Hi Andreas, Andreas Enge <[email protected]> writes:
> Am Mon, Feb 23, 2026 at 02:35:48PM +0900 schrieb Maxim Cournoyer: >> I think it's unavoidable that this kind of LLM usage happens in the >> community (which mainstream search engine doesn't show some LLM-produced >> summary these days?), and I think a good thing we can do is ask from our >> contributors to be transparent about it, by adding a disclaimer when >> they've used an LLM to author their changes. It could be just a box to >> check in the PR template, or some git trailer, or both. > > While this sounds like a reasonable thing to do, ironically it would > stop me from committing the pull request due to unclear licenses of LLM > generated code. As Anderson Torres pointed out, it can be seen as a > derivative work of all the data the LLM has been trained with (which > means, among others, all source code available on the Internet). If > interpreted like this, the result becomes non distributable. (Admittedly > this is not the only possible interpretation; for instance recounting > a copyrighted story in one's own words is not a violation of copyright > since only the individual "artistic expression" is copyrightable, the > general idea is not. So maybe this is what LLMs do. And then since they > are not natural persons, the outcome is not copyrightable and thus in > the public domain. Then whether we want to "pollute" our GPL code with > public domain code is yet another question.) There's also the argument that the creative part in the process of creating the LLM generated code was the prompt a human typed and refined to get the result they needed, plus the likely modifications needed to get the result properly working, would make the end result likely copyrightable by the author that used the help of an LLM to produce it. But the suggestion to have a "Got help from an LLM to author this" checkbox in our PR would serve a more practical purpose to me: give me a hint that I should pay extra attention to the code being reviewed as it's likely to be of a lesser quality/contain dubious things. > So if someone ticks the box "this code is not redistributable", I will > not redistribute it... If someone hides the fact that their code is not > redistributable, I may apply it since we assume that by making a pull > request, contributors take the stance that the code is under their > copyright and they take responsibility for it. We could start using 'Signed-of-by' git trailers the way the project that created them uses them, that is, as some kind of legally binding (I assume, not being a lawyer) confirmation from the author that they believe the copyright on the changes submitted is owned by them and clear of licensing issues. -- Thanks, Maxim
