Dear team members,

This message announces the start of the deliberation period on GCD 008,
entitled:

  Commitments and policy on the use of “generative artificial intelligence”

The final version of the document is attached and can also be viewed here:

  
https://codeberg.org/guix/guix-consensus-documents/src/commit/a24520c4147ffd67bb696c71f15ed4fb8521a791/008-genai.md

The history of discussions is available at the following places:

  https://codeberg.org/guix/guix-consensus-documents/pulls/13
  https://lists.gnu.org/archive/html/guix-devel/ (May, June, and July)

Excerpt from GCD 001, which specifies the process:

  Once the final version is published, team members have 14 days to send one
  of the following replies on the patch-tracking entry of the GCD:

  - “I support”, meaning that one supports the proposal;
  - “I accept”, meaning that one consents to the implementation of the
    proposal;
  - “I disapprove”, meaning that one opposes the implementation of the
    proposal.  A team member sending this reply should have made constructive
    comments during the discussion period.

To avoid ambiguity, replies from team members will be accepted until:

  Monday, July 27th, 2026, 23:59 AoE (anywhere on earth)

Replies are accepted both on the Codeberg pull request given above and
on the [email protected] mailing list.

Many thanks to everyone who participated so far!

Ludo’.

title: Commitments and policy on the use of “generative artificial intelligence”
id: 008
status: submitted
discussion: https://codeberg.org/guix/guix-consensus-documents/pulls/13
authors: Ludovic Courtès
sponsors: Janneke Nieuwenhuizen, Noé Lopez, Sharlatan Hellseher, Cayetano 
Santos, Tobias Geerinckx-Rice
date: 2026-05-17
draft-date: 2026-05-15
discussion-date: 2026-05-17
deliberation-date: 2026-07-13
SPDX-License-Identifier: CC-BY-SA-4.0 OR GFDL-1.3-no-invariants-or-later
---

# Summary

Guix is a free software project; it is made by people for people, with
knowledge sharing and empowerment as some of its core values.  This
document proposes commitments and a policy to support this vision.

# Motivation

Since large language models (LLMs) and “generative artificial
intelligence” (thereafter “genAI”) became available for the general
public a couple of years ago, they have been having a significant impact
on software development.

In the Guix community, genAI is viewed by some as an opportunity, in
particular as a way to help maintain and grow the package collection, at
a time where package updates are labor-intensive and often lagging
behind.  People reported experience using genAI for packaging-related
tasks both with commercial services and with locally-running pre-trained
models (Apertus, Qwen).

Others in Guix and in the broader free software community point out
immediate and longer-term risks associated with genAI that can be
broadly categorized as follows:

  - **Legal issues.** The copyright status of LLM output is unsettled
    worldwide, including [in the
    
EU](https://www.europarl.europa.eu/RegData/etudes/STUD/2025/774095/IUST_STU(2025)774095_EN.pdf)
    and [in the USA](https://www.congress.gov/crs-product/LSB10922).
    GenAI can also [regurgitate training
    data](https://arxiv.org/abs/2601.02671) thereby exposing their users
    to a risk of copyright infringement.
  - **Ethical issues.**  The leading commercial genAI services run
    non-free software and are to some extent “service as a software
    substitute” (SaaSS).  Other ethical concerns include the [unlawful
    extraction of
    content](https://www.amnesty.org/en/documents/pol40/0996/2026/en/)
    for training purposes for most models (also putting infrastructure
    [such as that of
    Guix](https://codeberg.org/guix/maintenance/issues/10) under high
    stress), and the [material, water, and
    
energy](https://www.technologyreview.com/2025/05/20/1116327/ai-energy-usage-climate-footprint-big-tech/)
    consumption of the genAI industry.
  - **Social issues.**  In many free software projects, genAI has led to
    an influx of low-effort contributions [putting an extra toll on
    maintainers and
    
reviewers](https://github.com/llvm/llvm-project/blob/main/llvm/docs/AIToolPolicy.md#extractive-contributions),
    leading those projects to regulate genAI use.  Other concerns
    include the risk of [growing cognitive
    debt](https://simonwillison.net/2026/Feb/15/cognitive-debt/) in
    projects where genAI plays an important role in code (co-)authoring,
    the risk of [weakening social ties as people look for guidance from
    chatbots instead of from other users and
    
