The problem I have as an end user is that I want the infrastructure protecting me to be invisible. Let's return to this after the following paragraphs. I will make some pretty far-reaching conclusions.
I think we can all agree that if some totally below-the-radar crypto anarchist who happens to have a few million dollars from bitcoins figured out that they actually have enough access via the dark web to bribe a few Russian generals and long story short detonate a nuclear bomb a few miles outside New York City, just for shits and giggles, then they should be stopped at some point along the way. This will seem like a made-up example to you but I purposefully don't want to confuse the issue with practical examples. We can all agree that at some point this should be stopped.
A reasonable time to stop it might be if intelligence agencies get a literal screenshot from a darkweb chatroom (from a concerned participant, where the participant thinks they're really going too far) where this is being planned in exacting detail but more information is needed to be precise. (For example, suppose the source of the nuclear bomb were not Russia but not enough information was given to identify it. There are actually quite a few nuclear states and many of them are quite corrupt. A short list includes India, North Korea, Pakistan.)
I would think that this kind of actionable urgent intelligence should unlock whatever privacy safeguards are in place, but the issue is that if there is a correct "technical" solution (if cryptography works 'correctly' and is not broken, in an academic sense), then there is no technical possibility to unlock anything. If Tor, crypto currencies, and encryption "work" (in a binary, yes it works, or no, it's broken sense) then following the receipt of such a screenshot there is no technical means of any further step.
Here I'm going to be philosophical for a second. The future of technology is nearly infinite human power. You can already in the next few seconds initiate a crypto currency transfer to anyone anywhere in the world, who can receive it without any banking infrastructure or oversight.
The arc of technology has been personal human enablement. When individuals become nearly God-like and all-powerful, it is dangerous to be in a position where, like the Muslims reporting the madman banned from his U.K. mosque for radical insanity, the status quo is that if you report your friend to the authorities saying, "My online friend, God-like in his powers, is planning to murder a million people just for shits and giggles, and he's kind of insane. Unfortunately, I don't know where he is or what he's doing, but I'm pretty concerned. He has a lot of money from a few ponzi schemes he ran. It's pretty credible for the following specific reasons (screenshots, quotes, etc)." And the only response from the authorities is, "Thanks for all this. We don't know where he is either, in the grand scheme of things a million deaths isn't that much and if it happens we will look at preventing another such case."
That's a pretty silly response, isn't it? That the only possible response is, sorry, nothing can be done.
Okay, now I've laid out why there should probably be some infrastructure on the back-end.
What I don't like is that this translates to humans literally reading people's private correspondence, web searches, etc. It's not very good.
What is a good middle ground?
Can't the NSA make things that run locally, so that no human is reading your correspondence or web traffic, but as you start researching nuclear weapons and making plans on how to murder a million people, and start making those transactions, all this starts adding up and, to quote the Constitution, its tools can receive instructions "particularly describing the place to be searched, and things to be seized", so that after such a report, its perpetrator can be found, or at least enough information can be collected to stop it if it is actually taking place?
I think that all of us here could be okay with being stopped at some point between purchasing a hundred million dollars in anonymous currency, and detonating a nuclear bomb. It's sensible. That can be part of the social contract.
It's difficult. Nobody wants to live with a judge, jury, and executioner in their home looking at everything they are doing in case they break some law.
I am glad that I personally don't have to answer these questions. But we can all agree on the need for privacy (no human looks at what you're doing), and also on the reasonableness, as each individual online progresses toward infinite personal power, for protecting the rest of society from credible and immediate, specific threats.
I agree with cryptographers who think of cryptography as a tool that is either working or broken. (If it has a back door, it's 'broken').
Perhaps if tools included a certain portion that runs locally they could increase the extent to which the tools are not actually 'broken' (i.e. they are actually working, and actually not backdoored), while also increasing the safety every single person has from other individuals being able to plan or pay for their specific death anonymously, and with impunity.
I realize that my suggestions here are not specific enough to be actionable, they are not clear recommendations. But I don't even see these possibilities being discussed (at least publicly), so I wanted to at least move the conversation a bit in this direction.
EDIT:
---
I'm getting downvoted pretty heavily. Let me ask point-blank: are you okay with someone being able to spend two weeks on the dark-web researching how to make and detonate a bomb using totally innocent chemical purchases, and then your spouse, parents, relatives, or you, being an innocent victim of my exploding the results, or would you want that person to be stopped at some point after they started doing that? The future of information is that it is ubiquitous and easy to access
[I edited this paragraph edited from first to third person.]
Actually secure communications would mean that it is technically impossible to see if someone has started communicating with people at ISIS who have overseen and helped people explode themselves. I am not saying communication should be weak and insecure, but should I really practically be able to start doing that if I want?
This is not some kind of false example, either.
Also, for downvoters: I think it is easier for you to agree with the other half of my statement, that nobody should be looking at our web traffic and correspondence, and that it should be actually secure, and also actually private.