They can't create them for old messages, but they could theoretically add them for messages going forward. So it could work like an old-fashioned wiretap, where

They can't create them for old messages, but they could theoretically add them for messages going forward. So it could work like an old-fashioned wiretap, where you get the subpoena, install a bug and start listening -- but can't go back in time and listen. How? iMessage encrypts and sends a copy of every message to every device registered to a recipient (iPads, iPhones, Macs, etc.), each of which has distinct public keys. It's actually a pretty ingenious aspect of iMessage, that it can support multiple devices without any private key exchange. But..... those recipient public keys are provided by Apple. So in theory Apple could add itself or law enforcement as another "device" -- let's call it "fbiPad" -- completely bypassing all that security. However the "good" news is whether Apple is actually doing that in a given instance could theoretically be detectable by analyzing network traffic, since it would result in an additional copy being sent. Even though it's encrypted you could probably tell by the amount of data.