On June 15, 2026, the UK government announced it will bar under-16s from social media platforms — Instagram, YouTube, TikTok, Snapchat, Facebook and X named explicitly — with the first regulations going before Parliament by the end of the year and enforcement beginning in Spring 2027. Livestreaming and stranger-contact features will be restricted for under-16s and "switched off by default" for 16- and 17-year-olds, and, in a companion move, AI "romantic companion" chatbots will be required to enforce an 18+ age floor on sexually or romantically oriented interactions. Technology Secretary Liz Kendall called it a step to "give kids their childhood back" (GOV.UK).
The legal vehicle is Section 70 of the Children's Wellbeing and Schools Act 2026, which amends the Online Safety Act 2023 to let the Secretary of State impose access or functionality restrictions on children by regulation — a power with no sunset clause (CMS Law). Messaging apps including WhatsApp and Signal are explicitly carved out; the target is platforms built around open social feeds and discovery (GOV.UK fact sheet).
The Case For It Is Not Frivolous
Before litigating the flaws, it's worth stating the strongest version of the government's case. The consultation behind this policy — "Growing Up in the Online World," run March to May 2026 — found roughly nine in ten parents backing a social media ban for under-16s, and about two-thirds of young people themselves agreeing under-16s shouldn't be on at least some platforms (GOV.UK fact sheet). That is not a fringe demand manufactured by tabloid pressure; it reflects a genuine, cross-generational anxiety about algorithmic feeds, stranger contact and livestreaming as vectors for grooming and self-harm content reaching children who are demonstrably too young to navigate them. Livestreaming in particular has been a recurring vector in real child-exploitation cases, and defaulting it off for 16- and 17-year-olds is a proportionate, narrowly-targeted fix that doesn't require a blanket ban to justify itself. Parliament also isn't writing on a blank slate: Australia has already shown that an under-16 access restriction is operationally possible to legislate, even if implementation details differ.
Where the Policy Outruns the Evidence
The trouble is not the goal but the mechanism. Age assurance "for proving whether someone is over 16" is explicitly still being designed: Ofcom has been asked to identify options that are "accurate, robust, reliable, and fair" only in the months since the announcement (GOV.UK fact sheet). In its own input to government ahead of the announcement, Ofcom flagged that verifying age at the 16 threshold has fewer available methods than verifying adulthood at 18, that current age-inference models lack evidence of an effective, privacy-preserving solution, and that age checks in any case "cannot stop all of the risks posed to children online" (CMS Law). That is the regulator responsible for enforcing the ban telling government, on the record, that the enabling technology is unproven. Building a nationwide compliance mandate on top of an admittedly unsolved verification problem inverts the normal order of tech regulation: prove the tool works, then mandate it — not the reverse.
The Electronic Frontier Foundation's critique, while framed more combatively than we'd put it, identifies the same structural gap: "there remains no reliable, privacy-preserving method of verifying the age of every internet user," which means the practical effect of the ban is to push facial scans, ID uploads or credit-card checks onto adults too, since platforms will over-verify to avoid liability (EFF). The government's own proxy signals for "probably over 16" — a 16-year-old account, a linked credit card, a pre-verified email — are reasonable-sounding shortcuts, but they're exactly the sort of ad hoc heuristics that create both false negatives (a 15-year-old with a parent's card) and privacy exposure (adults submitting ID to prove they're not children).
The Chatbot Rule Is Narrower Than It Sounds
The AI companion chatbot restriction is the more defensible half of the package on its face — an 18+ floor for bots whose function is explicitly sexual or romantic role-play addresses a real and growing harm as generative companion apps have proliferated. But Kendall's own framing constrains the rule's reach: chatbots must restrict features "specifically designed to enable sexually explicit interaction" before an age check applies (The Hollywood Reporter). A purpose-based test like that is easy to satisfy on paper while leaving general-purpose companion and role-play products — which can drift into identical intimate territory without being marketed that way — outside the gate. Regulators wrote a similar purpose-based carve-out into the Online Safety Act's original categorisation scheme, and it has already proven exploitable in practice.
The Better Sequencing
None of this argues against child-safety regulation on principle — livestreaming defaults and stranger-contact limits are sensible, low-cost interventions that don't depend on solving age verification first. But Parliament is being asked to legislate a hard access ban and an enforcement deadline (Spring 2027) before its own regulator has certified that the underlying verification technology is ready, private, or fair. The honest sequencing is Ofcom's technical findings first, a binding compliance deadline second — not the reverse, which is what's on the table now.