Why the Mutualist License doesn’t have ethical restrictions

We would like to explain in more detail why the Mutualist License (MutuaL) does not include ethical or field-of-use restriction clauses, as our stance is less rigid than many others in the F(L)OSS community.

The harms ethical source advocates highlight are real, and we share their concerns. We also reject the tech-libertarian myth that code is “above politics” in a world already designed to benefit us.

However, we believe addressing these harms through software licenses is the wrong approach. The answer starts with what happens when you attach moral conditions to different kinds of things, and why software makes those conditions dangerous in particular.

There’s an instinct that says: if you create something, you should have some say over how it’s used. This instinct is understandable. When you pour yourself into a creative work (whether that’s a film, a novel, a piece of music, or a codebase), that work carries something of you in it. It reflects your time, your effort, your beliefs, your aesthetic sensibilities.

But this carries something of me and I should control how it’s used are not the same claim, and the gap between them matters more than people often assume.

Consider cultural works, which are, by any measure, the strongest case for creator moral rights. The history of art, literature, and philosophy is largely a history of response: of people taking existing works and making something new from them, sometimes in ways the original creator would have hated. Collage, sampling, remix culture, fan edits. The entire tradition of critical and philosophical writing is built on adversarial readings: taking someone’s ideas and turning them against themselves. Half the French philosophical canon is someone doing this to Hegel, Lacan, Marx, Freud, or Nietzsche. This is not a bug in the system; it’s a feature. A work, once released into the world, becomes raw material for others’ expression—even expression you find objectionable.

This doesn’t mean creators have no protections. If an anti-war filmmaker’s footage is repurposed for pro-war propaganda in a way that implies their endorsement, they have legitimate recourse through defamation or libel law. But the thing being protected there is misattribution, not use. If the propaganda makes clear the filmmaker doesn’t endorse it, what remains is moral offense, and moral offense, however deeply felt, is not something we think the law should generally adjudicate. That path leads toward the logic of blasphemy and sedition.

So even for cultural works, the ground for moral-use restrictions is shakier than it first appears. The legitimate protections creators need—against misattribution, plagiarism, defamation—are already provided by existing legal mechanisms that don’t require moral-use restrictions on the work itself.

Software is not a different category of thing. The art/utility binary doesn’t hold up. Art has utility: propaganda, interior decoration, education, ritual. Tools have artistry: the craft of a well-made hammer, the aesthetics of a machine, the poetry of elegant code. Programming is often creative and expressive; software is also something people run to accomplish tasks.

The usual metaphor is “software is a tool, like a hammer.” That metaphor fails because it erases the expressive side of code: the craft, the style, the design decisions that reflect a programmer’s values and sensibilities. But the reverse metaphor, “software is art, like a novel,” fails for the same reason in the other direction: it forgets the people who depend on software to live and work.

We prefer to think of software as clothing.

A piece of clothing can carry expressive meaning, and you may pour a lot of yourself into its aesthetics—fashion design is a fine arts degree for a reason! But when someone puts on a piece of clothing, it becomes part of their daily life in a way that a book on a shelf doesn’t. They move through the world in it. They depend on it.

The relevant distinction isn’t between “expressive things” and “useful things.” It’s between things people contemplate and things people live in. When you attach moral-use conditions to something people live in—something they depend on for their daily work and activities—you create a form of ongoing moral leverage over their lives that does not exist for works people merely contemplate.

The power dynamic created by moral-use conditions on something people depend on is qualitatively different. Utility doesn’t “override” expression, and how someone receives a work doesn’t change what it is. You’re no longer just saying “this is my expression, respect it.” You’re saying “this is my expression, and because you depend on it, I now have ongoing leverage over what you’re allowed to do in your own life and work.” That is a different kind of claim, and we think it is usually unjustified: the leverage it creates is out of proportion to what a software license can legitimately claim, however right the moral impulse is.

Boring ol’ copyright law

To see why we adopt some clauses that feel like ethical restrictions, such as copyleft, you have to separate a moral request from a mechanical trigger.

Copyleft has a moral motivation, but the clause itself is a mechanical trigger. “If you use X, you must Y.”

• If you use this graphical asset, you must not resell it.
• If you rent this movie, you must not broadcast it.
• Copyleft: If you modify this software and offer it to others, you must relicense similarly.

There is no debate about your morality. It doesn’t require a judge to grade anyone’s soul. It asks a narrower question: did you meet concrete conditions? That’s boring, routine copyright law.

Ethical-use licenses typically operate on the logic of compliance:

• If you use this software, you must not violate human rights.

That invites the hardest kind of dispute: moral interpretation, shifting definitions, and fact patterns that are costly to prove.

