Run it in a VM. Don't use containers. Do you want to run a local model? If yes then you can put the model itself in a container for speed. If you want to use an external AI api, then this is all not worth it, as you will leak all that data to an external company. (And yes AIs do steal your data and code)