-->
[PATCH] docs/devel: relax policy on AI-generated contributions
qemu-devel
[Top][All Lists]
Advanced
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[PATCH] docs/devel: relax policy on AI-generated contributions
From :
Paolo Bonzini
Subject :
[PATCH] docs/devel: relax policy on AI-generated contributions
Date :
Thu, 28 May 2026 09:34:12 +0200
Until now QEMU's code provenance policy declined any contribution<br>believed to include or derive from AI-generated content. A blanket ban<br>was easy to maintain while LLM output was rarely usable on its own, but<br>as the tools improved an absolute prohibition has become harder to<br>justify.
The concern that motivated the policy is unchanged, and it is worth<br>stating precisely: the DCO is about whether the submitter has the legal<br>right to contribute the code, not about "creative expression". The<br>copyright and license status of LLM output remains unsettled, so that<br>question is still open. What has shifted is the balance of risk:
- projects accepting AI-assisted content have not run into serious<br>legal trouble so far, which suggests the probability of the risk<br>materializing is not high;
- other organizations, such as Red Hat[1], have assessed the risk as<br>acceptable -- though a community of individual developers does not<br>have the legal backing of a company, and even an unfounded dispute<br>would be a long-lasting distraction from work on QEMU.
Revise the policy to permit AI assistance where the ramifications of<br>copyright violations are at least easy to revert and unlikely to spread:<br>tests, documentation, mechanical changes, and small bug fixes. Core code<br>that other things depend on, and that cannot simply be thrown away once<br>a problem is noticed long after the fact, stays off-limits without prior<br>agreement from a maintainer.
Related to this, and already visible in the incredible uptick in<br>security requirements, is the question of maintainer burnout and the<br>shift in effort from the author to the reviewer of the code. AI lowers<br>the cost of producing a patch but does nothing to lower the cost of<br>understanding and reviewing one; if anything it raises it, since a<br>reviewer can no longer assume that the submitter has reasoned through<br>every line. The limits above work just as much to keep the volume of<br>review work sustainable.
Furthermore, introduce "AI-used-for:" as a trailer to record where AI<br>was used, and include other suggestions that help reviewers judge<br>the result. The standard is slightly different from the more usual<br>"Assisted-by", which doubles as a check that the author has read the<br>policy.
In any case, use of AI does not relax any other contribution requirement:<br>authors still comply with the DCO and take responsibility for the whole<br>patch via Signed-off-by.
[Commit message largely based on<br>ahXbxzB4C_lr6b0N@redhat.com/">https://lore.kernel.org/qemu-devel/ahXbxzB4C_lr6b0N@redhat.com/, by<br>Kevin Wolf. - Paolo]
[1]<br>https://www.redhat.com/en/blog/ai-assisted-development-and-open-source-navigating-legal-issues<br>Cc: Alex Bennée<br>Cc: Alistair Francis<br>Cc: Daniel P. Berrangé<br>Cc: Kevin Wolf<br>Cc: Michael S. Tsirkin<br>Cc: Peter Maydell<br>Cc: Warner Losh<br>Link:<br>20260524083329-mutt-send-email-mst@kernel.org/T/">https://lore.kernel.org/qemu-devel/20260524083329-mutt-send-email-mst@kernel.org/T/<br>Signed-off-by: Paolo Bonzini<br>docs/devel/code-provenance.rst | 123 ++++++++++++++++++++-------------<br>1 file changed, 75 insertions(+), 48 deletions(-)
diff --git a/docs/devel/code-provenance.rst b/docs/devel/code-provenance.rst<br>index 65b8f232a08..84f9f4a70fb 100644<br>--- a/docs/devel/code-provenance.rst<br>+++ b/docs/devel/code-provenance.rst<br>@@ -1,7 +1,7 @@<br>.. _code-provenance:
-Code provenance<br>+Code provenance and AI usage<br>+============================
Certifying patch submissions<br>~~~~~~~~~~~~~~~~~~~~~~~~~~~~<br>@@ -288,62 +288,89 @@ content generators below.<br>Use of AI-generated content<br>~~~~~~~~~~~~~~~~~~~~~~~~~~~
-TL;DR:<br>+**Please read the below policy before using AI to contribute code or<br>+documentation to QEMU. This applies to ChatGPT, Claude, Copilot,<br>+Llama, and similar tools.**
- **Current QEMU project policy is to DECLINE any contributions which are<br>- believed to include or derive from AI generated content. This includes<br>- ChatGPT, Claude, Copilot, Llama and similar tools.**<br>+The increasing prevalence of AI-assisted software development,<br>+and especially the use of content generated by `Large Language Models<br>+https://en.wikipedia.org/wiki/Large_language_model>`__ (LLMs),<br>+poses a number of difficult questions.
- **This policy does not apply to other uses of AI, such as researching APIs<br>- or algorithms, static analysis, or debugging, provided their output is not<br>- included in contributions.**<br>+Risks to open source projects include maintainer burnout from an<br>+increased number of contributions, as well as the risk to the project<br>+from unintentional inclusion of copyrighted material in the LLM's output.<br>+In order to mitigate these risks, the QEMU project...