UK Confirms Mandatory AI Chatbot Breaks for Under-18s and Default Midnight Curfews for Older Teens

Britain will require AI chatbot providers to force periodic breaks on users under 18, and will oblige social media services to switch on overnight curfews and turn off infinite-scroll mechanics by default for 16- and 17-year-olds — but it will not ban or age-gate VPNs. Technology Secretary Liz Kendall set out the package on 15 July 2026 in a written statement to Parliament and an announcement from the Department for Science, Innovation and Technology — the second and final instalment of the government's response to its Growing up in the online world consultation.

A month earlier, the government announced its headline measure: a ban on social media accounts for under-16s, modelled on Australia's but going further by also restricting livestreaming and stranger contact. The July package fills in either side of that line — older teenagers outside the ban, and AI services outside social media altogether.

The legal vehicle matters, because the shorthand of a "new bill" is wrong. This is not fresh primary legislation, nor is it Ofcom guidance — Ofcom is the enforcer, not the author. Section 70 of the Children's Wellbeing and Schools Act 2026 inserted a new section 214A into the Online Safety Act 2023, obliging ministers to make regulations imposing age or functionality restrictions on children's social media use — a concession extracted after government defeats in the Lords. July's measures will be delivered as secondary legislation under those powers, subject to approval by both Houses, with the first regulations due before the end of 2026 and the regime biting in early 2027.

What the Duties Actually Say

Social media services must apply default overnight restrictions for 16- and 17-year-olds, with alerts and push notifications muted between midnight and 6am, plus round-the-clock default restrictions on persuasive design features including autoplay and personalised recommender feeds. These are defaults, not prohibitions — older teenagers can toggle them off, a choice the government defends on autonomy grounds and critics have seized on.

For AI, the duty is narrower and less developed. Kendall told the Commons the government "will therefore require mandatory breaks for under-18s using chatbots and will work with experts on how frequently, and for how long, these should happen." No interval has been published. DSIT will also work with the Department of Health and Social Care and the Medicines and Healthcare products Regulatory Agency on chatbots offering "harmful, inaccurate or unverified mental health advice," and says it will "consider a full range of options, including, if necessary, banning certain services that present a threat to children." That builds on the June package, under which the UK became the first country to ban chatbots offering sexualised content to children.

The stated rationale is emotional dependency. Of the consultation's 116,211 responses, more than half of those aged 16 to 21 backed restrictions on extended chatbot use, and 18 percent raised concerns about reliance on chatbots for mental health support. Nearly one in three children said they want help managing screen time.

On circumvention, the government declined the most aggressive option. "VPNs have legitimate privacy and security uses and we will therefore not age-gate or ban them," Kendall's statement reads. The burden shifts instead to platforms. Ofcom has been asked to report by October on what highly effective age assurance looks like for verifying someone is over 16, and — with Information Commissioner's Office input — to research what services can do to detect VPN use. Enforcement runs through the Online Safety Act's existing machinery, where Ofcom can fine up to 10 percent of global annual turnover.

Two Cases, Both Partly Unproven

Child-safety advocates argue that defaults work. Online Safety Minister Kanishka Narayan, pushing back on Sky News against the charge that teenagers will simply switch the curfew off, cited a government pilot involving more than 300 teenagers and parents. "In October, for example, some platforms introduced these defaults of this sort — 90%-plus teenagers said to us that they've maintained those defaults as well," he said. "The evidence base is clear, the motivation is very clear and I wouldn't do the disservice to teenagers of saying they're all going to switch it off."

Charities are supportive but unsatisfied. NSPCC chief executive Chris Sherwood warned that unless followed up, the measures "will be a sticking plaster that fails to address the addictive design features which are driving high screentime and undermining children's wellbeing." Molly Rose Foundation chief executive Andy Burrows called it "yet another piecemeal set of announcements not the comprehensive plan for children's safety that's required." Children's Commissioner Rachel de Souza called it a "positive step" but wants "to know more about how the policies, such as a curfew, will be delivered."

The objections come from two directions. Politically, the opt-out is the target: shadow education secretary Laura Trott called the plans a "dog's dinner." Civil liberties groups attack the age-assurance infrastructure the whole edifice rests on. Big Brother Watch director Silkie Carlo has argued that "the only way to ban children from social media is through mandatory online ID checks for us all, adults and children alike," while head of advocacy Jack Coulson said the proposals "will force the public to trust their IDs to companies with serious track records of leaks and hacks." LSE professor Sonia Livingstone has raised a different worry: cutting vulnerable teenagers off from trusted support overnight could backfire.

Then there is workability. Australia's ban has been live since 10 December 2025; a March 2026 compliance update found many under-16 accounts removed or restricted, but a "substantial number" retained.

What to Watch

Ofcom's October report on age assurance for over-16 verification will set the technical bar for the whole regime. The first regulations, due before the end of 2026, will reveal whether the chatbot break duty carries a specific interval or is left to codes of practice — and, critically, which AI services fall in scope, since general-purpose assistants sit awkwardly within the Online Safety Act's definitions. Then comes the parliamentary vote, where the opt-out is the likeliest flashpoint; ministers have already indicated it will be kept under review.