A Line in the Sand, Drawn Early
On 15 June 2026, Prime Minister Keir Starmer announced that the UK will bar under-16s from social media platforms including TikTok, YouTube, Instagram, Facebook, Snapchat and X, exempting messaging apps and YouTube Kids, while setting a minimum age of 18 for AI "romantic companion" chatbots and restricting livestreaming and stranger-contact features for minors (GOV.UK). Legislation is due before Parliament by Christmas 2026, with the regime in force by spring 2027, and Ofcom will be the enforcer (NPR).
The steelman case is real. Parents and child-safety groups have spent a decade documenting how engagement-optimised feeds, autoplay and stranger-contact tools expose minors to grooming, self-harm content and compulsive use patterns no child-focused settings menu fully neutralises. Even after Ofcom's Protection of Children Codes under the Online Safety Act 2023 took legal effect, requiring "highly effective" age assurance against pornography, self-harm and suicide content, ministers evidently concluded that duty-of-care enforcement wasn't moving fast enough to blunt those harms for the youngest users. A hard age floor is simpler to explain to parents than a matrix of platform-by-platform risk assessments, and simplicity has real value when the target audience is a exhausted parent, not a regulator.
The Problem: Scrapping the Test Before Reading the Results
That steelman collapses on timing. The Online Safety Act's child-protection duties only became legally binding on 25 July 2025 — well under a year before this announcement. Ofcom itself, responding to the ban, noted it has "driven some of the strongest changes of any online safety regulation in the world, from widespread age checks to grooming protections for children," and said it would spend the coming months designing age-assurance mechanics for a policy that Parliament hasn't yet passed (Ofcom statement). In other words: the systemic, duty-of-care model Parliament built in the Online Safety Act 2023 (legislation.gov.uk) is being superseded by a blanket age wall before regulators have had a full enforcement cycle to show whether the original law worked.
That matters because the alternative model — a hard age cutoff rather than graduated content and design duties — has already been tried, at scale, in Australia. Its Online Safety Amendment (Social Media Minimum Age) Act took effect on 10 December 2025, and by mid-January 2026 the government reported more than 4.7 million accounts believed to belong to under-16s had been deactivated, removed or restricted (eSafety Commissioner). That is a genuinely large number — and also not proof of success. By late March 2026, Australia's communications minister confirmed investigations into Meta, TikTok and YouTube over children repeatedly attempting age verification after already declaring themselves under-age, meaning the ban's own enforcement data shows large-scale attempted evasion rather than settled compliance. A UK version, layered on top of the same major platforms, should expect the same friction: VPN circumvention, falsified ages, and migration to whatever isn't covered.
That last point cuts against the policy's own design. Exempting messaging apps is defensible — WhatsApp isn't a public feed — but it also creates the obvious escape valve: displaced under-16 social activity doesn't vanish, it moves into encrypted group chats that carry none of the Online Safety Act's content-moderation or reporting duties Ofcom currently enforces on the platforms being banned. As the Electronic Frontier Foundation argued shortly after the announcement, the ban "will cause more harm than it prevents" partly because the age-verification infrastructure it requires — uploaded ID, facial-age estimation, or third-party verification services — extends surveillance to the adults who still use these platforms, not just the minors being excluded (EFF).
"This week, politicians in the UK pushed forward with plans to eviscerate privacy and free speech on the internet by announcing a ban on social media for users under 16." — EFF, 19 June 2026
That is a strong claim, but it identifies a real asymmetry: age assurance is a compliance cost every platform must absorb, and it scales far more easily for incumbents with existing identity infrastructure than for new entrants trying to compete with them. A pro-innovation regulator should worry that a blunt age wall, badly implemented, entrenches Meta, Google and ByteDance precisely because they can afford the verification stack a startup cannot.
The Better Sequence
None of this argues for inaction on child safety online — the underlying harms are real and documented. It argues for sequencing: let Ofcom's Protection of Children Codes run through a full enforcement cycle, publish the compliance and harm-reduction data, and then decide whether a categorical age ban adds anything the existing duties don't. The AI companion chatbot age-gate, by contrast, targets a narrower and more clearly age-inappropriate product category and is easier to justify on current evidence. Parliament has until Christmas 2026 to legislate; that window should be used for evidentiary scrutiny of what Australia's rollout actually achieved, not treated as a formality before a deadline that was set before the debate began.