The finding that frames everything else: after the UK's new online safety duties took effect in July 2025, children's overall exposure to harmful content barely moved. About seven in ten children aged 11 to 17 reported encountering harmful material both before and after the rules came into force, according to Ofcom's children's online safety tracker. Legal obligations arrived. The numbers didn't.

Ofcom's data points directly at the mechanism. Personalized recommendation feeds were the primary route through which children encountered harmful content, including self-harm, eating-disorder, and age-inappropriate sexualized material, accounting for 37% of harmful exposure before the new duties and 35% after, per the same tracker. The near-identical figures bracket a period during which platforms were legally obligated to act.

The platform-level breakdown explains why TikTok and YouTube were specifically named. Among 11- to 17-year-olds, 53% reported seeing harmful content on TikTok, compared with 36% on YouTube and 34% on Instagram, according to Ofcom's research. Ofcom said its evidence showed their feeds "are still not safe enough," Reuters reported.

TikTok and YouTube's standard response to child safety scrutiny centers on the tools they offer: parental controls, screen-time reminders, age requirements, content filters. Ofcom is not disputing that these tools exist. It is asking whether they reduce harm at scale. The evidence suggests they don't.

Age gating illustrates the gap most clearly. More British children are being asked to verify their age the share reporting this rose from 41% to 51%, with facial scanning the most common method yet 84% of children aged 8 to 12 are still accessing services with a minimum age of 13, per Ofcom's research. Verification is increasing. Underage access is not decreasing.

Voluntary safety tools tell a similar story. Internal company documents surfaced through US litigation show that screen-time management tools, break reminders, and parental controls have adoption rates below 2% among minor users, according to a February 2026 analysis by the Knight-Georgetown Institute. The same analysis found internal records indicating TikTok imposed design constraints requiring that new screen-time tools reduce usage by no more than 5%, and that Meta's own projections anticipated 99% of teenagers would not use optional "take a break" features. The internal documents suggest some tools were designed with tight engagement constraints, raising questions about how much impact they were ever expected to have.

A separate NSF-backed analysis of 352 safety-related press releases and blog posts published by YouTube, TikTok, Meta, and Snapchat between 2019 and 2024 found the same pattern in company communications: platforms regularly described safety features but rarely provided evidence of their effectiveness once deployed. Claiming a feature exists and demonstrating it reduces harm are different things. Regulators are now insisting on the latter.

Ofcom's announcement drew a deliberate line between platforms that committed to new measures and those that did not. Snapchat agreed to block adult strangers from contacting children by default, remove prompts encouraging kids to connect with unknown adults, and roll out stronger age verification to all UK users over the summer. Roblox committed to giving parents the ability to disable direct messaging entirely for users under 16. Meta announced plans to use AI tools to detect sexualized conversations between adults and teenagers in Instagram direct messages all framed by Ofcom as substantive new commitments.

These concessions matter, but they address a different problem: contact-based risks, where adults can reach children directly. Ofcom's specific dispute with TikTok and YouTube centers on feed-driven content exposure harm that arrives through the algorithm without any human initiating contact. A platform can eliminate direct messaging risks entirely and still serve harmful content to children through a recommendation system that amplifies it at scale.

That distinction is why rival action doesn't resolve TikTok's and YouTube's position. Both were asked specifically to explain how they would make recommendation feeds safer. Neither committed to significant changes, Reuters reported. Ofcom also said it is not currently convinced that any of the major platforms' existing age-enforcement commitments will effectively prevent under-13s from accessing their services, per LBC a broader caveat that makes clear the problem extends beyond the two platforms named today, even if Thursday's sharpest criticism was targeted.

Thursday's announcement is a step toward enforcement, not a final warning. Chief Executive Dame Melanie Dawes said Ofcom is "determined to force through further changes, using the full extent of our powers," per Neowin. The regulator plans an independent audit of content moderation systems, algorithms, and age-verification processes, and said enforcement action, which could include a fine, is possible if platforms are found non-compliant with their duties to prevent children's exposure to harmful content.

The standard Ofcom is now applying is worth stating precisely. It is not enough to have safety features. Platforms must be able to show evidence that those features reduce harm, particularly in recommendation systems. That is a materially higher bar than listing tools in a risk assessment and it aligns with what the Knight-Georgetown Institute identified earlier this year as the central gap in platform governance: risks are increasingly well-described, but evidence that mitigations actually work remains largely absent.

There is also a legislative gap Ofcom is pushing the government to close. Current UK law does not clearly require platforms to prevent underage users from accessing their services at all. A government consultation on children's online experiences, including potential restrictions on addictive design features and age limits, closes next week, per LBC. The audit, the potential fines, and that consultation are all pointing in the same direction: toward requiring platforms to prove outcomes rather than describe intentions.