The GNU Compiler Collection will reject copyright-significant contributions derived from LLM output, with limited exceptions for trivial changes and test cases.
Sad to see that it’s accepted outside of the test cases at all. But good to see they’re severely limiting it, at least.
Why do I find even minor LLM changes sad?
Because it moves the goalpost about what is okay to copy. If you brought a one-line change to a FOSS project in the past that you took out of the leaked MS Windows source code, you would have been scolded for risking such an explosive origin for such low gain. Nowadays with LLMs people seem to be trying to make that the new normal. I don’t think that’s a good path to take for the ecosystem.
And because no project should have to think about what’s “copyright significant”. If you reach that state, I feel like you should perhaps just reject it and have somebody rewrite it cleanly.
the problem generally lies with doubt about the copyright relevance. in many projects luttle contributions (e.g. 2 lines) can make a difference but are not really creative work
I’m not a lawyer but I’m really unsure if there’s even some safe guess like two lines. E.g. if you look at lyrics, I think people have been sued less - not that I would know for sure, though, and no idea if that allows any conclusions for code…
Well, depends on what the line is, right? Like, you’d have a hard time copyrighting
int i = 0;
because millions of people write that line of code.
But if it’s something unique and, say, identifiably Microsoft’s code, then yes, you probably would not have legal grounds to stand on.
In the context of the OP, of course, if you’re using an LLM just to write int i = 0; then you might be stupid. But possibly if you’re using LLMs to do tedious tasks like writing unit tests then you might be able to argue that they’re generic enough to not infringe copyright even if the legal situation is resolved in favour of LLM output belonging to the IP holders of the training data. It’s just a legal grey area, which is presumably why so many projects are concerned about merging LLM output.
Sad to see that it’s accepted outside of the test cases at all. But good to see they’re severely limiting it, at least.
Why do I find even minor LLM changes sad?
Because it moves the goalpost about what is okay to copy. If you brought a one-line change to a FOSS project in the past that you took out of the leaked MS Windows source code, you would have been scolded for risking such an explosive origin for such low gain. Nowadays with LLMs people seem to be trying to make that the new normal. I don’t think that’s a good path to take for the ecosystem.
And because no project should have to think about what’s “copyright significant”. If you reach that state, I feel like you should perhaps just reject it and have somebody rewrite it cleanly.
i agree
the problem generally lies with doubt about the copyright relevance. in many projects luttle contributions (e.g. 2 lines) can make a difference but are not really creative work
I’m not a lawyer but I’m really unsure if there’s even some safe guess like two lines. E.g. if you look at lyrics, I think people have been sued less - not that I would know for sure, though, and no idea if that allows any conclusions for code…
Well, depends on what the line is, right? Like, you’d have a hard time copyrighting
int i = 0;because millions of people write that line of code.
But if it’s something unique and, say, identifiably Microsoft’s code, then yes, you probably would not have legal grounds to stand on.
In the context of the OP, of course, if you’re using an LLM just to write
int i = 0;then you might be stupid. But possibly if you’re using LLMs to do tedious tasks like writing unit tests then you might be able to argue that they’re generic enough to not infringe copyright even if the legal situation is resolved in favour of LLM output belonging to the IP holders of the training data. It’s just a legal grey area, which is presumably why so many projects are concerned about merging LLM output.