This is our first reason for not using ethical restrictions: the moral impulse may be right, but the legal mechanism is structurally mismatched to the goal.

The privilege of enforcement

MutuaL’s stance is that, for software, the honest approach is to protect technological freedoms and fight ethical battles through organizing, law, policy, and culture.

We often hear that this argument—“fight through law and policy”—is steeped in privilege. It ignores the reality that marginalized people, disabled individuals, and those excluded from power often cannot rely on legal or political arenas for protection. That criticism is fair. Infrastructure and the choices we make in designing it are never neutral.

However, the legal system itself is infrastructure. It is the foundation upon which licenses are built. If the infrastructure is corrupt, the courts are corrupt; if the courts are corrupt, all the masterfully worded legal rituals in the world will fail.

Any license, ethical or otherwise, is only a potential trigger for invoking the state. If you make someone sign a contract that says “you promise not to punch me”, there is nothing magical about that piece of paper that physically prevents them from punching you; it only creates a paper trail if they do punch you to later sue over.

Similarly, the most bulletproof ethical license won’t stop MegaCorp from using software in violation of the license unless someone sues them. A license that effectively requires expensive, prolonged litigation to enforce is least usable by the people it is often meant to protect: marginalized individuals, small collectives, precarious nonprofits, and under-resourced communities.

Giving a vulnerable person a license that requires a competent, expensive legal team to enforce is not true empowerment.

Meanwhile, the actors most likely to cause large-scale harm often have compliance departments, outside counsel, procedural patience, and a talent for turning ambiguity into delay. With an ethical clause, you are litigating real-world conduct, often far removed from the software itself; you are litigating politically contestable terms that allow defendants to reframe, delay, and outspend.

Copyleft is significantly more legible, unambiguous, and scope-limited. That’s not to say non-ethical software licenses somehow escape this power imbalance, or that perfect should be the enemy of good. All licenses are imperfect, including ours. Even in an imperfect courtroom, you can more readily prove whether source is offered, notices were preserved, or distribution conditions were met than you can prove that an unrelated employment contract (one you may not even have access to) is unethical in some subjective way.

A copyleft violation can be a quick summary judgment, or at least a relatively short lawsuit. An ethical investigation might take years just for the discovery phase.

The problem with deterrence and “chilling effects”

Some argue that even if lawsuits are rare, the point of ethical licenses is to “chill” compliance departments and create reputational risk for unethical entities.

We are talking about companies that are potentially willing to (or that you believe are potentially willing to, since you’re making it a condition in a license):

• Commit mass murder or automate drone strikes,
• enslave people or exploit labor,
• destroy the planet,
• and/or participate in mass surveillance for profit.

The claim is that these entities—willing to violate fundamental human rights for profit—will be stopped because they fear reputational damage from… violating a software license?

If you could prove these companies were enslaving people, we doubt the public would care more about license noncompliance than about the slavery.

We’re not denying that reputational and compliance risks matter, but they matter most to groups that are likely not your intended targets. Massive, multinational, evil companies like Palantir or Raytheon know people dislike them. They do not care.

By contrast, your local free clinic, underfunded nonprofit, or student group does care. If their overworked lawyer (if they have one at all) says they can’t risk the legal ambiguity of an ethical license, you end up cutting off exactly the people you most want to help.

When bad actors are asked to evaluate necessity and morality, that’s a fight they’re well equipped to win. State actors, defense contractors, and surveillance firms operate under frameworks that legally sanction and frame their actions as moral goods. They have entire legal teams whose job is to argue necessity and compliance. They will redefine “rights”, obfuscate their actions, and dare you to outspend them in court to prove otherwise.

The “chill” lands on the careful and under-resourced, and that trade-off is not clearly worth it to us.

Why we prefer strong copyleft

MutuaL includes a strong copyleft with a network use clause. This does not ask users whether they are “good”; it states that use requires transparency. This isn’t because we believe copyleft is morally pure, but because copyleft changes the economic and operational dynamics of software adoption in a way that depends less on moral interpretation.

Consider the entities we are concerned about: surveillance firms, defense contractors, data brokers, rent-seeking intermediaries, and so on. Their power comes from secrecy, monopoly, and proprietary advantage. Their business model relies entirely on privatizing the gains from software while keeping the source code closed. A data broker only has a valuable product if they’re the only ones with the data.

If a drone manufacturer integrates a MutuaL library, they trigger a legal obligation to publish their proprietary algorithms, tracking logic, and “secret sauce” for everyone to see and for their competitors to copy. This is not a moral request they can lawyer their way out of; it is a mechanical condition structurally incompatible with their existence. They are forced to destroy their competitive advantage (and likely violate their own security clearances) to comply.

