal right on a piece of land (Rousseau, 1964, p. 365, John T. Scott’s translation):

Each member of the community gives himself to it at the moment it is formed, such as he then is—himself and all his force, of which the goods he possesses make up a part. It is not the case that by this act possession, by changing hands, changes nature and becomes property in the sovereign’s hands.

Rousseau tells us that when a civil society is formed, every member of the community gives up all her power to the society. This does not mean, however, that she gives up her property too (her property does not change hands to now become the sovereign’s property). Rather, as Rousseau explains (somewhat confusingly) in the rest of the chapter, it is still her property, but now it is legally recognized by the community: this is the establishment of the property, a legal right. What is the basis for this legal right? For Rousseau it is the right of the first occupant, and Rousseau lists the necessary conditions for this right to be recognized:

In general, the following conditions are necessary to authorize the right of the first occupant to any piece of land whatsoever. First, that this land not yet be inhabited by anyone. Second, that one occupy only as much of it as one needs to subsist. In the third place, that one take possession of it not by an empty ceremony, but by labor and cultivation—the sole sign of property that should be respected by others in the absence of legal titles.

The first condition essentially means you do not steal. The second and third conditions are deeper than they seem. The second condition is truly only the barebones of an argument that we can piece together across all of Rousseau’s writings, which is that property should not become large enough to create dependence (whose ultimate form is of course slavery; see also Book 2, Chapter 11). Dependence is a major vice for Rousseau (Rousseau, 1969, p. 89, my translation):

There are two kinds of dependence: [i] on things, which is of nature, and [ii] on men, which is of the society. The dependence on things, not having any morality, does not harm freedom and engenders no vices; the dependence on men, being disordered, engenders all of them, and it is by it which the master and the slave are mutually corrupted. If there is any mean to remedy this ill in society, it is to substitute the law for the man, and to arm the general wills [volontés générales] with a real force superior to the action of every particular will.

It is here where we discern the limits of property for Rousseau. At the end of the day, any good whatsoever is a social good, and it is under the force of the general will, which is armed with “a real force superior to the action of every particular will.”

The third condition for the right of the first occupant is that this right is only truly acknowledged if the proprietor exerts his labor to cultivate the land (which of course is very similar to Locke’s labor theory of value). Thus, for Rousseau, the fact that one comes first—that fact alone—is not enough; it is through labor that the right is recognized (and supposedly maintained).

Let us now see if we can extrapolate Rousseau’s theory to software. It seems that the clearest argument we can apply is that of dependence. Today Big Tech creates dependence on multiple levels through software. First, social media is now a public sphere, using Habermas’ description. He writes that the public sphere is a space “made up of private people gathered together as a public and articulating the needs of society with the state” (Habermas, 1991, p. 176). But it is quite different from previous public spheres. Contrary to a park or the Ancient Agora, social media are privately owned. That alone is not that big a problem. After all, my parents’ and grandparents’ generations used the café as the physical space in which a public sphere was brought into life, and cafés are privately owned. However, one could switch relatively easily. In social media this is true only in theory, as in practice users feel they have no other choice (Dembrow, 2022). This was evident with the rise of Mastodon which had the ambition of replacing Twitter. But when all your friends are in a single unimaginably huge space—the whole world is in that one café—it proves quite hard to switch.

In that sense, it definitely creates unprecedented dependence. However, that is not the only level of dependence. One level deeper we find that social media design their platforms so as to cause addiction,6 as the landmark K.G.M. v. Meta et al. case (The Pisanchyn Law Firm, 2026) recognized:

[P]latforms’ design features like infinite scroll, algorithmic recommendations, and autoplay videos, fostered compulsive use and contributed to mental health struggles including depression, anxiety, and body image issues.

The lawsuit claimed that Meta and Google knowingly engineered features that hooked users, especially minors, and failed to warn about the psychological risks even though internal research showed addictive effects.

Moving away from dependence, can we analyze software from the perspective of Rousseau’s third condition? Any attempt should recognize one of the main differences between land and software: land precedes the human—it already exists, and at some point the human (the first occupant) cultivates it. Software, on the other hand, is created from scratch.7 But does that really change any of Rousseau’s claims? I do not think so. The “point 0” that matters is when cultivation starts, and for software that starts from its very creation.

But now we shall understand a crucial difference between the software we have discussed up to now and open-source software. The former is owned by someone, and because of that it can be viewed as private property. Rousseau’s analysis on property above really refers to private property; property that is owned by someone, the first occupant, who also cultivates the property. But in open-source software there is a first occupant but does not remain the only occupant and cultivator. Open-source projects (like Linux and LLVM) are not developed by any single person, or even any single legal person (e.g., a single company). It is a collaborative and distributed effort which in practice involves multiple companies but also many independent contributors.

Under a Rousseauian light, then, I believe that the only reasonable interpretation is to consider an open-source project as a social good, which is not only a good of the society at large, but it also creates its own mini-society. So, who does it belong to? To no one, just like for Rousseau a society is not owned by someone. And who has a right to the project? The society at large, but also its cultivators, which in this case are the contributors. Naturally, then, the future of a project should not be determined by any particular will, but it is not the sum of individual wills either. Rather, as we know, decisions should be made based on the volonté générale.

