Nowadays even consumer-level hardware can run some decent local LLMs, completely offline.

You might want to browse /r/LocalLLaMA/ if "security" is an issue for you.