UK encryption policy

Ofcom's Refusal to Mandate Client-Side Scanning Is the Right Call, but Group-Level Inference in Encrypted Chats Needs Scrutiny

Ofcom says it has no power to regulate devices, so its final child-safety rules stop short of client-side scanning. Private-chat inference and s.121 remain open.

Ofcom's Final CSAM Rules at a Glance People of Internet Research · UK Feb 2027 Rules take effect Subject to parliamentary approval. 0 Accredited s.121 scanning tools No tool has yet been accredited fo… 2022 Apple scanning plan shelved Announced in 2021, dropped a year … peopleofinternet.com
Ofcom's Final CSAM Rules at a Glance People of Internet Research · UK Feb 2027 Rules take effect 0 Accredited s.121 scanning tools 2022 Apple scanning plan shelved peopleofinternet.com

Key Takeaways

On 16 November 2026 Ofcom finalised a new package of Online Safety Act measures against child sexual abuse material (CSAM). In doing so it made a statement that matters well beyond the package. According to The Record, the regulator said it "does not have powers under the Act to regulate devices," and will therefore not require client-side scanning. That is the correct reading of its remit and the correct policy outcome. The surrounding rules still raise real questions about privacy in encrypted spaces.

The strongest case for stronger scanning

The case for scanning deserves a fair hearing. Child abuse is real, and offenders do use private and encrypted services to avoid detection. Supporters argue that on-device detection is the only technique that can reach material on end-to-end encrypted services, and that a regime which cannot see inside those services leaves the most vulnerable children unprotected. Campaigners pressed Ofcom to require it. That is a serious position held in good faith, and the regulator's refusal does not make the problem go away.

What Ofcom actually finalised

Ofcom's announcement describes "a far-reaching new package of measures." It includes:

Per The Record, the proactive detection duty applies to large platforms and covers material never seen before and never fingerprinted, but only for content users share publicly. All platforms hosting user-shared content must permanently ban adults found sharing CSAM or grooming children. The exceptions include victims, and potentially people who promptly report unwanted images, which addresses the risk that CSAM is sent to harass its recipient. Ofcom expects the duties to start from early 2027, subject to Parliament. The Record puts the date at February 2027.

The scope limit follows from the legal point. A duty on a platform to scan what is posted publicly can be met on the platform's own servers. A duty to inspect messages before encryption has to be met on the user's phone, and Ofcom says the Act gives it no power over devices.

Why client-side scanning was the wrong tool

The objection to client-side scanning is structural rather than a matter of tuning. The Electronic Frontier Foundation's 2021 analysis of Apple's on-device scanning plan argued that "even a thoroughly documented, carefully thought-out, and narrowly-scoped backdoor is still a backdoor." It also argued that scanning on the device means a messaging service can no longer honestly be described as end-to-end encrypted. EFF warned about mission creep, pointing to governments that would demand the same machinery for other content. Apple announced its system in 2021 and, as The Record notes, shelved it in 2022 after the criticism.

A pro-innovation approach should take that history seriously. A scanning mandate would force every messaging developer to build a capability that cannot be limited to one category of content or one jurisdiction. It would also be a precedent that authoritarian governments could cite. Ofcom's own limits on accuracy, discussed below, show that it knows the evidence for such tools is thin.

The part that should worry civil libertarians

The rules are not free of privacy costs. The Record reports that Ofcom's revised guidance lets platforms judge a whole encrypted group on the basis of its profile picture, its description, or coded terms backed by a British police report. Ofcom itself acknowledges that this kind of enforcement "may interfere significantly with users' privacy."

Group-level inference is a proportionality problem. A group label or a keyword list is a blunt proxy, and innocent groups can be caught by it: support communities, survivors, researchers, moderators, and parents discussing a news story. Police-backed terms improve reliability, but they also make state-supplied lists an input to platform enforcement without the oversight a warrant would bring. The sensible safeguards are:

Ofcom did show restraint elsewhere. The Record reports that it dropped a plan, floated in June 2025, to require automated fraud detection, saying it had "a high degree of confidence" only in detection tools for CSAM images. It will consult on other harms next year. That is how evidence-based regulation should work: duties follow demonstrated tool performance.

Section 121 keeps the question open

The bigger risk is not in the rules published this week. Ofcom did not rule out future scanning, leaving the matter to Section 121 of the Act. The Record reports that s.121 allows Ofcom to issue a "technology notice" requiring a company to use accredited software to find CSAM, including in private messages. Before one could be issued, the technology must be accredited against government accuracy standards, which the article says came into force in September. No tool has yet been accredited.

That is the right gate, but it is a thin one. Accreditation measures accuracy, not whether a tool can be confined to one purpose, and The Record reports that a notice would not be excused merely by high costs or the need for design changes. It would be excused only where it would "materially compromise the security" of a service. Who decides that, and on what evidence, is the question that matters. If a tool is ever accredited, Parliament should expect a public technical assessment of the tool's effect on encryption before any notice is issued.

Bottom line

Ofcom has drawn a defensible line. It will require platforms to act on what they can see, in public content and on public signals, and it declines to reach into the device. The UK should keep that line. If child-safety enforcement is to be credible without eroding secure messaging, the next step is evidence on how well the new tools work and transparency about how they are used, not a scanning mandate held in reserve under s.121.

Sources & Citations

  1. Ofcom: platforms must take stronger action on CSAM
  2. GOV.UK Online Safety Act collection (updated 16 Nov 2026)
  3. The Record: UK leaves client-side scanning out of new child safety rules
  4. EFF on Apple's on-device scanning plan (2021)