All I found on the matter is this article: http://www.scmagazineuk.com/symantec-re ... le/213609/
The gist of it seems to be that they keep a whitelist of known clean files, and that the dynamically add to this whitelist when one of their installs comes across a new file.
I would guess that they have the whitelist cached for offline use.
As far as sending the entire file, they don't. It's still signature based, so your personal information is still safe. (For now.)
But most of this is probably going to be slightly different for every AV vendor's implementation.