COSMIC Project Draws the Line on AI, Enforcing a Strict Ban on LLM-Generated Content in Pull Requests

Whether or not the broader tech community ultimately declares any specific year as the definitive "year of Linux," 2026 is shaping up to be a defining period when prominent open-source projects have definitively drawn lines in the sand regarding artificial intelligence. As maintainers across the ecosystem grapple with an influx of automated tooling, the COSMIC desktop environment has emerged as the latest development team to make its stance unmistakably clear. The project has announced a hard pass on all Large Language Model (LLM) created content, extending the ban across every corner of code submissions, including comments, documentation, and descriptions.

COSMIC brings the ban down on LLM-generated content

As initially highlighted by industry observers, the COSMIC project has officially updated its guidelines and requirements for anyone looking to contribute code through a pull request. With the sole exception of the cosmic-flatpak repository—where upstream projects manage their own distinct manifests—any developer wishing to submit code to the broader COSMIC ecosystem must now agree to a strict checklist that explicitly rules out AI-generated submissions.

The policy shift comes as open-source maintainers face mounting administrative and review burdens driven by the accessibility of modern generative AI technologies. According to Jeremy Soller of System76, the firm driving the development of COSMIC and Pop!_OS, this decisive rule was enacted primarily to alleviate the severe strain that LLM-submitted code and documentation place on project maintainers.

Soller noted that while artificial intelligence tools have successfully lowered the barrier to entry, allowing numerous first-time contributors to attempt to make their mark on the COSMIC project, the reality on the ground has proven challenging. According to the development team, these automated or AI-assisted entries are frequently unplanned, poorly aligned with the project’s architecture, and ultimately rarely accepted into the codebase. Despite this, the sheer volume of these incoming submissions demands significant human oversight and review time, pulling core developers away from critical architectural work.

The wider open-source dilemma and contrasting industry approaches

COSMIC is far from alone in facing this complex operational challenge. Across the broader Linux and open-source landscape, maintainers have increasingly spoken out about the friction caused by automated contributions. In previous months, prominent Linux kernel contributors voiced concerns over becoming swamped with low-effort patches, while core distributions like Ubuntu had to adapt their update schedules simply to process the rising tide of AI-discovered bugs and automated reports.

However, the strategy adopted by System76 and the COSMIC team represents a notably firm divergence from some of their peers. While major upstream Linux projects and distributions have generally chosen a more permissive route—effectively giving the green light to AI-generated content provided it meets high standards of quality and correctness—System76 has chosen an outright ban on generated code and text within pull requests.

Even with this rigorous stance against generative AI output in submissions, the COSMIC project’s policy is nuanced. The team does not appear to be strictly opposed to the judicious use of AI tools for non-generative, analytical use cases—such as utilizing machine learning models locally to help developers analyze codebases or hunt down elusive bugs. The primary target of the ban remains the automated generation of pull request code, comments, and descriptions that flood maintainer queues with unverified or low-context material.

As open-source ecosystems continue to navigate the rapid evolution of artificial intelligence, the policy implementation by COSMIC serves as a significant bellwether for how desktop environments and independent projects may choose to protect their maintainer workflows and code quality moving forward.