Introduction of an anti-LLM policy?

[cloudskater]

Hi!

I'm a user of Liberapay who is really appreciative of this service that allows me to donate money to people and projects without the use of a corporation like Patreon. I mean, aside from the payment processors, but that sadly can't be helped at the moment.

I really appreciate your lack of support for cryptocurrencies, as they are quite harmful, and I would be ecstatic to see this project adopt a policy outlawing the use of LLMs, which most commonly take the form of "generative" AI today. Even small scale use is incredibly destructive to the environment and those already living in less-then-ideal situations, and that's not mentioning the theft of people's writing, visual art, code, music, you name it. It's all fed into a machine that can only thoughtlessly reproduce what it's given without attribution, under the guise if it somehow being new.

So you can see why I, and I'm sure many others, would be happy to see a strict anti-LLM policy adopted by Liberapay as a project. If you want more context and proof on the topic, as well as issues I didn't cover in this blurb, I hope the following details are useful!

Environmental Impact

The Wikipedia page on the environmental impact of artificial intelligence is a good place to start, but here are some more specific sources:

• MIT News: Generative AI's environmental impact

• MIT Technology Review: We did the math on AI's energy footprint. Here's the story you haven't heard.

Effect on Hardware Prices

The demand for construction and outfitting of new data-centers to host AI/LLM compute capacity has overwhelmed global supply and production of multiple hardware components. Memory, Storage, and GPUs have seen massive increases in price for both consumer and enterprise models, upwards of 400% in some cases.

The lack of supply has led large system-builders to purchase production capacity from OEMs well in advance of delivery, leading some manufacturers to end consumer-oriented product lines in favor of enterprise capacity.

The downstream effects for consumers is that nearly all electronic devices which contain memory and storage will see their prices rise and availability decline. Those who already own existing electronics and computer hardware components may also find themselves unable to have their devices repaired or replaced under warranty.

• The RAM shortage is coming for everything you care about

• Western Digital is already sold out of hard drives for all of 2026 — chief says some long-term agreements for 2027 and 2028 already in place

• The 2026 storage crisis: Why AI data centers are hoarding every hard drive on the market

This has resulted in pricing historically marginalized individuals, who are more likely to have less financial resources, out of learning about computers at home and shrinking the diversity of the computing and programming industry.

Ties to the War Industry

A lot of AI companies also work directly with nation states for use in their War Departments, which in turn leads to further AI usage during war and invasions. This is coupled with NYT: Palantir, Anthropic and small start-ups are generating rewards from their investments in defense tech.

As another example NPR: OpenAI announced Pentagon deal after Trump banned Anthropic which was due to the USA Department of War launching an AI acceleration strategy.

Due to the nature of LLMs being only kind of as good as the data they are trained on, this can lead to additional civilian deaths and housing/infrastructure damage either intentionally or not. Examples:

Content Warning: War details, death

• Gaza: UN experts deplore use of purported AI to commit ‘domicide’ in Gaza, call for reparative approach to rebuilding
• Lavender & Where’s Daddy: How Israel Used AI to Form Kill Lists & Bomb Palestinians in Their Homes
• Microsoft says it provided AI to Israeli military for war
• Google has dropped its promise not to use AI for weapons
• "We want to use it for everything": How Project Maven became central to America’s AI-powered warfare

What You Can Do

I'd like to kindly suggest implementing a strict "No LLM" policy. This can also help with what seems to be larger and larger influxes of low quality drive-by PRs by people using LLMs.
While it is impossible to fully vet every commit for LLM usage (nor should it be expected), a strict policy will deter those tainted pull requests. It also gives moderators a rule to point to in the case that usage is discovered, allowing them to more easily take action against those who violate the guideline.

A list of some projects that have taken steps to refuse AI usage can be found here:
https://noai.starlightnet.work/list.html

As for a specific policy, I can recommend the following policies for inspiration:

• wafrn's Contributring Doc

• Loupe's Use of Generative AI Policy

Thank you and kind regards,

cloudskater

[MixusMinimax]

Throwing my name in the hat. I'd also like to see this.

First of all, great project, I find it very important. Thank you for all your work.

