THE CRUNCH
KDE Plasma has begun drafting community guidelines for Large Language Model usage, aiming to balance AI assistance with human oversight. The proposed rules, submitted by project lead Nate Graham, enforce a 'human in the loop' principle to prevent developers from acting as 'meat proxies' for AI tools. The draft explicitly forbids AI-generated text for merge requests, commit messages, and replies, while also requiring
The guidelines prioritise code quality and maintainers' time by discouraging 'vibe-coded' changes that require significant cleanup. Developers are expected to take full responsibility for AI-generated code, with the draft explicitly forbidding adding 'Assisted-by' tags to commits as a way to excuse low-quality work. The project is currently seeking community feedback on the draft, which is expected to be finalised and added to KDE documentation in the coming days.
The stance aligns KDE with the Linux Kernel project, which permits AI-assisted code only if a human demonstrates understanding through a proper merge request write-up. This approach contrasts with other open-source communities that have taken more extreme positions, such as GIMP's total ban on AI-generated code. The guidelines also address the issue of AI-generated text, arguing that 'what you are willing to write is generally the upper limit of what others are willing to read.'