This is very difficult to think about as someone who's very pro-individual liberty and privacy, but our current way of thinking may become obsolete if it won't be that difficult for a single, ordinary civilian to kill millions if they become motivated to do so. "Give me liberty, or give me death" will make a lot less sense if liberty implies almost certain death. We should start thinking about how a surveillance state could be created and maintained in the most ethical way possible. For example, automated collection and detection without human analysts invading privacy until something is already flagged by the system (to prevent "LOVEINT"-type abuses).
More people need to be exposed to Bostrom's ideas. I hope he does more podcasts.