Put another way, it doesn’t demand goodness: it taxes secrecy. The good-faith actors ethical licenses often crush don’t derive their power from proprietary advantage. They barely have power to begin with, and, by and large, are happy to share.

Transparency is not a magic wand. Publishing source code can both spread capability and dilute accountability. We still prefer copyleft because:

• it reduces monopoly control,
• enables independent auditing,
• grows a commons that non-powerful actors can build on,
• and increases the odds of real oversight, or at least enforcement.

We don’t claim it solves every risk. We claim it’s a clearer, more structural lever than moral-use clauses in a license.

The problem of the spying shirt

A common counter-argument to our clothing metaphor is that some software is clearly designed for surveillance, targeting, or violence-adjacent purposes. A shirt cannot spy on you, bias your hiring decisions, or automate drone strikes.

Yes, software can be inherently violence-adjacent. But licensing still cannot reliably distinguish between ‘intrinsically harmful’ and ‘potentially harmful’ in a way that is enforceable, predictable, and free of collateral damage. The more dangerous a tool truly is, the more the appropriate governance layers tend to be regulation, procurement bans, professional standards, and oversight, not private copyright conditions based on contested moral judgments.

This also shifts the conversation in two ways.

Illegal behavior

For illegal behavior, this shifts the conversation from licensing to criminality. It is already against the law to violate the law. You do not reduce crime by writing “do not use for theft” on a crowbar. The person buying it for a crime will commit the crime anyway, and if they’re caught, the prosecutor will bring charges for theft regardless of your preferences.

Trying to make an illegal act even more illegal is not a coherent strategy. It also raises uncomfortable questions about presumed innocence and when exactly the license should terminate: upon arrest? Upon charges being filed? Upon conviction? What about when a case is dismissed but the court of public opinion still believes the party is guilty, justly or unjustly?

If they ignore the license entirely because they are multinational billionaires with battalions of lawyers and a belief that might makes right, then no license, ethical, copyleft, or permissive, would have stopped them anyway, and attempts to do so are wasted effort.

Legal behavior

If human rights violations or environmental destruction are legal for the entities committing them, the conversation shifts from licensing to private veto power.

At that point, you are no longer “adding ethics” to software use. You are attempting to use copyright—a tool designed to regulate copying, modification, and distribution—as a substitute for public regulation across jurisdictions.

Sometimes this is emotionally satisfying because it feels like refusing complicity. But structurally, the problems compound.

It asks courts to decide contested moral and political questions as a condition of a copyright grant. Even if a judge personally agrees with your ethics, which is already taking a lot on faith, they still have to interpret vague terms (“human rights,” “harm,” “facilitate”) and decide what evidence is sufficient. This is not a narrow dispute about notices or source availability; it is a full-blown trial about real-world conduct.

When the conduct is legal, the clause has to lean on definitions that go beyond the law. Either you tether it to existing legal standards (in which case you mostly reproduce whatever the state already permits), or you define “human rights” more broadly than local law (in which case you ask a court to enforce a private moral regime that the jurisdiction has yet to adopt). The result is unpredictability: what counts as “a violation” will vary by country, court, decade, and the political weather that week.

And it concentrates power in the hands of those who can afford to litigate. If the state will not stop a harmful practice, your license does not provide additional enforcement power. It becomes a tool invoked selectively—sometimes against easy targets, often not against the worst offenders—and selective enforcement is not the same as harm reduction.

None of this means your moral outrage is wrong. It means that when the conduct is legal, the fight must focus on changing laws, procurement standards, professional regulations, and material incentives, not on trying to smuggle governance in through a simple permission slip.

The danger of performativity

Licenses are certainly declarations of values, but we must ask: how many of these declarations are non-performative?

If you hate war, you probably aren’t contributing to the @ParamilitaryCo/Kill-Bot-Efficiency repository in the first place. If you’re working on an emulator for an arcade machine you loved as a child, no one is using it for harm anyway. If you’re writing a tiny utility library, it will likely be buried so deep in a dependency chain that licensing decisions become theoretical calculus.

The danger lies in mistaking a value signal for actual harm reduction.

If a clause is too ambiguous to enforce, too costly to litigate, or too easy for powerful actors to circumvent, it may serve mostly as a declaration that makes the author feel (understandably) less helpless. Helplessness, however, is not cured by clumsy attempts to litigate one’s soul.

Signal your values. Sing them loudly! But we can’t think of anything less punk rock than a contract meant to be enforced by a broken system. An ethical covenant, a code of conduct, or a set of community standards already does the signaling and the community building. What do clauses in a legal document add? We doubt they unlock collaboration you wouldn’t get by stating your values, and we do not find “someone needs a legally binding ethical restriction before they contribute” a compelling practical consideration.

