System76, the company behind the Pop!_OS Linux distribution, has drawn a hard line against AI slop in its flagship desktop project: the COSMIC desktop environment will no longer accept pull requests that contain large language model-generated content.

As first reported by Linuxiac and picked up by Neowin this week, the COSMIC pull request template has been updated to require contributors to confirm that their submissions contain no LLM-generated material of any kind — including source code, code comments, and the pull request description itself. If that confirmation is missing, maintainers may close the pull request outright.

The move makes System76 one of the most prominent open source projects to switch from asking contributors to disclose AI assistance to banning it from contributed code altogether. For developers tracking how the AI industry is reshaping software engineering, it is one of the clearest signals yet that maintainers are reaching their limit. For more stories like this, see our latest AI developments coverage.

What the New Rules Require

Under the updated contribution policy, COSMIC contributors must now confirm several things when opening a pull request:

  • that the submission contains no LLM-generated code, comments, or descriptions
  • that they understand the changes they are submitting
  • that they can answer review feedback from maintainers
  • that they have tested their work
  • that the contribution is certified under the Developer Certificate of Origin, the standard attestation that the contributor is allowed to submit the code

The policy applies across the COSMIC codebases, the family of Rust-based repositories — from the cosmic-comp compositor to the cosmic-files file manager — that together make up the desktop environment shipping in Pop!_OS 24.04.

Why System76 Pulled the Plug

Jeremy Soller, System76's principal engineer, said the driving factor is the extra review burden that AI-generated submissions place on the COSMIC team, according to Linuxiac. Maintainers have seen a growing volume of submissions from first-time contributors relying on LLMs, and many of those pull requests were unplanned and rarely accepted.

The project had already been quietly filtering this traffic. In August, contributor Jacob Gkau noted that most large LLM-written pull requests were already being rejected, particularly where they raised copyright or maintainability concerns. Until now, the project's formal rule required contributors to disclose AI-generated code and demonstrate they understood their submissions. The new rule is stricter: LLM-generated content is not disclosed and reviewed — it is not accepted at all.

Review bandwidth is the scarcest resource in open source, and the economics explain the shift. A machine-generated pull request takes seconds to produce and minutes to reject, but hours to review properly — and every hour spent triaging a speculative AI contribution is an hour not spent on vetted work from trusted maintainers.

Not a Ban on AI Tools — a Ban on Unsolicited AI Contributions

The policy is carefully scoped. It does not forbid System76 engineers or contributors from using AI tools in their own workflows; it governs what lands in the project's repositories from outside. A contributor who uses an LLM to learn a Rust pattern and then writes and genuinely understands the patch themselves is not what the rule targets — the confirmation requirement is aimed at submissions where the code, comments, or description were generated wholesale by a model.

There is one carve-out. The cosmic-flatpak repository, where upstream projects manage their own Flatpak manifests, follows a lighter-touch process: the COSMIC team primarily checks those manifests for correct sandboxing before accepting them.

Open Source Draws a Line on AI Slop

The debate spilled into the open this week on Hacker News, where the story gathered more than 75 points and a long discussion thread. One commenter said their own pull request — a hand-polished fix for a password-handling issue in COSMIC's network applet, developed with Claude's help — was closed under the new policy. "I respect their decision and no hard feelings," the contributor wrote, while acknowledging the setback for someone trying to break into open source contribution.

Other voices in the thread were less diplomatic, describing unmaintainable "AI messes" in their own projects and praising System76 for acting. Some questioned whether the policy could hold in practice, noting that any Linux distribution is built from thousands of upstream packages whose provenance no maintainer can fully audit. Reports in recent weeks indicate that Redox OS, the Rust-based operating system project, has adopted a similar ban on AI-generated code, and maintainers across the ecosystem are watching to see whether formal no-AI policies spread.

What It Means for Pop!_OS Users and Developers

For users, nothing changes immediately — Pop!_OS and COSMIC continue to ship as normal, and the policy is about contribution hygiene, not features. For prospective contributors, the bar is now explicit: write it yourself, understand it, test it, and be ready to defend it in review.

The deeper question is whether other major projects follow. AI-assisted coding is now mainstream, and every maintainer-facing project must decide what to do with the resulting wave of machine-generated contributions. System76's answer — a template checkbox, a possible closure, and a Developer Certificate of Origin — offers one template. Whether it becomes the industry's default remains to be seen.

---

Stay Ahead of AI

Get the latest AI news, analysis, and breakthroughs — all in one place.

Read more AI news →