What was announced
On July 15, 2026, Technology Secretary Liz Kendall confirmed that platforms including Instagram, TikTok and YouTube will be required to make 16- and 17-year-olds' accounts unavailable by default between midnight and 6am, and to switch off autoplay and personalised, algorithmically-driven feeds for that age group by default (GOV.UK). Teenagers can turn either setting back on themselves. The first regulations go to Parliament before the end of 2026, using powers under the Children's Wellbeing and Schools Act 2026, with the changes taking effect in spring 2027 alongside the previously announced ban on social media accounts for under-16s (GOV.UK).
Kendall framed the policy as closing a "cliff edge": rather than 16-year-olds moving overnight from a total platform ban to unrestricted access, they get a softer, opt-out regime. "Even as young people gain greater independence at 16, they should still be protected from the most addictive online features that can have a harmful impact on their wellbeing," she said in the announcement.
The steelman
There is a real, evidence-backed case for this. Autoplay and infinitely-scrolling personalised feeds are not neutral product choices — they are engagement-maximising defaults, deliberately engineered to extend session length past the point a user would otherwise stop. For an age group still developing impulse control, defaulting those features off rather than banning the platform outright is a proportionate intervention: it targets the specific mechanism of harm (compulsive, late-night use) rather than the medium itself, and it preserves teen agency by letting them switch settings back on. That is a meaningfully different, more liberal design than the blanket under-16 ban running alongside it. Parents, unsurprisingly, like it: nine in ten surveyed backed the broader package of under-16 restrictions (GOV.UK).
Where the evidence thins out
The government's public justification leans heavily on a DSIT-commissioned pilot run by the polling firm Savanta, which surveyed 309 households before and 307 after a one-month trial in May 2026 involving more than 300 teenagers and parents (MediaNama). That pilot is doing more rhetorical work than it can bear. It tested a 9pm-to-7am curfew — three hours wider than the midnight-to-6am window DSIT actually announced — on a sample skewed younger than the policy's target group (200 participants aged 13-15, versus 109 aged 16-17). It did not test the autoplay or personalised-feed restrictions at all. In other words, no one in the pilot experienced the policy the government just confirmed. Families reported that curfews "quickly became part of their routine" and improved sleep, but that finding describes a different intervention, on a different population, for a different set of hours.
That matters because defaults are powerful precisely because most people don't change them — behavioural economics is unambiguous on this point, which is the entire logic behind making curfews and feed restrictions "off by default" rather than optional add-ons. A default-based policy is a soft mandate wearing an opt-out's clothing, and it deserves to be justified by evidence of what that specific default does, not a pilot of a stricter, untested variant.
The accountability and enforcement gap
Running this through secondary legislation under the Children's Wellbeing and Schools Act 2026, rather than a fresh bill, is efficient — but it also means a significant behavioural regulation of a legal product for legal adults-in-waiting gets a statutory instrument's level of parliamentary scrutiny, not a full debate. Ofcom, for its part, has confirmed only that it is "preparing to deliver the expansion of social media restrictions" and will detail age-assurance options in the coming months (Ofcom); as of publication there is no public detail on how platforms verify that a given account belongs to a 16- or 17-year-old specifically, as opposed to an adult or a younger child.
That gap is exactly what civil liberties groups have been warning about. Open Rights Group has argued that expanding age-gating "would multiply the scale" of privacy risk already exposed by breaches at age-verification vendors, and that the government is "playing whack-a-mole with online safety, focusing on individual harms and product features instead of confronting the structural power of dominant tech companies" (Open Rights Group). That critique is aimed primarily at the under-16 ban, but it applies with equal force here: if age-tiering 16- and 17-year-olds requires the same biometric or document-based age assurance Ofcom is exploring for the under-16 ban, the privacy cost of this "softer" policy converges with the harder one it's meant to be distinguished from.
The proportionate path
None of this means the underlying instinct is wrong. Regulating dark-pattern design — autoplay, infinite scroll, engagement-optimised feeds — for a specific, vulnerable age band is a more defensible target than blanket access bans, and defaults-plus-opt-out is a legitimately lighter-touch tool than prohibition. But DSIT should publish the Savanta pilot data in full rather than a summary, commission a trial of the actual midnight-to-6am, feed-restriction policy before the under-16 ban locks in a parallel age-assurance infrastructure, and Ofcom should specify — before regulations are laid this year — exactly what age-verification method 16- and 17-year-olds will be subject to. A default worth flipping for millions of teenagers should be tested as the default it actually is.