If your goal is to reduce harm, the most effective approach we’ve seen is:

• choose a license that builds a durable commons and discourages reckless pursuit of profiteering from proprietary advantage, and
• fight ethical battles where they can actually make an impact: labor organizing, procurement standards, public policy, professional regulation, coalition pressure, and community governance.

What would change our mind

We do not treat this as an article of faith. If ethical-use restrictions reliably reduced harm in practice, we would want to know.

The kind of evidence that would move us would look like:

• documented cases where a major harmful actor stopped or changed behavior primarily because of an ethical-use clause, or
• a meaningful body of court decisions interpreting and enforcing such clauses consistently across at least one jurisdiction, or
• clear patterns showing these licenses do not disproportionately exclude good-faith, under-resourced users.

We have not seen this evidence. We have seen strong structural reasons to expect uneven enforcement and unacceptable collateral damage.

First, do no harm

The current head of MutuaL is a healthcare worker. Medicine’s “rule zero” is primum non nocere: first, do no harm.

It is not better to try medications at random for a difficult-to-treat problem. Without a clear, evidence-based plan, you will harm the patient more than you help them.

There is no evidence that ethical clauses work, and good reason to doubt their efficacy and, more importantly, their safety. We don’t want to see well-meaning people metaphorically prescribe broad-spectrum antibiotics when they haven’t yet ruled out the common cold.

We want to build a world, and a software commons, where harmful power is harder to consolidate, secrecy is harder to monetize, and exploitation is less profitable.

---

Specific ethical licenses

We really wanted to be able to end this with “but if you feel you must use an ethical license anyway, please use a well-designed one like…”

But when we looked for a well-designed one, no matter how many we went through, we found nothing that satisfied us. Here’s a list of every notable ethical source license we found, as well as a non-exhaustive list of “terminal” problems with each.

Hippocratic License

The license is modular, with sixteen optional modules you can include or leave out. HL3-BDS-ECO-EXTR-FFD-MY-TAL-USTA-XUAR has vastly different requirements than HL3-BOD-CL-LAW-MEDIA-MIL-SOC-SUP-SV.

There is quite literally nowhere on Earth less suitable for a Choose Your Own Adventure® book than a legal document in a courtroom. It makes “licensed under The Hippocratic License v3” a meaningless shorthand, and it stops you from knowing your rights at a glance.

Learn what the Apache 2.0 License requires once, and you know how you may use an Apache-2.0-licensed work every time you spot it in a repo or a file header.

No one can remember every possible combination of modules. There are 2^16 unique configurations, or 65,536. That many variations is a compliance nightmare.

If we put that questionable design decision aside, that mostly leaves two meaningful configurations: HL3-CORE and HL3-FULL (no modules and all modules, respectively).

We’ll take them one at a time.

HL3-CORE

Most of HL3‘s ethical clauses seem to be based on international definitions that are reasonably immutable and limited in scope, such as the United Nations Universal Declaration of Human Rights. We have some issues with the clauses:

Harm the environment in a manner inconsistent with local, state, national, or international law.

It is already against the law to violate the law.

The Licensee SHALL:

• Provide equal pay for equal work where the performance of such work requires equal skill, effort, and responsibility, and which are performed under similar working conditions, except where such payment is made pursuant to:
  • A seniority system;
  • A merit system;
  • A system which measures earnings by quantity or quality of production;

“Merit”, here, is so broad and subjective as to allow nearly any justification for any pay discrepancy.

But that belies the much larger issue:

SUPPLY CHAIN IMPACTED PARTIES:

In addition to the above Ethical Standards, Licensee voluntarily accepts a Duty of Care for Supply Chain Impacted Parties of this License, including individuals and communities impacted by violations of the Ethical Standards. The Duty of Care is breached when a provision within the Ethical Standards section is violated by a Licensee, one of its successors or assigns, or by an individual or entity that exists within the Supply Chain prior to a good or service reaching the Licensee.

Breaches of the Duty of Care, as stated within this section, shall create a private right of action, allowing any Supply Chain Impacted Party harmed by the Licensee to take legal action against the Licensee in accordance with applicable negligence laws, whether they be in tort law, delict law, and/or similar bodies of law closely related to tort and/or delict law, regardless if Licensee is directly responsible for the harms suffered by a Supply Chain Impacted Party. Nothing in this section shall be interpreted to include acts committed by individuals outside of the scope of his/her/their employment.

This requires a little bit of breaking down.

