Client-Side Scanning: Reading Your Messages Before You Send Them

CaptainDumbass · July 6, 2026, 4:11am

This post explains a surveillance technique and the rights questions it raises. The technical conclusions draw on work by leading security and cryptography researchers, linked throughout so you can verify every claim. Details are accurate as of publication on July 5, 2026. Proposed laws like the EU’s Chat Control change quickly, so check the linked sources for the current picture.

End-to-end encryption means a message is scrambled on your device and only unscrambled on the recipient’s device. Nobody in the middle can read it. Governments that dislike that have stopped trying to break the encryption directly. Instead they want to read your messages before they get encrypted, right on your own phone. That is client-side scanning, and security experts consider it one of the most dangerous ideas in modern technology policy.

The neat description comes from the encryption debate itself. The lock on the door still works. Someone is just standing inside the room.

How Client-Side Scanning Works

A client-side scanning system runs on your device. Before a photo or message is encrypted and sent, the software checks it against a list of forbidden content. The usual method is perceptual hashing, which turns an image into a short fingerprint and compares it to fingerprints of known banned material. If it matches, the system can block the content or report it to authorities automatically.

Because the scan happens on your device before encryption, the people pushing it claim the encryption is untouched. That is a technicality. If something on your phone can read your message before it is sealed, then flag it and report you, the privacy guarantee of end-to-end encryption is functionally gone. More than 300 scientists said exactly this in a 2023 open letter, as did a group of fourteen of the world’s leading cryptographers in a paper titled “Bugs in Our Pockets: The Risks of Client-Side Scanning.” Their conclusion was blunt. This is bulk surveillance installed on everyone’s personal devices.

Apple Already Tried and Backed Off

This is not theoretical. In August 2021, Apple announced a client-side scanning system to check photos headed for iCloud against a list of known child ■■■■■■ abuse material. The backlash from security researchers was immediate and overwhelming. They showed the system could be fooled, could produce false matches and could be quietly repurposed. Apple abandoned the plan in 2022.

The technique did not die with Apple’s retreat. The European Union has been pushing a regulation, widely called Chat Control, that would require messaging platforms like WhatsApp, Signal and Telegram to scan content on the device. Signal’s president, Meredith Whittaker, has said Signal would leave the EU market rather than break its own privacy promises.

Why Experts Say It Cannot Be Made Safe

The objections are technical, not political, and they hold no matter how good the intentions.

It does not catch the people it targets. Perceptual hashing is not a strong cryptographic function. Anyone with moderate skill can alter an image so it slips past the filter. So the serious offenders evade it while ordinary users get scanned.
It produces false matches at scale. The same weakness lets attackers craft an innocent-looking image whose fingerprint matches banned material, then send it to a target. The victim, a journalist, an ex-partner or a political opponent, gets flagged and reported. The system becomes a weapon for harassment.
It cannot be limited to one kind of content. The scanner does not understand what it is looking for. It checks against a list. Change the list and it scans for anything: terrorism, protest organizing, anti-government material, leaked documents. The list can be swapped with a remote update, and because the contents are secret, no user can verify what their own device is scanning for. The EU has already floated expanding scanning beyond CSAM to terrorism and organized crime.

That last point is the one that matters most. Once the scanning infrastructure exists on every device, the only thing standing between “scan for child abuse” and “scan for dissent” is a policy decision and a software update.

The Rights in Question

The Fourth Amendment and its equivalents. Searching everyone’s private content with no warrant and no suspicion is the definition of a general search, the exact abuse that warrant requirements exist to prevent. The cryptographers’ paper frames it directly: the direction of travel is bulk scanning of everyone’s private data, all the time, without warrant or suspicion.

Free expression. People who know their private messages are being scanned censor themselves. Journalists cannot protect sources. Dissidents cannot organize. The chilling effect is the point, whether or not anyone intends it.

The same pattern you have seen before. An on-device scanner that checks your files against a secret list and reports violations is the identical architecture proposed for print blocking on 3D printers. One scans photos, the other scans design files, but the design is the same: mandatory scanning on hardware you own, against a list you cannot see, that can be expanded at will. When you recognize the shape of it, you can spot it wherever it shows up next.

What You Can Do

This technology only ships if lawmakers mandate it, so it gets decided in legislatures.

If you are in the EU, the Chat Control proposal is the immediate fight. Tell your representative you oppose mandatory scanning of private messages. In the US, watch for bills that pressure companies to weaken encryption or scan around it, and tell your representatives you want strong encryption protected.

In the meantime, use end-to-end encrypted messaging and support the apps and organizations that have said they will not build scanning into their products. The more people who depend on real encryption, the harder it is to take away.