Who Can Read Your Messages When the App Says End-to-End Encrypted?
"End-to-end encrypted" makes one narrow promise: the server relaying your message cannot read it. That is a real and valuable property. It is also not the same as "only the person I sent it to can ever read it", and the gap between those two sentences is where most real-world exposure lives.
So here is the list of who can still see your messages, roughly from most to least likely.
The recipient, and whoever holds their phone
Encryption ends at the recipient's screen. Once the message is decrypted there, it is plain text on a device you do not control. The recipient can screenshot it, forward it, paste it into another chat or hand their unlocked phone to someone else.
No protocol fixes this. Disappearing-message timers are a polite request to the other client, not an enforcement mechanism. If the other end is hostile or just careless, the cryptography has already done its whole job and is now irrelevant.
Your own backups
This is the one that catches people out most often. Many apps offer chat backups to a cloud account, and whether those backups are end-to-end encrypted depends on the app and on a setting you may never have opened.
- A backup encrypted with a key the cloud provider holds can be read by the provider, and handed over on a valid legal request.
- A backup encrypted with a key only you hold (a passphrase, a recovery key) cannot.
- Your contacts' backups count too. Your side of the conversation sits in their backup, under their settings, not yours.
Check what your own app does by default. Some platforms keep the message keys inside a standard cloud backup unless you switch on an extra protection mode, and the chat itself can look perfectly encrypted while that happens.
Other devices on the account
Linked devices (desktop clients, tablets, web sessions) each hold their own keys and a copy of your history. Each one is another place a message can be read.
Look at your linked-devices list now. A forgotten laptop from two jobs ago, still logged in, is a full copy of your conversations with a lock on it that you have not thought about for years.
The server, via the keys it hands out
Here is the bit that is genuinely interesting. When you message someone for the first time, your app asks the server for their public key. The server could, in principle, answer with a key it controls, sit in the middle and re-encrypt onwards. Nothing in the encryption itself stops that.
What stops it is verification. Most apps use trust on first use: they accept the key they are given and warn you if it later changes. Comparing safety numbers or scanning a QR code out of band closes the hole, but very few people do it. Key-change warnings are the realistic safeguard, and they are often switched off or ignored.
Quick detour: group chats
Groups make this worse in a subtle way. In many designs the server helps manage membership, so a compromised or compelled server could try to add a member. Good clients show a visible notice when someone joins, and some make the existing members approve it. If your app adds people quietly, that is a design weakness worth knowing about. Then back to the main list.
The app vendor, through the app itself
You are trusting the software you run, not just the protocol it claims to implement. The vendor ships updates, and an update could change what the client does with plaintext before it is encrypted or after it is decrypted. That need not be malicious. It can be a feature.
- Cloud-hosted AI features that send message text to a server for summarising.
- A "report this message" button that forwards recent messages to the vendor's moderators. The reporter chooses to send them, so the encryption was not broken, but those messages are now readable.
- Scanning on the device before encryption, which is a policy and legal argument in its own right.
Open-source clients with reproducible builds let outsiders check the shipped code matches the published source. Closed-source ones ask you to take it on trust.
Your lock screen and your notifications
Push notifications are an easy leak. If the preview shows message text on a locked screen, anyone standing nearby can read it. On some setups the text passes through the phone's operating system and its notification service, so a vendor other than the messaging app may handle it too.
The fix is dull and effective: turn previews off on the lock screen, and ideally set notifications to show only the sender or nothing at all.
Everyone, for the metadata
Even a perfectly encrypted conversation leaves a trail: who talked to whom, when, how often, from which network, and sometimes roughly where. The server needs some of that to deliver anything at all. Apps differ a lot in how little they keep, and some have deliberately reduced what the server can see about the sender.
If your threat is "someone learns who I talk to", content encryption alone does not address it. That is a separate problem with separate tools.
A quick audit you can do tonight
- Find out whether your chat backups are end-to-end encrypted, and turn that on if the app offers it.
- Remove linked devices you no longer use.
- Turn off message previews on the lock screen.
- Verify the safety number with the one or two people where it genuinely matters.
- Ask yourself who the recipient might show the conversation to.
None of this makes end-to-end encryption less worth having. It stops the server in the middle reading your messages, which is a large and important job. It just does not stop everyone else.