COSMIC's Stance on LLM/AI Code Contributions?

[Commander-Zal]

Hi there,

There is a growing concern in the FLOSS community around accepting code contributions that were directly generated from an LLM/AI. As they are trained on code that is not license compatible with GPL, it is quite possible for them to recreate that incompatible code verbatim. If that code was accepted in a pull request to a GPL licensed project, it would unknowingly to both the submitter and the project itself introduce plagiarized, copyrighted code into the codebase, potentially opening itself up to legal liability.

This talk from FOSS Backstage Berlin does an excellent job of showing how LLM/AI Generated code can directly copy infringing code in its output: https://youtu.be/xvuiSgXfqc4?t=247

In Europe, a court ruled that LLM/AI output is able to be found as plagiarizing: https://www.twobirds.com/en/insights/2025/landmark-ruling-of-the-munich-regional-court-(gema-v-openai)-on-copyright-and-ai-training

According to an article on the subject, about 10% of LLM code output is likely to be plagiarized, copyrighted code: https://www.theatlantic.com/technology/2026/01/ai-memorization-research/685552/

In response to this threat, many FLOSS projects have decided to officially stop knowingly accepting AI generated code submissions, to reduce the risk of legal action in the first place, as well as to lessen their liability and give them a stronger defense if brought to court from an undisclosed AI contribution. The latest to do so is the GCC project, and previously the SDL library did the same.

What is COSMIC's official stance on AI generated code contributions submitted to COSMIC? (and I want to emphasize generated code contributions, as I think an LLM pointing out bugs, or asking it about things in general do not render the same sort of legal threat to FLOSS projects).

Thank you for your time and consideration!

[mmstick]

We have a policy in place that we use with GitHub pull request templates: https://github.com/pop-os/pop/blob/master/.github/PULL_REQUEST_TEMPLATE.md

[Commander-Zal]

Thank you for your response and info! Is there any possibility of that policy changing in the future?

[jacobgkau]

There's always the possibility of policies changing in the future, but nothing is currently planned.

If you browse our PRs, you'll see that our engineers usually already choose not to accept LLM-written PRs if they're very substantial (which would be the ones more likely to have copyright issues, as you're concerned with).

[actionless]

it's some sort of anti-ai spam campaign (ex. awesomeWM/awesome#4113 linuxmint/cinnamon#13914 qtile/qtile#6011 #3774)

[Commander-Zal]

it's some sort of anti-ai spam campaign

In light of Xfce recently accepting AI code, and then discovering many other DE's followed suit, I am now trying to find a desktop environment that shares these concerns. The only way to know a project's stance on these issues is to ask. Asking multiple projects at once does not, in my mind, make it spam.

You yourself appear to have very strong opinions on AI use, and accused me of being a "biological-supremacy extremist" in the AwesomeWM discussion thread for simply having concerns about the long-term health of FLOSS projects from potential copyright concerns, so you appear to be rather biased yourself.

And to clarify, I have no issue with a developer using a self-hosted model for bug checking a program, and if an open model existed that was able to ensure it was trained exclusively on code that didn't require attribution or incompatible licenses, I would even be okay with it being used for code generation in some capacity, but that model does not currently exist.

[Commander-Zal]

There's always the possibility of policies changing in the future, but nothing is currently planned.

If you browse our PRs, you'll see that our engineers usually already choose not to accept LLM-written PRs if they're very substantial (which would be the ones more likely to have copyright issues, as you're concerned with).

Thank you for your response :)