Richard Stallman's vision involved an exploitation of copyright to ensure that software would become free and remain free, and get even freer as we went along. The GPL and its derivatives have been, for decades, a really potent weapon to wield against closed-source, corporate interests, and proprietary software. Especially in the form of GNU/Linux: the kernel and its major distros are ultimate bastions of GPL software.
Creative Commons relies on copyright to work. BSD, MIT, Apache, all the F/OSS source licenses, and permissive licenses: they all rely on copyright. So how are LLMs going to help here?
If LLM output remains uncopyrightable, then we're in trouble. It seems that only a minority of jurisdictions confer any rights to the "prompters" or "vibe coders" of genAI works. If any given contributor to an F/OSS project cannot own, copyright, or assign rights from their works, then the project cannot remain "free".
Oh, sure, any F/OSS can become public domain, and purely public domain works are indeed "free software" but they can't be GPL'd and they can't be subject to any additional "freedoms" such as requiring attributions. Public domain software can be subsumed into trade secrets and closed-source projects. What company will want to contribute public domain code to any public project that they've written with LLMs... they will simply go closed-source.
I think any F/OSS dev should be concerned, at this point, with LLM contamination of code bases. Many projects are wisely clamming up and prohibiting AI-generated code, to the extent possible, but that's sort of impossible, isn't it? Contributors will simply claim ownership and claim manual output, when it wasn't, and who will prove them wrong?