Now that we have established a Rousseauian viewpoint of software, let us apply it to evaluate an implementation of free software—that is its claim—the GNU Public License. GNU8 is a collection of free software whose legal establishment rests on the Free Software Foundation (FSF), a 501(c)(3) non-profit organization. The GNU actors played the most important role in the history of the term “free software,” forming a revolutionary and opinionated conception. This was then concretized by major advancements in Software Law through a software license they came up with: the GNU Public License (GPL).9

The most crucial clauses of the GPL are the following two conditions: any software licensed under GPL must disclose its source, and any software that uses source code from a GPL software must have the same license. Nota bene that GPL still allows software to be used commercially (i.e., sold).10 But for example if I work at Meta and I want to copy some code from a piece of GPL-licensed software, then my software must be licensed under GPL too. This, in turn, means it has to be open-source. As one can expect, this is not something private companies love at all.11 Now we can get further insights into what I mentioned in the Background about software licenses. There, I mentioned that the license of LLVM was instrumental to its success. The main open-source alternative at the time was GCC. But GCC stands for GNU C Compiler, which means it has a GPL license; a detail rather unattractive for companies.

Despite the dislike of the commercial world for GPL, these conditions impose not only that a piece of software is free, but also that any derivative software is free too. When I introduced free software in Background, I defined it as “software that imposes as few restrictions as possible on how users use it.” Some claim, then, that GNU software is not truly free because of these conditions. But for GNU people, and probably for Rousseau too, these conditions are special in that—to use Rousseau’s language—they force us to be free. They provide the general will with the legal force which is required to overcome any particular will that may go against it.

Whatever we have mentioned so far is also relevant to Artificial Intelligence (AI), but in AI there is one important caveat. Remember that licenses capture modes of use such as commercial use, redistribution, modification, and whether you should attribute the original creator. But what about learning? Imagine a restrictive license which does not allow the user to redistribute or modify or use the code in their projects either commercially or privately. Yet, the code is publicly available, with a license that allows one to read it.12

Until pretty recently, no one would even think that there is anything wrong with looking at the code, reading it, and even learning from it (obviously as long as it's true learning—whatever that means—and one does not e.g., simply copy and paste it verbatim). That was mainly, albeit unconsciously to most, because it was a human that used to do all these things, and the only one who could do them. Perhaps equally important is that a human, contrary to an AI model, is not a product.13 As a result, AI introduced a general sentiment of uneasiness into the world as now machines could do all that.14


Don't want to miss any updates? You can follow this RSS feed.

Bibliography



Footnotes

  1. Unless it is explicitly released in the “public domain.”↩
  2. On the webpage choosealicense.com you can take a quick look at some of the most popular licenses and their terms.↩
  3. What one can do with the executable code is also relevant in real-life scenarios. For example, there is a lot of expensive software that is available to university students free of charge, but almost always under the condition that they will not use it commercially. However, for the purposes of this essay, it is source code that is relevant because this is what is used in training AI.↩
  4. This was originally posted as part of a FAQ webpage of the Chief Information Officer (CIO) of DoD, a CIO which was now moved to the Department of War. The original author, Dan Risacher, moved the FAQ on Github to allow collaborative editing, so the most up-to-date version at the time of writing is located at: https://risacher.github.io/DoD-OSS-FAQ/↩
  5. Which is a challenge to translate because of the word “domaine.” Some translate it as On Real Property. On the one hand the word “property” is too restrictive a translation, but on the other, the seemingly more natural and general—which is convenient for our purpose—term “domain” makes little sense.↩
  6. We should note, however, that “social-media addiction” is not a medically recognized condition. Health professionals still use the term “problematic social-media use.”↩
  7. Which, note, is not necessarily the same as ex nihilo.↩
  8. Which is a recursive acronym that expands to “GNU’s Not Unix!” because in many ways that are primarily technical, and so outside the scope of this essay, GNU resembles the Unix operating system, a landmark in the history of software whose creators got a Turing Award.↩
  9. The main actors were, first of all, Richard Stallman—the software engineer who created GNU—who in 1989 worked closely with the attorney Jerry Cohen to provide the first version of the GNU Public License (GPLv1). Then, Eben Moglen, a law professor and attorney, served as the primary legal counsel for the FSF and played a crucial role in drafting versions 2 and 3.↩
  10. At first sight, this might not make sense: why would one pay for something that they can just download for free? The most common case is that someone needs a modified version of the software for their purposes, or continuous support, but they do not have the expertise to do it. This where companies like Embecosm come into play. It is paid to modify open-source software to meet the needs of its customers, and to also provide customer support.↩
  11. And so even researchers avoid the GPL in the hope that their software will be used by companies.↩
  12. This example is made up, but it is close to real-life examples, like the original license for the game Doom.↩
  13. Thanks to George Liontos for raising this point. ↩
  14. A realistic scenario, which I do not engage here but which is nevertheless interesting, is when AI contributions are provided to human-created software (Cro, 2026).↩