Usually, a “Duty of Care” is established by laws and social norms (e.g., a doctor has a duty to a patient; a driver has a duty to pedestrians). Here, the license says that the “Licensee voluntarily accepts a Duty of Care” to anyone affected in the “Supply Chain”. They define “Supply Chain” like this:

“Supply Chain” means the sequence of processes involved in the production and/or distribution of a commodity, good, or service offered by the Licensee.

Simple English translation: the series of steps that happen to make and deliver a product or service.

It expressly includes any harm done by anyone upstream of those who adopt the license. In other words, by potentially doing nothing more than npm installing a package, you voluntarily accept the Duty of Care. Then, the moment you offer a product or service, you become liable for the entire supply chain behind that product. The “and/or distribution” clause could be interpreted to include liability for someone who happens to be injured by a delivery truck carrying something you shipped, given that the Ethical Standards include “Infringe upon any person’s right to life or security of person”.

This is not hyperbole or bad faith interpretation. The license specifically goes out of its way to indicate that they can sue, quote, “regardless if Licensee is directly responsible for the harms suffered”. And by framing it as tort law, it means that anyone who believes they are affected can sue not just for the usual license standard of ceasing to use the software, but for uncapped (from the license’s standpoint) monetary damages.

The intent is clear and noble: they want organizations to put forth some effort to audit their supply chain. You can’t just say, “Well, I treat my workers well,” if you are knowingly buying your servers or raw materials from a company that uses forced labor (e.g., conflict minerals).

Unfortunately, the execution is horrific. The phrase “regardless if Licensee is directly responsible for the harms suffered” is legal nuclear waste. No one, at any size or level of morality, is going to sign a contract that says “you may be sued at any time, for any reason, for anything that someone else did, as long as you happened to rely on them in some material way”.

Consider: you wouldn’t sign a contract stating that you’re liable for any crime any of your friends commit, now or in the future, whether or not they are still your friends at that time.

Will they be successful in suing you? Maybe, maybe not. But defending yourself takes money, time, and willpower. Innocent people have been bankrupted by legal fees for cases that ended up getting thrown out.

The license doesn’t even cap how many “layers deep” the supply chain liability goes. Let’s say you are willing to take on the responsibility of regularly, endlessly auditing every single action every single one of your suppliers or logistics partners take. Are you also going to audit every single one of their suppliers? As-written, this covers anything up to and including the manufacturer of the paper you printed someone’s invoice on (and the company that sold that paper to you).

We appreciate the intent, and we know lots of companies love to whitewash their reputation by kicking responsibility down or upstream. This is an absolutely radioactive way to handle that, though, and we could not recommend it for any use by anyone, even private individuals pursuing hobbies.

We’re not even getting into the question of whether it would be grounds to just throw the whole license out. It has a severability clause, sure, but legal systems generally do not like when contracts try to rope in arbitrary, uncountable third parties who never signed (or were even made aware of) a contract. You cannot use a copyright license for a Rust library to create a global duty to unnamed villagers continents away—and if you try, a judge might get snippy enough to throw the entire thing out on the basis of sheer absurdity.

HL3-FULL

The license was already unusable, but the modules introduce additional problems.

Be an individual or entity, or a representative, agent, affiliate, successor, attorney, or assign of an individual or entity, identified by the Boycott, Divestment, Sanctions (“BDS”) movement on its website as a target for boycott;

Be an individual or entity, or a representative, agent, affiliate, successor, attorney, or assign of an individual or entity, on the FFI Solutions Carbon Underground 200 list;

Be an individual or entity: which U.S. Customs and Border Protection (CBP) has currently issued a Withhold Release Order (WRO) or finding against based on reasonable suspicion of forced labor;

Contracts should never have clauses which can pull the rug out from under you based on mutable, external criteria. No one should ever be compliant at one point and then liable for a violation of the license at a later point based on no change in their actions or behavior.

Regardless of personal beliefs regarding these clauses, they all rely on living third-party documents that are regularly updated as circumstances require. Anyone at that third party can, at any time, insert anyone onto the lists with zero oversight.

Saying some legally precise version of “don’t use forced labor” is static: you can accept the contract once after reading it and know that you’re in the clear as long as you never choose to use forced labor. You’d have to consciously change your behavior to be in violation. You can’t say the same when it is based on the opinions of others.

We listed this license first for a reason. Every license below this point isn’t half as egregious as the Hippocratic License. Even if you disagree with every single point we made in the first half of this article, at least don’t use this ethical source license.

Similarly:

THE ANTIRACIST ETHICAL SOURCE LICENSE/At The Root

Based on the HL3 and includes the same duty-of-care non-starter.