I want to add a couple more points:

1. Software development worked fine before LLMs even existed, so there really isn't an argument strongly in favor of a need for LLMs. Offsetting all the negatives is in my mind a pretty high burden to overcome.

2. Human developers can be, and often are, very skilled and experienced. Offloading work (or even parts of it) onto an LLM has a high likelihood of losing those skills; so-called "deskilling" is a real issue. Ask yourselves this: Would you want your contributors to learn from mistakes, to improve themselves, and take pride in their work? I think we all do. That way, it reduces the burden on reviewers, too.

   It doesn't matter if you put more importance on ethical issues, code quality, deskilling, human rights, legal issues (copyright), it really doesn't matter what angle you approach the topic from, the answer is always the same. I'm sure someone is willing to elaborate, if needed.

3. But maybe you already agree with this, but think "but can we even enforce this?" To that I say: Doesn't matter. A policy is just a policy, it should be written based on what is right, not what is enforceable. How or if to enforce it, is a completely separate discussion.

   A policy based on the honor system and the judgement of reviewers is better than no policy. I would argue, it's usually actually quite effective.

4. Another common question: "But we don't want to tell people how to do their work? We're scared of backlash, we can't make everyone happy."
   To that I say: Don't make everyone happy, make the people you care about happy. You will either lose vibe coders, or lose experienced developers, by forbidding or allowing AI. There is no middle ground.

Please consider this proposal, I find it very reasonable. It can't really hurt you either.

[cloudskater]

A policy is just a policy, it should be written based on what is right, not what is enforceable. How or if to enforce it, is a completely separate discussion.
   A policy based on the honor system and the judgement of reviewers is better than no policy. I would argue, it's usually actually quite effective.

Not strictly relevant but this is the best way I've ever seen this point expressed, I love your phrasing. Hope you don't mind if I replicate it in the future! lol

[MixusMinimax]

I'm honored! I guess I am a professional rambler after all haha

[abdlfc11]

I think SDL’s policy is quite fair:

AI can be used to identify issues with contributions to this project, but the solutions to those issues should be authored by humans.

Using AI to find problems is fine. Accepting fully AI-generated PRs, however, is a different matter, it tends to introduce subtle errors, dilute accountability, and make review harder.
That said, I don’t think the only options are “ban AI completely” or “allow it with no rules.” A middle path seems more practical:

• Require contributors to disclose when AI was used (and how)
• Insist that the actual code and design decisions remain human-authored
• Keep the usual standards: PRs must be well-scoped, clearly explained, and technically solid

This still rejects pure AI spam while avoiding the rigidity of a total ban. Not every useful fix needs to wait for someone who has unlimited free time to rewrite everything by hand.

[cloudskater]

Using AI to find problems is fine. Accepting fully AI-generated PRs, however, is a different matter, it tends to introduce subtle errors, dilute accountability, and make review harder. That said, I don’t think the only options are “ban AI completely” or “allow it with no rules.” A middle path seems more practical:

* Require contributors to disclose when AI was used (and how)

* Insist that the actual code and design decisions remain human-authored

* Keep the usual standards: PRs must be well-scoped, clearly explained, and technically solid

This still rejects pure AI spam while avoiding the rigidity of a total ban. Not every useful fix needs to wait for someone who has unlimited free time to rewrite everything by hand.

A rigid ban is necessary because any use of an LLM further normalizes the practice. Many argue over LLM use and if they're justified for one purpose over another, but in terms of the ethical problems detailed above, the result is essentially the same.
This also informs my answer to the common counterpoint: "Won't banning AI just make those who use it, hide it?". Putting aside the deterrent that a strict anti-LLM policy is on it's face, said people can be (pardon the pun) sloppy, and leave behind fairly obvious evidence. Action can then more easily be taken against these violations due to them indisputably being defined as such. Leaving an opening just allows room for denial and the like.
I guess it all comes down to what @MixusMinimax said: "A policy [...] should be written based on what is right, not what is enforceable. How or if to enforce it, is a completely separate discussion." Kinda like any rule, anywhere. It's never 100% enforceable, but you don't loosen your definition of "wrong" based on your ability to stop the "wrong" from happening, even if that ability is low, which I don't believe it is in this case.