I think we need to make a distinction between automated services and humans eavesdropping. I'd feel weird if someone was snooping in on my conversations and clicking my links, but on the other hand I very much appreciate the little bot that sits in my IRC channel and displays the title of any page linked. Both monitor the chat and access links, but I value one and feel weird about another. I don't think there's a way to truly make a distinction between the two though, and I think saying, "nothing is ever allowed to access your communications" removes the possibilities for a lot of added functionality (the link bot being just the base camp of the mountain of things that are possible). I think the better choice here is to ensure that the public has a way to communicate securely, and that our mental model of "trust usually, distrust as the exception" needs to move to "distrust usually, trust as the exception". This is similar to how sudo works in a way - we maintain a lower level of security usually for the convenience, then escalate only when needed.