|
Getting your Trinity Audio player ready...
|
The KDE open source community, best known for developing the Plasma desktop environment used across many Linux distributions, found itself in the middle of a heated internal fight this week over how the project should handle contributions generated with the help of large language models. A discussion thread proposing formal guidelines for LLM use within KDE devolved so thoroughly into conflict that its own creator ended up locking it down and asking participants to cool off before trying again.
KDE Plasma developer Nate Graham, who also serves as CEO of Techpaladin Software, opened the thread on KDE Invent on September 19 under the title “Proposed KDE LLM guidelines,” aiming to draft and refine rules governing how AI-generated content should be created and handled across KDE’s projects. The core of Graham’s proposal centered on what he called a golden rule for LLM usage: contributors should not use AI tools to replace their own judgment, interpersonal communication or learning process, should avoid taking unsustainable shortcuts, and should not skip the process of growing as a developer. The underlying rationale, according to the draft, was that ignoring those principles tends to produce poor-quality work that ultimately becomes someone else’s problem to clean up later.
The proposal also addressed what should happen when contributions show obvious signs of careless or unreviewed LLM use, suggesting such submissions could reasonably be ignored or closed by maintainers rather than requiring extensive review, and it left open questions around disclosure, including whether contributors would need to state they used an LLM, identify which parts of a patch came from a model, and confirm whether they had actually verified the output before submitting it.
According to GamingOnLinux, which first spotted and reported on the discussion, the conversation started off relatively civil, with most early pushback focused on refining specific wording and charged terminology in the draft rather than objecting to the underlying premise of allowing some AI-assisted contributions at all. That tone shifted considerably as the thread progressed. Participants increasingly raised broader objections to LLM use within KDE as a whole, including concerns about the environmental impact of training and running AI models, which some argued could conflict with the goals of KDE Eco, the project’s existing sustainability initiative.
The debate eventually moved well beyond the policy’s specific wording into a wider ethical argument over whether AI-generated contributions belong in KDE’s projects at all. According to reporting from Linux Compatible, one contributor pushed for a mandatory “Assisted-by” disclosure line specifically to flag concerns about LLM output potentially reproducing content verbatim from other projects in ways that could later create uncopyrightable code, while another countered that a disclosure tag only has value if people actually read it. KDE developer David Edmundson warned the thread would be locked if the conversation stopped being constructive, and the situation escalated further when a user posted personal information about another commenter, prompting fellow developer Tobias Fella to shut the discussion down entirely, stating plainly that the space was meant for policy discussion rather than trolling. Graham separately asked everyone to take 24 hours to cool off while he prepared a revised draft.
The controversy did not stay contained to the original thread. A sub-community calling itself KDE for People emerged in its aftermath, organizing around a petition demanding a complete ban on generative AI across Plasma and its components. According to XDA Developers, the group’s position statement frames the proposed ban as rooted in a pro-people outlook toward community, ethics and respect for the decades of work that has gone into KDE’s software and projects. The petition’s scope extends well beyond code contributions specifically, calling for a ban covering AI-generated assets, merge request descriptions, issue text and even translation files, and it points to other open source projects, including Zig, elementaryOS and OBS Studio, that have already implemented outright bans on LLM submissions as precedent for the approach it’s advocating.
The campaign behind the petition has acknowledged its own limitations, noting that a ban of this kind cannot be perfectly enforced and that some AI-assisted contributions will inevitably slip through regardless of the policy in place. Organizers have framed the effort less as a guarantee of total compliance and more as a signal of the kind of community KDE wants to represent, while leaving open the possibility that any ban adopted would not necessarily need to be permanent. Reaction to the petition itself has been mixed rather than uniformly supportive, with the original discussion on Reddit leaning neutral to negative and the broader debate spilling out across the fediverse, where positions on both sides have become fairly entrenched.
KDE’s current draft policy, as it stood before the thread was locked, tracks most closely with the approach taken by the Linux kernel project, which recently began accepting LLM-assisted code submissions under a framework that is permissive in principle but conditioned on careful review and personal accountability from the contributor submitting the work. That puts KDE in a middle position within the broader open source ecosystem, which remains genuinely split on the question. Gentoo and Void Linux have taken firmly anti-AI stances rejecting LLM contributions outright, while Debian has adopted a more permissive policy allowing high-quality LLM-generated content specifically, illustrating just how unsettled this question remains across major open source projects even as AI coding tools become increasingly common in software development generally.
With Graham working on a revised policy draft and a second attempt at community discussion expected in the coming days, KDE’s path forward remains genuinely uncertain. Whether the project ultimately lands closer to the Linux kernel’s permissive-with-oversight model, adopts something closer to Debian’s quality-gated approach, or moves toward the outright ban being pushed by KDE for People will likely depend heavily on how the next round of discussion unfolds, and whether the community can separate disagreements over specific policy wording from the broader, more emotionally charged debate over AI’s role in open source software that the first thread ultimately became consumed by. Continuing coverage of how open source communities are navigating AI-assisted development is available on Business Tech. Additional detail on the original policy discussion is available through KDE’s official mailing list archive, and further reporting on the controversy can be found through GamingOnLinux’s coverage of the story.
