To get code into this project, you must agree to the terms of a few documents:
Mostly, these documents assert that you won't be a jerk, and you won't submit code you don't have rights to.
And, of course, you must agree to publish your contributions as Open Source under the same license as this project's source code, or as Public Domain with no copyright.
Please sign your commits to certify the DCO:
git commit -s
This applies to AI-assisted contributions as well. You remain responsible for review, testing, and having the right to submit the code.
See AI Contribution Policy for details.
AI-assisted code is welcome. Treat AI like any other tool: you are responsible for the result. The Linux Foundation's AI Policies are a useful reference; our practical expectations are:
-
Contributors should ensure that the terms and conditions of the generative AI tool do not place any contractual restrictions on how the tool’s output can be used that are inconsistent with the project’s open source software license, the project’s intellectual property policies, or the Open Source Definition.
-
If a contribution is later found to include third-party copyrighted material, the contributor must help identify the source and provide any required notice and attribution, or help replace the code with a clean alternative. If you already know an AI output is derived from a specific source, disclose it and include the required license/attribution at submission time.
Additionally:
- You must understand and test the code you submit.
- If AI assistance was substantial, disclosure in the PR or commit message is strongly recommended.
- Do not submit AI output you did not review, test, or have rights to.
The US Copyright Office has some pretty relevant findings on the topic:
-
"Copyright does not extend to purely AI-generated material, or material where there is insufficient human control over the expressive elements."
-
"Whether human contributions to AI-generated outputs are sufficient to constitute authorship must be analyzed on a case-by-case basis."
-
"Based on the functioning of current generally available technology, prompts do not alone provide sufficient control."
This means it is complicated to determine if contributors have sufficient "authorship" rights to be able to claim copyright over the code being submitted.
GPL/AGPL projects can include public domain code, but those parts of the code remain public domain and are not subject to the terms of the project license. There is no clear and practical way to denote which parts of the code are copyrighted vs public domain though, so that generally needs to be addressed on a case by case basis if/when any contention occurs.
The analyses I've found from lawyers and prominent Open-Source companies mostly concluded that it's complicated and clear rules don't exist yet, so their recommendation is to be very careful with AI-generated content (like avoid things which look like they may have been copied instead of original), and otherwise assume it's public domain unless clear evidence of authorship is provided (like by documenting the author's creative process and how they have substantially transformed the AI's output).
So in this project, AI-generated code will be subject to the following guidelines:
-
If it looks like it might be copyrighted code from somewhere else, reject it or rewrite it. Non-trivial effort may be required to look for original sources, so things may stall out at this step.
-
If no clear evidence was provided that a human has done a legally meaningful amount of work to transform the AI's output, assume the code is not copyrightable. Strip any copyright notices from the submitted code, and treat it as public domain. If the patches modify existing source files, the pre-patch copyright headers on those files will remain unchanged. If the patches create entirely new files, they will have no copyright headers until/unless a human does a meaningful amount of modifications to the file. Copyright will be assigned to the person who did the modifications, not the submitter who AI-generated the initial version.
-
If any questionable content exists in the commits being submitted, such as potentially plagiarized code or copyright headers added to uncopyrightable content, those commits will not be included in the project's history graph. Instead, patches will be applied manually in fresh commits, without a reference to the original commits. The submitter will generally be acknowledged in a commit message, but not listed in the commit headers, like 'Author:' or 'Committer:'.
-
AI-generated code will need additional testing and validation, to understand it and ensure it does not introduce any bizarre bugs. This will take additional time. In many cases, it may also need to be significantly modified or rewritten, to align with the project's style and goals. This will also take additional time.
-
Given the difficulty of detecting whether code was written by AI or by the person submitting patches, it will be important to communicate about the origin of the code. Pull requests and patches should clearly communicate the code's origin and authorship.
If a maintainer treats code you wrote as if it was AI-generated, please contact them to resolve the misunderstanding.
If you believe your copyrighted code has been included in violation of your rights, please contact the project maintainer to resolve the situation. We will make best-effort attempts to prevent any code from being included without permission, but it's difficult and complicated to determine original authorship, and mistakes may be made occasionally.