contributors](https://blog.stdlib.io/ai-and-the-invisible-newcomer-in-open-source/),
    the risk of [competition for financial resources between humans and
    genAI vendors](https://arxiv.org/pdf/2601.15494), and the risk of
    exclusion of people who cannot afford genAI or are prevented from
    using it for geopolitical reasons.
  - **Technical issues.** Reproducibility is a core property of Guix but
    genAI, whose process is non-deterministic, puts the reproducibility
    of development workflows at risk.  Another concern is code rot:
    should genAI be widely used in Guix, existing automation tools not
    indispensable to LLM agents, such as `guix import`, `guix refresh`,
    and `guix style`, are at risk of becoming unmaintained, resulting in
    a loss of autonomy for the project and its contributors.

This proposal aims to address some of these concerns through the
adoption of commitments and a policy as precautionary measures to
safeguard our production from a legal standpoint as well as the social
fabric built over almost 15 years around the project.

# Detailed Design

In a context where genAI is already widely used and is often hard to
escape, we propose the adoption of commitments and of a policy on behalf of the 
project,
with the following goals:

  - Safeguarding the project’s main assets—its community and its code—by
    following the *precautionary principle* considering irreversible
    changes that unregulated genAI adoption may cause, socially and
    technically.
  - Questioning the reasons that make genAI feel necessary for people
        using Guix, and finding ways to fill the gap.
  - Strengthening support for craftspeople the project interacts
    with—translators, artists, developers, and so on.
  - Contributing to the public debate on these matters and creating ties
    with like-minded organizations and grassroots movements.

## Project Commitments

We propose the following project commitments:

  1. The project members (defined as maintainers, team members, and anyone with
     write access to a Guix repository or to Guix
     resources such as the build farm) **will not use
     genAI** to author code or packages, to interact with other
     participants (e.g., to explain code changes or to review code), to
     produce artwork, or any other artifact.
  2. The project will keep working to **provide people of all levels of
         experience with the resources to use Guix and to contribute to
         Guix** without feeling the need to resort to genAI:
         - by facilitating access to documentation and improving it to
           better suit the needs to people with different levels of
           experience—newcomers discovering free software, people with prior
           exposure to GNU/Linux, developers;
         - by developing and improving tools that make it easier to get
           started such as [Guix
           Packager](https://guix-hpc.gitlabpages.inria.fr/guix-packager/)
           and `guix import`;
         - by improving diagnostics to make our tools more approachable;
         - by providing communication channels anyone can use to look for
           help or to offer support, where all and everyone can feel safe to
           participate;
         - by improving tools for mechanical translation, refactoring, and
           updates such as `guix import`, `guix style`, and `guix refresh`,
           so that tedious packaging tasks can be automated in a way that is
           transparent, deterministic, and maintains the project’s
           sovereignty.
  3. We acknowledge that the project’s sustainability depends on
         automation for all the mechanical, labor-intensive tasks such as
         package updates.  We will keep **improving hackable tools and services 
to
         automate some of the package collection maintenance work**.

The Guix project will publicize these commitments, by referring to it in its
contribution guidelines, by making it easy to find, by presenting it in
blog posts, and through any other communication deemed appropriate.

## Policy for Contributions

We propose that contributions to Guix be subject to the following
policy:

  1. **GenAI use.** Contributors who are *not project members* (as
     defined above) MAY use genAI to assist in their authoring process
     for some contributions, subject to these restrictions:
     1. **GenAI use disclosure.**  Contributors using genAI to assist them
        in their authoring process MUST disclose its use and how they used
        it.  This includes cases where the contribution was produced by
        genAI or where the contribution derives from changes produced by
        genAI.
         2. **Contribution acceptance.**  Changes produced by genAI or deriving
        from genAI output that can be viewed as “creative”, and [could thus
        fall under
        copyright](https://en.wikipedia.org/wiki/Threshold_of_originality),
        MUST be rejected to avoid any risk of copyright infringement.
        Conversely, when reviewers consider a genAI-produced change
        “non-creative”, and thus not falling under copyright, the change
        MAY be accepted.  Examples of non-creative changes include
        mechanical conversions of package metadata from other repositories
        similar to those made by `guix import`, mechanical changes similar
        to those made by `guix refresh` or `guix style`, changes that
        merely follow suggestions made by `guix lint`, integration of
        upstream patches, changes of a package’s `#:configure-flags`, and
        similar package definition adjustments that are arguably below the
        [threshold of
        originality](https://en.wikipedia.org/wiki/Threshold_of_originality).
  2. **Interaction.** Contributors and project members MUST NOT use genAI for 
direct
     interaction with other participants.  Examples of prohibited uses include: 
letting
     an agent comment in pull requests, issues, or mailing list threads;
     copying text provided by genAI as answers or explanations in a
     discussion; more generally, removing the human from the loop.
  3. **Exploratory analysis.** Contributors and project members are free to use 
genAI
     as part of their exploratory process as long their final contribution
     respects the above rules.  For instance, use of genAI to identify
     the cause of a bug or the reason for a package build failure is
     permitted.