Do No Harm License

This software must not be used by any organisation […] that: promotes, lobbies for or derives a majority of income from:

“Derives a majority of income” strangely overspecifies the moral condition and creates a loophole as a result. As long as your “slavery or indentured servitude”, “war crimes”, or “extraction or sale of fossil fuels” only makes up 49% of your revenue in total, you are apparently in the clear. You could run a human trafficking ring, according to this license, just as long as it remains a side hustle.

The entire license seems to oscillate between vagueness and overspecificity, honestly. In one section, it forbids something as vague as “industrial processes that generate waste products that threaten life” (so a pharmaceutical company manufacturing curative chemotherapy for cancer?); in another, it goes out of its way to define these:

“Forests” shall mean 0.5 or more hectares of trees that were either planted more than 50 years ago or were not planted by humans or human made equipment.

“Deforestation” shall mean the clearing, burning or destruction of 0.5 or more hectares of forests within a 1 year period.

Now we’re back to the side hustle problem. Just make sure you only clear 0.49 hectares of ancient redwood trees, we suppose?

Opinionated Queer License

Okay, we love this name. Unfortunately, the license itself is somewhat of a highlight reel of every problem typical to ethical source licenses.

No one may use this Work for prejudiced or bigoted purposes

Against whom, where, how, why, and in what way?

including but not limited to: racism, xenophobia, queerphobia, queer exclusionism, homophobia, transphobia, enbyphobia, misogyny.

If the author’s an “opinionated queer”, they should be more aware than anyone of how much in-fighting and definitions-based arguing goes on in the queer community. “Enbyphobia” and “queer exclusion” is going to start no shortage of arguments, and the terms themselves have no particularly clear legal boundaries. Courts will do what courts always do with vague moral language: launder it into whatever the judge already thinks.

If you’re non-binary, do you really want a potentially 75-year-old conservative judge who thinks “cisgender” is a slur litigating what discrimination against you looks like? What about in another jurisdiction more conservative than yours?

The license then tries to anchor at least some of its terms, but in horrific ways:

No one may use this Work to inflict or facilitate violence or abuse of human rights, as defined in either of the following documents: Universal Declaration of Human Rights, European Convention on Human Rights along with the rulings of the European Court of Human Rights.

Like with the Hippocratic License, we’re willing to give a pass to the mostly stable and limited UN/EU definitions, but “rulings of the European Court of Human Rights” is instantly disqualifying. The rulings of a court (that you may share no jurisdiction with) are a living, expanding body of case law that changes over time and is not realistically knowable “at a glance” by ordinary users. It’s barely knowable by lawyers.

The HL3-FULL’s fatal flaw was pinning license terms to mutable third-party lists, but their mutable third-party lists at least were run by organizations that had a stable ‘purpose’ you could broadly anticipate. The OQL went with an entire court’s past and future history of jurisprudence.

Then we go categorical in the ways you’d expect (law enforcement, military, immigration enforcement, etc.), but with a subtly problematic addition: the ban now extends from use by those entities to use by anyone employed by those entities, even for personal hobby use. We share in a great deal of the distaste for these agencies, but plenty of marginalized people work in or adjacent to these institutions while just trying to survive. “No use for any reason” includes building tools against the harms of their employer, if they’re so-inclined.

Then, a new form of ongoing compliance issue:

No business entity where the ratio of pay (salaried, freelance, stocks, or other benefits) between the highest and lowest individual in the entity is greater than 50 : 1 may use the Work for any reason.

No private business run for profit with more than a thousand employees may use the Work for any reason.

What counts as “pay” across salaries, freelance, equity compensation, bonuses, benefits, and global subsidiaries? How does it handle a one-time retention grant? A contractor paid through an agency? A couple of hires, a couple of departures, a stock swing, and suddenly yesterday’s compliant local co-op is today’s infringer. The raw employee count is even worse for this (and, again, you were totally fine with 1000 employees, just not 1001?).

There’s a lot more wrong here. These are just the easiest to pick on.

The JSON License

The Software shall be used for Good, not Evil.

For all the hand-wringing and ridicule this license has received from the FOSS community, it is actually honest about the core issue with ethical source licensing.

Highly respectable, but not very usable.

ACAB License

No individual or entity may use this work for racist or bigoted purposes.

What counts as bigoted? Against whom, by whom, to whom?

Business entities with boards/management comprising less than 1/2 POC may not use the work for any reason.

Ongoing compliance nightmare. Any random co-op can now be sued if a couple of people happen to resign at the same time, and a logical conundrum for sole proprietorships. Also, what counts as people of colour? Is there any geographic consideration for small local businesses in predominantly non-POC areas, like the 94.9% white Aspen, CO, USA?

