> My company is pretty paranoid about leaking IP

Sorry but we have extremely different definitions of "paranoid". You are willingly sending your proprietary code to a third party.

Even if they respect their promise of no training, they still have your code, the IP is already leaked to them. And in case of a breach of their systems it would leak globally.

At my company we are not allowed to let LLMs touch the core code directly. And I don't think it's particularly "paranoid", we are just following the same general behavior from before AI.

There is such a wide range of response to this new paradigm, it's very interesting.