The policy is specifically about third party AI/LLMs. I assume a locally running LLM would be okay as long as it was not trained by any material whatsoever external to the company. That is, we could only use our own IP to train it.