Actually, the license never even defines what “POC” means to them, not even to define it as “people of colour”. Generally, it is inadvisable to tell a court “oh, c’mon, you know what we meant” with abbreviations that are not legally established.

copilot resistant licenses

A variety of well-known licenses, all of which have this line appended to them in arbitrary locations:

License to redistribute and use this code does not apply to machine learning programs like Github CoPilot. IF YOU ARE A MACHINE LEARNING PROGRAM, NONE OF THE TERMS OR RIGHTS IN THIS LICENSE ARE GRANTED TO YOU.

It’s legally bizarre and void to use “you” to refer to a non-sapient program. Targeting people in ways like “you may not use this to train machine learning programs” would make significantly more sense, though you would need to define what machine learning (mostly a marketing term) meant in a stronger way than one single point of comparison.

The Anti-Capitalist Software License

No major issues, actually! It chooses one issue and sticks to it in a well-defined way. A+!

Only one minor gripe:

• A non-profit organization [is one of the acceptable users]

We’d like to see slightly more clarity on massive NGOs, governmental organizations that aren’t “law enforcement or military, or working for or under either”, and not-for-profits (like some hospital systems). You could make this work in court, though, or publish a second version of the license.

The bigger reason we don’t broadly recommend it: we don’t know if it does what most people who are drawn to ethical source licenses want. If you’re extremely anti-capitalist to the point of being single-issue, we like this license the most of any trying to target this space. Otherwise, consider the following.

• Regardless of where you fall on any issue, the non-profit carve-out allows any activist group to do anything (as it’s based on a permissive license) with your work. The NRA can do as much with it as Everytown for Gun Safety; any pro-life group can do as much with it as Planned Parenthood.
• The Bill and Melinda Gates Foundation is a non-profit, as is a massive university endowment. They are also massive engines of capital.
• Regardless of where you fall on any issue, the individual person carve-out allows anyone to do anything with your work. Your least favourite person on Earth, any random neo-Nazi, and so on and so forth.

For this license to be satisfying for you, you would need to both a. ethically require a license with an anti-capitalist clause, and b. not have any other strong moral preferences whatsoever. If that describes you, however, we do endorse this one. It’s reasonably close enough to a mechanical trigger.

Similarly:

"Anti 996" License

If you’re fully “single-issue” for this individual labor right and nothing else, we have no issues with and endorse this license.

ml5.js License

You must use this software in compliance with the ml5.js Code of Conduct as it exists at the time of your use. This includes cooperating with any investigations of your compliance with the ml5.js Code of Conduct.

“Embed the Code of Conduct into your license” isn’t really what we meant when we endorsed Codes of Conduct.

To start with the positives: “as it exists at the time of your use”, as well as the fact that it’s their own code of conduct (not external), effectively solves the ‘mutable external resource’ critique. It’s an internal resource with immutability baked in.

(If we wanted to be maximally cautious, we could nitpick that “at the time of your use” could be more specific: with how modern software tooling works, it’s unclear if npm update years after you first imported ml5.js counts as a reset on “time of your use”. Even if you fully read the Code of Conduct at the time of first installation, there’s not a lot of precedent to resolve interpretational divergence such as this, potentially making it more threatening than desired. This is a nitpick, though, and is well within standard contract ambiguity.)

However, merely kicking the terminal imprecision to a separate document does not fix the core issue of imprecision. From the CoC itself:

Do

• Be mindful of your language.

Do Not

• Engage in any of the following types of behavior:
  • Offensive comments related to gender identity and expression, sexual orientation, race, ethnicity, language, neuro-type, size, ability, class, religion, culture, subculture, political opinion, age, skill level, occupation, or background
    […]
  • Or any other kinds of harassment

As a Code of Conduct itself, these are great. They’re values signals and not legally binding. As legal text, “be mindful”, “offensive comments” (offensive to whom?), and “or any other kinds of harassment” (a liability wildcard depending on who defines harassment) are non-ideal. These are bound to “Behavior within the community”, however, so choosing not to engage in the community is an easy enough bandage.

Worse, though, is the “Use of the ml5.js software library” section:

Do

• Teach with ml5.js
• Make creative projects with ml5.js
• Use ml5.js for social good

This is “for Good, not Evil” in fancier language.

[Do Not] Use ml5.js to build tools of mass surveillance and prediction to repress the rights of people

• Machine learning can be used to analyze data about large numbers of people and to use that data to predict future trends. Although there can be legitimate uses of these tools, they are too often used to repress the rights of others.

