The US Post Office takes digital images of most letters simply as a part of the automated sorting process. If the mail’s pre-sorted with bar codes, it doesn’t need to do much with them; if it’s typed it does easier OCR, and if it’s hand-written it does much harder OCR. 20 years ago, OCRing hand-written zip codes was cutting edge, but Moore’s Law is your friend for technologies like that.
Has the FBI or NSA gotten them to keep all those images? Doing a mail cover of just about everybody is easier today than doing mail covers of Usual Suspects was 30 years ago. Getting laws passed or court orders issued to let them do it has also gotten much easier. Keeping all the images isn’t that hard, and storing all the text when that’s available is trivial. And if Moore’s Law continues to cut storage costs in half every year or two, for roughly double your initial storage budget, you can keep it all forever.
