I agree. There's an epic productivity boost, but there's a valid concern about prompts leaking sensitive data to a competitor. Fortunately, I think we are fairly close to being able to run our own instances of highly capable LLM's without too much friction.