The Guix project will make this policy visible to contributors: as a
file in the main repository, by adding it to the “Contributing” chapter
of the manual, or in any other way deemed appropriate.

## Related Work

A number of free software projects have adopted a policy with respect to
genAI, revealing different sensibilities and choices.  Here are some of
those we looked at:

* Asahi Linux
  https://asahilinux.org/docs/project/policies/slop/
* Debian
  https://lwn.net/Articles/1020968/ (May 2025)
  https://lwn.net/Articles/1061544/ (March 2026)
* FreeBSD
  https://www.freebsd.org/status/report-2025-04-2025-06/
* Gentoo
  https://wiki.gentoo.org/wiki/Project:Council/AI_policy
* GCC (work in progress)
  https://gcc.gnu.org/wiki/working-group-ai-policy
* GNOME (individual packages)
  https://gitlab.gnome.org/GNOME/gnome-calendar/-/merge_requests/725
  
https://discourse.gnome.org/t/loupe-no-longer-allows-generative-ai-contributions/27327
  
https://gitlab.gnome.org/GNOME/libadwaita/-/blob/main/CONTRIBUTING.md#use-of-generative-ai
* GNU Binutils
  https://sourceware.org/binutils/wiki/LLM_Generated_Content
* GNU Gnulib
  https://lists.gnu.org/archive/html/bug-gnulib/2026-02/msg00064.html
* Linux
  
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/process/generated-content.rst
  https://lwn.net/Articles/1032612/
  https://docs.kernel.org/process/coding-assistants.html
* LLVM
  https://github.com/llvm/llvm-project/blob/main/llvm/docs/AIToolPolicy.md
* Mastodon
  https://github.com/mastodon/.github/blob/main/AI_POLICY.md
* NetBSD
  https://www.netbsd.org/developers/commit-guidelines.html
* NixOS
  https://github.com/NixOS/nixpkgs/pull/514587
  
https://github.com/NixOS/nixpkgs/blob/master/CONTRIBUTING.md#automationai-policy
* NLnet Labs
  https://nlnetlabs.nl/llm-policy/
* Pepper & Carrot (comics)
  
https://www.peppercarrot.com/en/documentation/409_Code_of_Conduct.html#use-of-generative-ai
* Servo
  
https://web.archive.org/web/20260311122512/https://book.servo.org/contributing/getting-started.html#ai-contributions
* Software Freedom Conservancy (recommendations)
  
https://sfconservancy.org/llm-gen-ai/llm-backed-generative-ai-recommendations.html
* Zig
  https://ziglang.org/code-of-conduct/#strict-no-llm-no-ai-policy

Software package lists:

* "No AI" list
  https://noai.starlightnet.work/list.html
* software projects that accept LLM contributions
  https://codeberg.org/small-hack/open-slopware

## Cost of Reverting

A new GCD could propose a different set of commitments, a different
policy, or abandoning both altogether; should consensus be found on such
a proposal, it would override the decisions of this GCD.

What would be costly to revert is the *lack* of any form of regulation
on genAI use in Guix.

# Drawbacks and Open Issues

This proposal has seen significant changes since the initial version
that was submitted; it is the result of discussions among people
covering a wide spectrum of opinions on the subject.  As such, it might
feel “too weak” to some and “too restrictive” to others.  Yet, it is
hopefully a good reflection of where we stand today as a community.

Attachment: signature.asc
Description: PGP signature

Reply via email to