A variant of the “side hustle” problem: we can build tools of mass surveillance as long as they don’t “repress the rights of people”? Worse, “repress the rights of people” is totally undefined. ‘Any right of anyone’ is a complete wildcard in the context of legal liability. It’s not anchored to anything (like a UN declaration), nor is it given any guidance.

“This includes cooperating with any investigations of your compliance” is equally undefined, and unlimited in scope in ways we find uncomfortable. As-written, anyone on ml5.js’s team can launch “any investigation” for any reason, and it’s unclear exactly how much “cooperation” you must provide and what that cooperation looks like to them. You’re placing significant trust in a group that you are merely importing for your own work because it came up in an npm search for “machine learning”.

Again: as a Code of Conduct, it’s good. The world does not end if you get banned from their forum, and “we’ll know it when we see it” is fine for informal community standards. The issue is constructing a legal document with ‘please don’t harsh our vibes’ in it by proxy.

Each contribution to this software is licensed under this license for three years from the date it is contributed to the software. After that three year period the contribution is licensed under the Blue Oak Model License v 1.0.0.

This also strikes us as a somewhat odd decision. There are other modern attempts to make licenses that include time-limited requirements, such as copyleft-next, but that license makes no attempt at ethical restriction. We’re unclear why someone would wish to prevent what they see as unethical use only temporarily. (Also, copyleft-next’s sunsetting is after fifteen years; three years seems unreasonably short either way, even just as copyleft.)

The Non-Violent Public License Family

This license tries to be a hybrid cultural works/software license, which we're not generally fans of. The two fields have different needs and wording requirements, and trying to combine the two usually either fails one or significantly bloats the license (making it harder to understand your rights and obligations).

Beyond that, we like this one the best of all of the ones we’ve listed here (at least as far as “multi-issue” licenses go), and its Cooperative variant manages an anti-business clause with reasonable wording.

We have a couple of issues, but seeing as how this license was updated just two months before this was first written, we have hope it evolves into a state where it can be the one we definitively recommend for those unconvinced by our primary argument.

Incarceration is Confinement in a jail, prison, or any other place where individuals of any kind are held against either their will or (if their will cannot be determined) the will of their legal guardian or guardians unless otherwise specified or in cases where there has been known antagonism between them and their guardian or guardians. In the case of a conflict between the will of the individual and the will of their legal guardian or guardians, the will of the individual will take precedence.

Written in an ambiguous way that leaves us unsure about edge cases, but more importantly, appears to expressly include acts like necessary involuntary psychiatric holds. Though there’s no doubt that some involuntary psychiatric holds are abusive, there are necessary cases, and this definition makes no attempt to include even subjective (e.g., “non-necessary”) exceptions.

Surveilling is the use of the Work to either overtly or covertly observe and record persons and or their activities.
[…]
You do not use the Work for the purpose of Surveilling or tracking individuals for financial or private gain without revokable consent. Where revokable consent requires presenting the option to decline Surveillance or tracking forever before collecting any information, and allows withdrawing consent forever at any point.

More issues with ambiguous writing. Also, “financial or private gain” is doing a lot of heavy lifting it may not be able to sustain for things like forensics and fraud prevention, and “allows withdrawing consent forever at any point” creates an infinite liability loop we’re not comfortable with as-written. If the surveilling is consented-to and for a medical study, does this include after publication of the final research paper?

To Discriminate is use of a work to differentiate between humans in a such a way which prioritizes some above others on the basis of percieved membership within certain groups.

A license that defines discrimination! See, NVPL, this is why you’re our favorite candidate.

… Unfortunately, though, it’s not the best definition. It’s really, really vague, and “differentiate” and “prioritizes” are verbs that don’t cleanly map to specific actions or violations. What about population research that excludes certain groups for not being relevant to the study’s end goal, such as for sickle cell anemia?

Massive points for effort, but the execution needs work.

Miscellaneous personal exclusions (e.g., this license modification)

There exist an uncountable number of edited standard licenses in various personal repos with additions such as:

This license is granted to everyone except for the following entities and any of their subsidiaries:
[…]

• “United Parcel Service Co” (for working with ICE)
• “Walmart Inc” (for abusive treatment of workers)
  […]

These seem unsustainable to us. Are you going to make a new license version every time a new company makes the headlines for unethical behavior? Any attempt to manually enumerate every potential unethical actor is doomed to rapidly becoming outdated, ineffective, and unrepresentative of one’s views.

Beyond that, we fear potential confusion for developers who skim and see, e.g., “MIT License” in the text and don’t investigate further. Distinct licenses should clearly be distinct licenses.

---

Please get in contact if you know of more ethical source licenses you’d like us to review, whether for inclusion or analysis.