What it takes to build and secure private AI infrastructure — from defining what “private AI” means, through hardware, quantization and the software stack, to threat modeling the working LLM environment and deciding what should run locally versus in the cloud.