I suspect my question is dumb, but still - after 30 minutes of active search I haven't found an answer, so here it goes.
We know that Trezor's architecture is such that it keeps private keys inside the device and never exposes them (that's what we read in the FAQ:
http://doc.satoshilabs.com/trezor-faq/ )
We also know that whenever new Bitcoin address is generated, there is a pair of keys - public and private associated with each address, and the only way to spend balance from the given address is via its private key.
So in my understanding, if the wallet is guarded by Trezor, then each new address generated for such wallet should be generated by Trezor (otherwize private keys would be exposed).
However, my experience with Trezor proves that it is not the case.
After I set up my account and wallet at
https://mytrezor.com/ with the Trezor device, I detached the device and had no problem generating as many new addresses for my wallet as I like (with no Trezor attached). Consequently, private keys for these new addresses could NOT have been stored in Trezor.
So my question is - where are the private keys for the generated addresses are stored, and if not in the device - then what's the purpose of the device? What other private keys does it keep and how does it help protect the balance in my Bitcoin addresses, if their private keys are not in Trezor?
Maybe I should check Trezor's source for an answer, but I figured I post here first...