LLM Code Output Jeopardizes Copyleft Licensing and Open-Source Ownership

RE: https://wandering.shop/@xgranade/116430602464418667 I wrote this by way of disagreement with the argument that the ethical issues with using LLMs are mitigated when writing code, because code is not art. I disagree about code and art in at least some cases, but that's not really the point. Yes, art means something, and LLMs are devoid of meaning by construction — it's worth discussing that when considering the use of AI-generated assets in place of art. But that's not really an *ethical* argument. The ethical argument is that code is labor.

(The ethical argument is also that LLMs are based on eugenicist thought, that LLMs profit fascists, that LLMs deprive us of the critical backstop of saying no to fascists, and that LLMs take unconsionable amounts of energy. But the pro-labor argument against LLMs is the one under immediate discussion in the context at hand, so that's the one I focused on.)

Interesting blog post by @yoasif about #GenAI, #LLM-generated code, #copyright and #copyleft licenses (e.g. #GPL): https://www.quippd.com/writing/2026/04/08/ai-code-is-hollowing-out-open-source-and-maintainers-are-looking-the-other-way.html TLDR: LLM-generated code is not protected by copyright - because it was not created by a human. Therefore, it is essentially public-domain. Now, if you start accepting LLM-generated code into a project under a copyleft license, eventually it should also become public domain (e.g. imagine an open-source project where after a few years of LLM contributions 90% of the code has been generated by an LLM). If that project essentially doesn't fall under copyright anymore, it becomes public domain, so the author(s) can't put a license on it anymore - and hence the #copyleft license doesn't apply anymore. Everyone is free to use it, even if the original contributors intended for the project only to be usable in a #copyleft way. The blog post touches on some interesting, but quite philosophical questions, mainly: - Where do we draw the line between code written by a human with the help of a machine and code written by an LLM? - How much editing of LLM-generated code is necessary until the code can be considered to have been written by a human?