The main reason that the infosec folks have paused the use of LLM's to generate code is copyright concerns: who "owns" the code that's generated, the LLM, or, the company?