A spy who is the only person using an encrypted network isn’t hidden. He’s the most visible person on it.
That single, almost obvious problem is the reason Tor exists in something close to its current form. I’ve spent enough time around research on structure, safety, and control to recognize this pattern the moment I saw it: a system built to protect the few only worked because it had to protect everyone, including people its own designers would never have chosen.
That’s not a footnote to the story. It’s the whole story, and it’s the part that keeps getting cut.
The problem American intelligence actually had to solve
In the mid-1990s, researchers at the U.S. Naval Research Laboratory were trying to solve something narrower than “how do we protect privacy.” Their actual question was closer to: how do you stop a watching adversary from noticing that a message even exists, let alone reading it. Encryption scrambles content. It does nothing to hide the fact that a signal moved from a known government address to a particular server at a particular time. That pattern, who is talking to whom, is often more revealing than the message itself.
David Goldschlag, Michael Reed, and Paul Syverson built a system that routed traffic through layers of encrypted relays, so no single point on the network could see both the origin and the destination of a message. They named it onion routing. The Tor Project’s own account of its history is direct about the goal: connections that don’t reveal who is talking to whom, even to someone monitoring the network.
Why a network built only for spies defeats itself
Here’s the part I find genuinely strange, and the part almost nobody repeats when they tell this story. Solving the cryptography wasn’t the hard problem. The hard problem showed up after.
If the only people using this system were intelligence officers, the system would announce them. Anyone watching global traffic patterns could simply flag every connection to this obscure network and conclude, correctly, spy. A crowd of one isn’t cover. It’s a spotlight with extra steps.
I don’t think this gets treated with the weight it deserves. Anonymity isn’t something you build into a wire or an algorithm. It’s something that only exists in relation to other people doing the same ordinary thing you’re doing. Which means the researchers hadn’t actually solved their problem yet. They’d just moved it.
The alliance nobody would have predicted
So the people building a classified tool for the U.S. government went looking for cypherpunks, civil libertarians, and privacy advocates, the exact community most likely to distrust the U.S. government on principle, and invited them into the design process.
I keep coming back to this detail because it isn’t ideological generosity dressed up after the fact. According to a detailed account in the MIT Press Reader, the Navy couldn’t even run the infrastructure itself, because a network visibly tied to the U.S. military would only be trusted by people who already trusted the U.S. military. The tool needed to look independent because it needed to be independent. There was no version of this that worked as a purely internal government project.
The code went public in 2004, and the Electronic Frontier Foundation began funding Dingledine and Mathewson’s continued work soon after. In 2006, they formally founded the Tor Project as a nonprofit. A classified research question turned into free software maintained by the same nonprofit ecosystem that now defends ordinary people’s privacy.
Where I think the popular version of this story goes wrong
Most retellings of this history land in one of two places. Either it’s a conspiracy (“the government built the dark web to watch you”) or a redemption story (“a military project became a gift to the world”). I don’t buy either framing, because both skip the mechanism that actually explains what happened.
The dissidents, journalists, and everyday browsers who use Tor now aren’t lucky bystanders who happened to benefit from someone else’s tool. They are the tool. Structurally, they’re the reason it functions at all, even for the officer it was originally built for. Without a large, mixed, ordinary crowd, the system fails at its one original job. So this was never intelligence agencies generously sharing something. It was intelligence agencies discovering their invention was mathematically useless unless they gave it away.
That’s the detail I think almost nobody remembers, and once you see it, you can’t read the rest of the story the same way.
What this says about who actually gets protected, and why
I find this pattern uncomfortable, honestly, because of what it implies about protection more broadly. The dissident in an authoritarian country gets real anonymity partly because a government elsewhere needed cover traffic for its own operatives. The journalist protecting a source benefits from infrastructure that exists for reasons that have nothing to do with journalism. Nobody sat down and decided these groups deserved protection on their own merits. They inherited it as a side effect of someone else’s requirement.
That should unsettle anyone who assumes privacy is something you either have a right to or don’t, as a standalone fact. Here, it was never standalone. It was borrowed, contingent on a crowd holding steady. Shrink that crowd and the most vulnerable users, not the intelligence officers, are the ones left exposed.
The tension I don’t think resolves cleanly
I’ll be honest about where this leaves me. It’s tempting to pick a side, either the dark web is a surveillance trap or a privacy triumph, and defend it. I don’t think either holds up.
The exact design that lets a source talk to a journalist without being identified is the same design that lets an intelligence operation move undetected, and the same one that let marketplaces like Silk Road operate for years before law enforcement caught up. You cannot keep the parts of that crowd you approve of and remove the rest. The protection and the exposure come from the same mechanism, and pretending otherwise is just a more comfortable way of not thinking about it.
Sovereign Mind lens
At Ideapod, we use a framework called the Sovereign Mind to work through exactly this kind of story, one where the popular version turns out to be missing the mechanism that made it true in the first place.
- Unlearning: the assumption that privacy tools and surveillance tools sit on opposite sides of a line. Here, the same design served both, and had to, by necessity rather than accident.
- Restoration: rebuilding the habit of asking who actually needed a system to exist before asking who benefits from it now, since those are rarely the same answer.
- Defense: resisting whichever simplified version of this story is easier to repeat, “the government spied on you” or “the government gave you a gift,” when both are true and neither is the point.
Why I think this still matters
This isn’t trivia about a browser. It’s a useful test for how we think about independence more broadly, about how much of what protects our ability to communicate freely was built for us versus built around us as an unavoidable side effect.
Most people who value privacy today didn’t get it because someone decided ordinary people deserved anonymous communication as a right. They got it because a narrow military research problem happened to require exactly the same thing. I don’t find that cynical. I find it clarifying, and clarifying in a way that should make us less sentimental about how durable any of this actually is. Infrastructure built as someone else’s side effect can be withdrawn the moment that someone else’s need changes.
Closing reflection
The part worth keeping isn’t that intelligence agencies built Tor. It’s that the tool only works because it no longer belongs to them — the moment it stopped being exclusively theirs is the same moment it started doing its job.
I think about this whenever privacy gets discussed as a personal setting, something you opt into or configure. It rarely works that way in practice. It depends on who else is standing next to you, whether they’re allowed to stay, and whether you’re willing to accept that the protection you want for yourself only exists because it also covers people you’d rather it didn’t. That’s the tangled version. It’s also the true one.
