I mean, if you ignore the fact there would be no LLM's without wholesale scraping of the corpus of all software ever written.
LLM's are the least ethically sourced pieces of technology I've ever seen. That they have businesses built that haven't been sued out of existence for not asking for permission to train first is positively mind boggling.
> all software ever written
LLMs aren't usually trained on large proprietary codebases like the ones from Google, Microsoft or Apple?