GNOME Circle has moved to protect the quality of applications and libraries that orbit the GNOME desktop. The committee behind the program updated its criteria this week to reject submissions that show clear signs of low-effort machine-generated output. The change comes as reviewers face a growing pile of applications, some lingering for years.
The new rule does not ban AI tools entirely. Developers may still turn to large language models for learning or small code completions. Yet they must stand behind every line. They need to explain their work. Submissions filled with unnecessary functions, mismatched styles, invented API calls or leftover prompt comments face outright rejection. The policy mirrors one already in force for GNOME Shell extensions.
“While it is not prohibited to use AI as a learning aid or a development tool (i.e. code completions), app developers should be able to justify and explain the code they submit, within reason. Submissions with large amounts of unnecessary code, inconsistent code style, imaginary API usage, comments serving as LLM prompts, or other indications of AI-generated output will be rejected,” reads the updated text in the GNOME Circle criteria.
Sophie Herold, writing on the official GNOME blog, laid out the committee’s thinking. “As a precaution, we have adopted the AI policy that already applies to GNOME Shell extensions,” she explained. “With the rise of low-quality machine-generated software, there is a risk that GNOME Circle could be overwhelmed by new submissions of this nature, further clogging the review queue.” (GNOME Blogs)
A quick survey of existing Circle maintainers delivered a clear picture. Sixty-two percent said they use no LLMs at all. Thirty-four percent turn to them for small questions or code snippets. Only three percent rely on them for larger sections. Not one maintainer reported that AI had replaced their own coding. The committee remains open to feedback on the approach.
The timing feels deliberate. GNOME Circle also announced it would stop accepting new submissions for now. The backlog has grown too large. Some projects waited years for review. “We will stop accepting new submissions for now. We will reopen submissions when we have made a considerable dent in our backlog,” Herold wrote. The pause gives reviewers breathing room to clear older entries first.
This decision lands amid broader moves across the open-source world. Flathub tightened its own rules around AI-generated material, a step that matters because most Circle apps land there for distribution. Earlier, GNOME’s extension review process had already drawn a firm boundary against AI slop. Reports from Phoronix and Linuxiac captured the shift within hours of the announcement.
GNOME Circle exists to champion independent software built for the GNOME platform. Approved projects gain visibility, advertising support, GNOME Foundation membership and other perks. The program helps smaller developers reach users who might otherwise never discover their work. Yet that openness created vulnerability. A flood of low-quality entries threatened to bury the good ones.
Reviewers already spend considerable time hunting for malicious code, runtime problems and broken dependencies. Machine-generated submissions add another layer. They often compile but behave strangely. They introduce bloat. They leave maintainers unable to fix bugs because the original author never fully understood the code.
The committee did not stop at policy. It outlined several other changes. Issues awaiting maintainer replies will now close automatically. Developers must reopen them when they respond. SDK update reminders will arrive months before end-of-life rather than weeks. The goal is to reduce last-minute scrambles that consume reviewer time.
Benefits for accepted projects continue to expand. Students can work on Circle apps through Google Summer of Code or Outreachy. Help pages can live on the official help.gnome.org domain. Projects receive dedicated tags on GNOME Discourse. All contributors qualify for foundation membership. The committee actively seeks more reviewers and invited interested parties to join the Matrix channel.
One bright spot emerged from the same update cycle. The Resources application earned acceptance into the GNOME Incubator. The tool, which offers detailed system monitoring, could eventually replace the aging System Monitor in a future GNOME release. Its inclusion shows the review process still rewards solid, human-led development.
Other projects made progress too. GNOME Maps gained the ability to download map areas for offline use. The RustConn application saw continued refinement. A fresh release of the classic Solitaire game arrived for desktop users. These items, highlighted in This Week in GNOME, remind observers that innovation continues even as gatekeepers tighten standards.
The move reflects a growing skepticism inside parts of the open-source community. Many developers view current AI coding tools as useful for boilerplate but risky for anything complex. They produce plausible but often incorrect results. In a volunteer-driven project like GNOME, where trust and long-term maintainability matter, that risk carries weight.
Critics may argue the policy could discourage new contributors who experiment with AI. Supporters counter that the requirement to understand and explain code simply demands basic competence. The survey results suggest most active maintainers already work that way. They treat AI as an assistant, not a replacement.
Whether the policy spreads further remains to be seen. The Linux kernel has adopted disclosure rules for AI-assisted patches. Some distributions have taken harder lines. QEMU recently signaled it might loosen restrictions for minor contributions. The conversation is active and likely to evolve.
For GNOME Circle the immediate priority is clearing the queue. Once that happens, new submissions will resume under stricter scrutiny. The committee hopes the combination of policy, process tweaks and additional reviewers will keep the program healthy. Quality, not quantity, now sits at the center of its decisions.
And that choice may define how the next wave of GNOME software looks. Human judgment, accountability and deep familiarity with the code base have regained explicit value. In an age of instant generation, GNOME Circle has chosen to slow down and insist on the real thing.
GNOME Circle Draws a Line Against AI-Generated Code first appeared on Web and IT News.