Given how the models were trained for coding, every single code base that uses any code generated by an LLM should be required to be open sourced, or at least source available.

I'm not saying there is currently a legal president to enforce this, I'm saying ethically it make sense.