Human Emacs
Emacs as a free software project focuses on empowering its users<br>without compromising its ethical principles. As users of Emacs, we are<br>interested in ensuring that we can continue to use it and contribute to<br>it in good faith in the future.
Background
Living thru 2026 as a software enthusiast has been a constant stream<br>of betrayals; it seems like every week we're faced with the news that<br>another program developed by people we thought we could trust has<br>succumbed to the LLM onslaught. We have watched as even stalwarts like<br>Vim and rsync have aggressively adopted LLM-driven development and<br>alienated their core user base, leading to forks like vim-classic:
https://vim-classic.org/
The current leadership of the Emacs project has indicated that<br>LLM-based contributions will not currently be accepted to the project,<br>which we are glad to see. However, this stance is a temporary measure<br>that is only meant to hold us over until the GNU project finalizes their<br>policy on LLM-based contributions:
We are awaiting the decision by the GNU Project on these matters,<br>which will define the policy for all the GNU packages, and in the<br>meantime we don't accept LLM-generated code, as a precaution.
[...]
Such policy discussions and decisions are not for the Emacs project<br>to make, they are global for all the GNU packages, and so here is the<br>wrong place to look for them.
https://lists.gnu.org/archive/html/emacs-devel/2026-03/msg00425.html
We are concerned that this upcoming GNU policy may not hold Emacs to<br>the high ethical standard that GNU has historically held. Given the<br>overwhelming negative externalities of large language models, it's<br>concerning that we haven't seen clear direction around this yet, and we<br>are worried that when a policy does arrive it may not protect us from<br>these problems.
Intent
Therefore we are here to declare our intent to use and develop a<br>version of Emacs that is free from LLM-generated contributions<br>regardless of whatever policy GNU decides upon. We<br>invite them to weigh the costs and do the right thing, in which case we<br>will continue to support them. But if they are not willing to do so, we<br>will continue to use and develop an LLM-free version of Emacs anyway,<br>even if it is not GNU Emacs.
It is not without hesitation that we drop the F-bomb (fork) here.<br>Forks of Emacs have historically been very contentious and have not<br>always had a good deal of success. However, in the case that GNU Emacs<br>accepts LLM-generated contributions, we do not feel like we have much of<br>a choice; giving up using Emacs for many of us would feel tantamount to<br>giving up on computing altogether, or at least giving up on the parts of<br>computing that bring us joy.
We also note that even if the top level GNU policy fails to address<br>this problem, there is no reason Emacs maintainers cannot have<br>additional policies on top of GNU's; contrary to the claim in the<br>mailing list thread above there would not be any conflict in making the<br>temporary ban permanent.
Motivation
The Guix consensus document "Standing<br>up for human crafting" does a great job of describing the motivation<br>of why we require such a policy for a program that we rely on as much as<br>Emacs. (Though the fact that they did not have the foresight to place a<br>temporary ban the way Emacs did means that they felt forced to leave<br>some unfortunate exceptions in their policy.)
We feel the harms are clear enough that it's not necessary to<br>enumerate them afresh here; please refer to that document if you wish to<br>read it laid out in detail.
Not Under Discussion
We are not here to discuss whether LLMs are effective at what they<br>are claimed to be able to do; their effectiveness is not at all relevant<br>to the question of whether their use can be part of a principled<br>software movement dedicated to user empowerment.
We are not here to discuss "open weight" models; these are still<br>built on a foundation of companies destructively mining the web from<br>data centers that wreck communities. Such models cannot exist without<br>exploitation. When a model can be fully trained by end users using data<br>that was collected with consent, then we can talk about that, but right<br>now that is nothing but science fiction and speculation.
We are not here to discuss how bad-faith contributors can lie about<br>the provenance of their patches. This risk is not new; bad-faith<br>contributors have always been able to lie about the licensing and<br>copyright implications of a patch. It is enough to treat LLM-generated<br>patches the same as other forms of plagiarism.
Organizing
Discussion takes place on the mailing list and<br>on the #human-emacs channel on Libera Chat. You do not have to join the<br>mailing list to post to it.
Signatories
Phil Hagelberg (aka<br>technomancy)
Jon Levin (aka mhcat)
David Bremner
Christine Lemmer-Webber
Geoff Wozniak
Shae Erisson (aka<br>shapr)
Tushar (aka<br>tusharhero)
Brit Butler (aka kingcons)
Gergely Nagy (aka<br>algernon)
Anna Liberty
jamesnvc
Leon Henrik<br>Plickat
Spencer Williams...