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.
signature.asc
Description: PGP signature
