There is a particular kind of policy proposal that never quite dies. You defeat it, watch it fall, and a season later it is back wearing a slightly different jacket. Client-side scanning is one of those proposals. The United Kingdom has been drawn back to it again, and the pitch this time is delivered with a reassuring technicality: nobody is asking to break encryption. They only want your device to take a look at your messages before the encryption happens.

That distinction is meant to sound modest. It is not. It is the whole game.

What client-side scanning actually does

End-to-end encryption works because a message is scrambled on the sender’s device and can only be unscrambled on the recipient’s. In between — on the wire, on the company’s servers, in the hands of anyone who intercepts it — the content is unreadable. Signal, Apple’s iMessage, WhatsApp and others build their privacy promise on exactly this property: not even the provider can read what you send.

Client-side scanning (CSS) does not attack the math. It steps around it. The plan is to install software on your phone that examines your photos and messages in the clear — on your device, while they are still readable — and compares them against a list of prohibited content, typically using “hash matching” against known illegal material. Only after that inspection does the message get encrypted and sent.

The official line is that this leaves encryption “intact.” Technically, the cipher is untouched. But the purpose of encryption was never to protect a stream of meaningless bytes. It was to keep your communication private from everyone but the person you chose to talk to. A system that reads your content before it is encrypted has already defeated that purpose. The lock on the door is irrelevant if there is an inspector standing inside the room, watching you pack the box, reporting on what they see.

The UK’s renewed push

The legal machinery for this lives in the Online Safety Act, which gives the regulator Ofcom the power to require services to use “accredited technologies” to detect child sexual abuse material — including in private, encrypted messaging. No such technology has actually been accredited or mandated yet, and Ofcom’s final guidance has been expected around spring 2026. The power, though, sits on the books, loaded and waiting for a future government to decide the political moment is right.

Running alongside it is the Investigatory Powers Act, the regime under which the Home Office reportedly served Apple with a secret order — a Technical Capability Notice — demanding access to encrypted iCloud data. Apple’s response was to withdraw its Advanced Data Protection feature from UK users rather than build a way in, and to challenge the order before the Investigatory Powers Tribunal. By reporting in 2025, after pressure that reportedly included the US intelligence community, the UK appeared to step back from that particular demand. But “stepped back” is not “abandoned.” The statutory powers remain; only the will to use them publicly has, for now, wavered.

This is the pattern worth watching. The UK has not passed a single dramatic “scan everyone” law. It has assembled a set of powers — Online Safety Act accreditation, Investigatory Powers notices — that could, in combination, compel device-side scanning whenever an administration chooses. Signal has said plainly it would leave the UK market rather than implement scanning. Element and others have signalled the same. Apple has already shown it will pull features rather than comply. These are not bluffs born of stubbornness; they are companies recognising that a single backdoor cannot be made to exist only inside one country’s borders.

Why cryptographers refuse to bless it

This is not a fight between technologists who “don’t care about child safety” and officials who do. The most respected names in the field have examined CSS specifically and concluded it cannot be made safe.

In Bugs in Our Pockets: The Risks of Client-Side Scanning, a group including Ross Anderson, Whitfield Diffie, Ronald Rivest, Bruce Schneier, Susan Landau and Carmela Troncoso worked through the design space and found, in their words, no room for a system that delivers real benefits to law enforcement without unduly endangering the privacy and security of ordinary, law-abiding people. The paper was peer-reviewed and published in the Journal of Cybersecurity. Its conclusion is not a slogan; it is the considered judgement of people who have spent careers building the very systems being discussed.

The reasoning is structural, and it comes down to two facts.

First, the scanning apparatus is the danger, not the policy attached to it. Once every device ships with the capacity to inspect content against a centrally controlled list and report matches, you have built a general-purpose surveillance instrument. Today the list holds hashes of abuse imagery. The list is updated remotely, in secret, and the user cannot inspect it. Nothing in the technology limits it to that purpose. A future home secretary, a different government, a court order in another jurisdiction, or simply a quiet expansion of scope can repurpose it toward leaked documents, protest organising, dissident speech, or whatever the anxiety of the moment demands. The capability, once present on billions of phones, is the thing that can never be un-built.

Second, the matching is imperfect, and the failure modes are severe. Hash systems produce false positives. Perceptual hashing can be deliberately fooled, and “adversaries” who genuinely want to evade it can. So you arrive at the worst of both worlds: a system that can be gamed by the people you most want to catch, while routinely flagging innocent images from everyone else — family photos, medical pictures, intimate messages — and routing them to human review.

The European mirror

Britain is not alone in chasing this, and the European experience is instructive. The EU’s “Chat Control” proposal pursued essentially the same idea — mandatory scanning of private messages — and met sustained resistance. In March 2026 the European Parliament voted to restrict untargeted mass scanning, limiting it to people specifically suspected of abuse and subject to judicial authorisation, and rejected a clean extension of the older voluntary-scanning regime. It was, by a thin margin, a privacy win.

But notice how narrow that victory is. The voluntary derogation lapsed; the underlying impulse did not. Each retreat is framed as temporary, and the proposal returns. “Chat Control is dead, long live Chat Control” is not cynicism — it is an accurate description of how these measures behave.

That is the real lesson for the UK debate. The choice in front of us is not “scan a little to protect children” versus “do nothing.” It is whether to manufacture, install and normalise a content-inspection mechanism on every personal device — and then trust every government that ever holds the keys to it, indefinitely, to use it only as promised. Encryption is valuable precisely because it does not require that trust. Scanning before the lock clicks throws that away, and asks us to be grateful for the technicality that the lock itself was left intact.

Sources: State of Surveillance — UK encryption legal challenge tracker 2026; Computer Weekly — The UK’s Online Safety Act explained; EFF — The UK Is Still Trying to Backdoor Encryption for Apple Users; Computer Weekly — EU Parliament rejects Chat Control message scanning; Patrick Breyer — Historic Chat Control vote; Bugs in Our Pockets: The Risks of Client-Side Scanning (arXiv).