If you draw an element (or a collection of elements) from your own data, there are three potential scenarios:
If you're dealing with markers, polylines, polygons. circles, rectangles or features, you don't send any data to Google. All those elements reside in their own layer, overlayed on top of the map.
If you need to draw a KML Layer, Google will load and parse the contents of the url parameter. That URL must be accesible for Google and won't work with urls on an intranet or behind a login wall. So yes, in that case they will have the contents of that KML in their servers.
If you want to use Fusion Tables Layer, the actual Fusion Table, residing in your google drive, must be publicly available.
For the second and third scenarios, you don't get to manage the analytic data in your browser context. Instead you provide Google with means to access that data, parse it and serve it to you in the form of pre rendered tiles with your KML or Fusiontable embedded on them. The illusion of them being a separate object is given by the custom click listeners they provide.
I wouldn't worry about Google sniffing whatever information you're drawing. They already know everything and they do not care.
Regarding your use case, the internal/intranet is only allowed during development phase. You might get your site or IP banned for violating the TOS. Situations like yours have spanned several startups building their own mapping solutions or porting to open source tools.
TL/DR: use the browser console to inspect requests made whenever you add data.