Who holds the key, and why it’s the whole game.

Last time I said the method is public and the key is the secret. This is the part where that stops being a nice idea and starts telling you which apps you can actually trust with your writing.

Any company that encrypts your notes has to make one decision before it writes a line of code. Does it keep a copy of the key, or not?

Everything you can and can’t trust about that company follows from the answer and it’s almost never on the pricing page.

A key is not a password

Let’s separate those two, because they get used as if they were the same thing, which they’re not.

Your password is something you might remember. A key is a very long number, so long, in fact, that writing it out would fill a line or two of this page, and nobody’s memory can hold that.

They are connected though. Plenty of systems take the password you typed and run it through a slow, deliberate process to produce a key. Password in, same key out every time. The password’s job is to be the starting point. The algorithmic processing turns the password into something that can be used for encryption.

That’s why most sites these days make a fuss about the strength of your password. If it’s used to generate a key, it’s not just the password that can be guessed. It produces a guessable key too.

Catchlight goes a different route, twelve ordinary English words generated on the phone the first time you open it, and those words derive the key. I’ll come back to why we don’t use a password.

The locksmith who kept a copy

Picture a locksmith fitting your front door. He hands you a key, you pay him, he leaves. What you don’t know is whether he dropped a second key in a drawer at the shop on the way home.

He’s telling the truth about the lock. It’s a good lock and it’s exactly as strong as he said it was. The lock was never really the problem, the possibility of a copy in the drawer is.

Encrypted software has the same story. Two apps can use identical encryption, the same published, tested method, and one of them can read everything you write, while the other can’t. Not because one has better maths or systems, but because one kept a copy of your key.

Where your key actually lives

In practice, there are only a few ways to handle an encryption key and it’s useful to be able to spot which one you have.

Only you hold it. The key is made on your device, from something only you have, and nothing that could rebuild that key ever leaves. The company just stores ciphertext it can’t decrypt. This is what’s meant by “zero-knowledge” and, somewhat disappointingly, it’s rarer than most of the marketing would have you believe.

You hold one and so does the company. By far the most common setup, and the one that hides most comfortably behind the word “encrypted”, sees your notes encrypted, both on the disk and in the cloud. Sounds good, but the company can also open them whenever it has a reason to. Maybe for a support request, a subpoena or any number of reasons, possibly including the sale or use of the data itself.

The company holds it and you never touch one. Common in web apps, where you sign in with an email and a password, and everything happens on the company servers. The encryption is real and it’s protecting their storage from other people, but it’s not protecting your notes from them.

Somebody holds a spare on your behalf. Recovery keys or a copy kept in another part of the same company, sometimes optional and clearly explained, sometimes on by default. What you end up with is the second variation, disguised as the first.

Only the first type survives the question that matters, which is, what happens when someone with power leans on that company for access?

The test that takes two minutes

Get locked out of the house, the locksmith returns, but does he have a key, must he pick the lock or smash a window?

Open the app you keep your private writing in and start the forgotten password process. Don’t finish it, just watch what you’re offered.

If it emails a link, lets you set a new password, and you then find all your old notes are sitting there afterwards, exactly as they were, then somebody other than you can open those notes. Your old password produced the key that unlocked the old writing. When you reset the password and the writing came back regardless, the key was never really yours to begin with.

If it warns you that recovering the account means losing the data, or it asks for a recovery phrase you were told to write down on day one, that’s a system where you’re the one holding the key.

No policy reading, no cryptographic expertise, no marketing to unravel. The recovery possibility gives it away every time, because a company can only hand back what it’s already able to read.

What “we can’t read your data” has to mean

It’s a strong claim and one you can confirm, now you know what to look for. For it to be true, rather than aspirational, a few things have to be in place so the company simply doesn’t have any options.

The key has to be made on your device, rather than issued to you. If it arrived from a server, it existed on a server, and it can be read from a server.

Nothing capable of rebuilding or recovering your key can ever leave your device. Not the phrase, not the password, not a helpful backup copy stored “for your convenience”.

The company also has to have explained the consequences of you losing it. That is the uncomfortable bit, because it means the honest answer to “I’ve forgotten my phrase”, is that the notes are gone.

If a company promises both that it can’t read your notes and that it can get you back in, chances are the first is false, simply because the second is true.

The cost, simply put

That’s the trade you make, and I’d rather just say it than dress it up.

Catchlight hands you 12 words the first time you open it, the Privacy Phrase, and tells you to store them somewhere safe, somewhere that isn’t the phone. Those words re-derive the key on any device, which is how a second phone picks up your writing without me being involved at any point. There’s no account, so there’s nothing at my end to reset. Lose the words and no support ticket fixes it, because I’ve got nothing to fix it with.

I use words, rather than a password, for a fairly boring and somewhat obvious, reason. A password you invent tends to be one you’ve used somewhere before and it also has to survive being made into a key. Randomly generated words carry far more variety than anything you’d think up, and you can copy them onto a card without getting them wrong. You’ll even see 24 words recommended at times, but the reality is 12 gives you more than enough randomness and it’s nowhere near as daunting a job as re-entering twice as many.

The words do feel like a bit of an odd dance for about 30 seconds. What you should remember is this, they are the entire reason nobody else can read your notes.

Why this matters for your notes

Hold any app’s privacy page against a single question. Not whether it’s encrypted, because everyone can say yes to that nowadays and be telling the truth. The only question worth asking is, who holds the key?

I pulled that thread for one company in can Apple read your notes, and watched a whole country find out the hard way when the UK lost Advanced Data Protection. If you want the short version of the term itself, what zero-knowledge actually means covers it, the glossary keeps the definitions in one place, and the comparison pages line the apps up on this exact point. For the app most people already have, there’s Catchlight compared with Apple Notes.

Remember, the lock on the front door is rarely the issue, the drawer at the back of the shop, with a key, just might be.

Next time, three phrases that get used as if they were interchangeable, end-to-end, zero-knowledge and on-device. They promise different things and one of them is doing less work than you think.