RAIL: Nonfree and unethical

For over 20 years, the FSF Licensing & Compliance Team has been the preeminent resource of free licensing for free software developers.

Software freedom means that other people cannot use software to exercise power over users, like tying them to a particular vendor or holding their data or privacy hostage. Free software licenses, and especially strong copyleft licenses such as the GNU General Public License (GNU GPL), have been instrumental in preventing the social injustices that result from using software to control what users can do on their computers.

It is probably due to the success of free software licensing that some people have started to perceive licenses as a tool for addressing social injustices generally. Every now and then, though, someone makes an attempt to draft a software license that, at first glance, looks commendable: it lists a number of anti-social activities and requires that licensees refrain from these activities as a condition of the license. Such licenses are often advertised as "ethical," but make no mistake: they deny users their software freedom and therefore are unethical. Software freedom includes freedom 0 (the freedom to use the program for any purpose). Clearly, any use restriction in a software license makes the program nonfree. RAIL are an example of such unethical licenses, and we urge people not to use them.

The Free Software Foundation's (FSF) Licensing and Compliance Lab maintains a list of licenses classified by their freedom, including whether they are copyleft licenses and compatibility with the GNU GPL. It would be impractical for us to list all existing nonfree licenses, and we do not hurry to include every new license in this list. But, in this case, we decided to explicitly state that RAIL are nonfree by including them in our nonfree list because we still receive questions about them, despite these licenses having not gained much popularity. An additional important reason for us to list RAIL is that they are marketed as addressing ethical challenges related to machine learning. Machine learning has become an important area from the point of user freedom, and it is essential to stop the spread of unethical machine learning licensing. We explain threats of use-restricting licenses in more detail below.

Restricting software freedom increases social injustice

Free software has became a worldwide standard and is the basis of many implementations. Anyone can choose from thousands of programs available for any purpose, and indeed, almost everyone builds their tools with free software. One of the results of freedom being the standard is that there are many developers who can contribute to free software even if they gained experience on private, in-house implementations. Widespread interoperability relies on free software being the default. Notably, this reality is only possible because of half a century of challenging work and commitment from everyone in the free software community, in spite of SaaSS and companies that appropriate free software and contribute nothing in return.

If we lived in a world where software subject to use restrictions was the norm, we would be much worse off. Many law-abiding individuals would need to expend significant resources on auditing their software use or seeking differently-licensed replacements. With restrictions phrased so that they can be interpreted to include almost anything, a user could face a copyright lawsuit whenever the licensor does not approve of a particular use. If more people started drafting licenses with custom use restrictions, the risk would grow even more dire. Users would have to check all licensors' policies (subject to change without notice), all restrictions, and exclude programs under conflicting ones before running a given piece of software. It would be very difficult to form a community collaborating on and sharing programs if use-restricted software becomes standard.

Notably, bad actors (those that RAIL and other similar licenses try to exclude) would not have much difficulty in using software regardless of license conditions. Most, if not all, activities prohibited in RAIL are already illegal in most countries. If some people or organizations are not stopped by the law as enforced by the government, why would they yield to a copyright holder threatening to sue them over breach of a license? Restricting uses of existing software would also not stop the wealthy from developing their own software and locking users in with proprietary interfaces. With licenses such as RAIL, there can be no shared code between proprietary and free software, and developers taught to work on nonfree software cannot readily contribute to free software. Normalizing licenses such as RAIL would add to the difficulties already faced by the free software community coming from SaaSS and free software appropriation.

Additionally, introducing license restrictions on users with morally loaded terms ("harming," "discriminating," "purposes of deception," to use examples from RAIL) means that licensors claim legal authority for themselves that can be exercised justly only through democratically-elected lawmaking. The outcome of use-restricting licenses is outsourcing the law to individuals, organizations, or businesses; the opposite of democracy. Licensors, driven by private interest would quite likely invoke the anti-harming, anti-discriminating, and anti-deception language of RAIL to protect themselves instead of the public.

Use-restricting licenses do not address ethical issues of machine learning

RAIL are presented as being designed to address ethical challenges related to machine learning applications. However, they do not require anything that is really necessary for users to control their computing done with machine learning, including: complete training inputs, training configuration settings, trained model, or — last, but not least — the source code of software used for training, testing, and running tools based on machine learning. Thus, RAILed machine learning can be, and most probably will be, unethical. Use restrictions do not prevent these licenses from being used to exercise power over users.

As a side note, RAIL refer to machine learning as "artificial intelligence" and completely ignore the fact that such applications have no actual intelligence as they do not have an understanding of anything. Using this term does a lot of harm, especially to less technologically-savvy users, many of whom have already started to perceive tools developed with machine learning as credible sources of knowledge and trust them when looking for complete solutions for real-life problems. RAIL contribute to unethical marketing of machine learning, again under the disguise of morally-loaded restrictions they purport to enforce.

Copyleft licenses are essential for public good

If we want software to help decrease social injustice, we should oppose licenses that restrict how software can be used. We should focus on effective ways of addressing injustices: government and community support for freedom-respecting tools and services; releasing programs under strong copyleft licenses; and entrusting copyrights to organizations that have the resources to enforce copyleft.

Software freedom must be defended, not denied. More specifically, the more free software is out there, the more likely people will collaborate on tools and services that do not pose moral dangers and help solve existing ones. Free software also makes it more likely that users have real choices when looking for freedom-respecting ethical programs and tools based on machine learning. Denying people the freedom to a particular program, as RAIL or similar licenses would have it, prevents them from using such program for the common good.

If you liked this article, consider supporting our work to protect essential freedoms for all computer users.

Donate | Become an associate member

"Licensing & Compliance Lab logo" © Free Software Foundation, Inc., is licensed under a Creative Commons Attribution 4.0 International license.