Right, of course. I'm talking about decrypting the past messages. Obviously they can modify their software to capture your decryption password, my question is

Right, of course. I'm talking about decrypting the past messages. Obviously they can modify their software to capture your decryption password, my question is more out of curiosity than it being a serious advantage against a state actor targeting you specifically. I am fairly sure that protonmail cannot do this without modifying the software to target me by capturing the decryption password. Of course they can always capture plaintext messages as they come in, we can only assume they don't keep records of that. If that's true, and we only have their word for that, it would make any requests forward focused, like a wire tap rather than a search of old bank records. They can't necessarily access old emails without explicit effort. It's not some insurmountable barrier, and I don't mean to suggest it is. It's trivial to think of at least three ways to work around it, assuming you are still logging in. But it is a difference in design.