Britain is choosing a behavioural default, not a lockout, for 16- and 17-year-olds. Under the Department for Science, Innovation and Technology’s new plan, selected social platforms would block access by default from midnight to 6am and switch off autoplay and continuously personalised feeds. Older teens could reverse those settings.

That legal and product distinction matters. A ban removes access; a default establishes the starting state while preserving a user’s choice. Britain’s earlier debate focused heavily on whether under-16s should be kept off large platforms. TECHi’s coverage of British teenagers opposing that proposal captured the rights and enforcement objections. The new question is narrower and more measurable: can a protective setting that is deliberately easy to override still alter behaviour?

Robert Booth’s Guardian report gives the sharpest criticism to Baroness Beeban Kidron, founder of the 5Rights Foundation. Kidron calls the measure cosmetic because a teenager can undo it in a few clicks. That objection identifies a serious enforcement gap, but it treats all overrideable defaults as empty. Ofcom’s newest behavioural evidence does not support so absolute a verdict. The policy may have a mechanism; it does not yet have proof of an outcome.

The rule is narrower than the headline

The operative proposal is a configuration mandate. Social services in scope would have to place 16- and 17-year-old accounts into a midnight-to-6am block, stop videos from automatically rolling into the next clip, and disable feeds that continually serve personalised material. The account holder would retain the ability to change those settings.

The government says the first set of regulations should reach Parliament by the end of 2026 and take effect in spring 2027. Its implementation fact sheet leaves important operational choices to the next stage, including which services and functions will be covered and how Ofcom will judge acceptable age assurance.

That sequence is consequential. A ministerial announcement tells platforms the intended destination; regulations define the legal duty; Ofcom’s codes and enforcement practice determine what compliance looks like. Until those later documents exist, claims that Britain has already “introduced a curfew” overstate the present position. It has announced a rule it intends to make.

Kidron is on firm ground when she rejects the announcement as protection in itself. Teenagers cannot be protected by a press release. Platforms must identify the relevant age group, apply the setting to the correct services, preserve it across devices and updates, and avoid interface design that quietly steers users toward opting out.

Defaults can change choices without removing choice

An override button does not make a default meaningless. Defaults work through inertia, attention and the perceived normal state of a product. A user who must actively choose the less protective setting faces a different decision from a user who receives unrestricted access automatically.

Ofcom published a randomised controlled trial of protective social-media defaults on July 13, 2026. It tested children aged 13 to 15, older teenagers aged 16 to 17, and adults on a mock platform. Across settings involving location sharing, direct messages and network expansion, safer defaults had high retention. Ofcom also points to an earlier child trial in which 69% kept a preselected “don’t recommend harmful content” option, compared with 51% in the control condition and 54% when the safer option was listed first but not preselected.

Those experiments do not prove that a six-hour social-media block will survive contact with a real teenager’s social circle. Ofcom says as much: a mock platform cannot reproduce peer pressure, habit, notifications or the fear of missing a conversation. Still, the trials establish a mechanism. Where the starting state changes, a meaningful share of users may leave it unchanged even when changing it is possible.

The policy dispute should therefore move past the binary claim that a setting is either compulsory or useless. The better question is how much friction is legitimate, how clearly the choice is presented, and whether the safer state persists. A default can influence conduct without becoming a prohibition. That is precisely why regulators must measure it.

The government’s pilot cannot carry the outcome claim

The supporting pilot is much weaker evidence than the Ofcom trial for causal claims. DSIT’s full social-media intervention report involved 309 families and compared three household interventions: a 15-minute daily limit, an overnight block from 9pm to 7am, and removing selected social apps. Families took part for a month.

The overnight option was described as the most manageable. Participants most consistently associated it with better sleep, earlier bedtimes and feeling more rested. Those accounts are relevant. They explain why ministers selected a curfew rather than a severe daily cap or complete removal.

They do not establish that the intervention caused the reported improvements. The research team expressly describes the work as qualitative, self-reported and exploratory. Participants were not randomly assigned in a way that supports a population estimate; many expressed a preferred intervention, some changed groups, and the study did not directly monitor device or app usage. A relatively small, self-selecting sample cannot show how all British 16- and 17-year-olds will respond.

The policy also differs from the pilot. The research tested a 9pm-to-7am household curfew; the proposal uses midnight to 6am. The pilot restricted Snapchat, TikTok, Instagram, Facebook, X and Reddit while allowing YouTube, WhatsApp, iMessage and school tools. Future regulations have not yet fixed an equivalent service list. Participants also reported workarounds, movement to other devices or services, and daytime use that remained unchanged or shifted around the blocked period.

This is the central evidence gap. Government can fairly say the pilot made an overnight restriction look more tolerable than the alternatives. It cannot fairly present the pilot as proof that the proposed national default will improve sleep, concentration or family life. Those are hypotheses to test after implementation.

Compliance needs more than an on switch

A credible rule would make platform performance visible without collecting a new trove of sensitive information about individual teenagers. Ofcom can require aggregated, privacy-preserving reporting and independent testing. At minimum, the regulator should be able to answer:

  • What share of accounts identified as 16 or 17 received every required default?
  • What share disabled each setting after 24 hours, seven days and 30 days?
  • Did the safer state survive app updates, new devices, account recovery and a change in age?
  • Did usage shift into exempt apps, web versions, secondary accounts or VPN-routed sessions?
  • Did platforms use prompts, colours or button placement that made opting out easier than opting back in?

These measures distinguish a real safety control from a compliance screenshot. They also create an honest failure condition. If most users immediately opt out, platforms steer them out, or activity simply migrates to services outside the rule, the policy will not have delivered the result ministers claim.

Australia’s experience is a warning against counting an account action as a social outcome. TECHi reported when millions of under-16 accounts were deactivated under Australia’s ban. That figure demonstrated platform action, not an end to adolescent social-media use. Britain should not repeat the measurement error by treating the number of accounts initially defaulted as proof of better sleep or lower exposure to harmful material.

The chatbot promise runs ahead of the present legal perimeter

The announcement also says ministers intend to require breaks for under-18 chatbot users and are considering restrictions on systems that pose serious threats or give dangerous, misleading or unverified mental-health advice. Those statements are policy intentions, not an existing chatbot code.

The legal perimeter is still being built. A government delegated-powers memorandum explains that some standalone generative-AI services remain outside the Online Safety Act when they neither let users share content nor search the live internet. The new power is designed to bring currently unregulated services into scope through later regulations.

That gap matters because a social platform’s embedded assistant and a standalone companion bot may present similar risks while falling under different legal definitions. TECHi previously examined the government’s child-safety response to the Grok image scandal. The current package moves the argument from outrage toward rulemaking, but its chatbot portion will remain aspirational until Parliament and Ofcom translate it into duties that reach the services ministers are describing.

Conflating the mature and immature parts of the package would mislead readers. The social-media defaults have a stated timetable and an existing statutory route. Chatbot breaks, mental-health controls and possible bans still require sharper definitions, scope and proportionality tests.

Verdict: the weakness is real, but the mechanism is not imaginary

Kidron’s criticism catches three defects: an easy override limits the force of the rule; the covered service list is unsettled; and ministers are using a qualitative pilot to support language that sounds causal. Calling the default pure theatre goes too far.

Ofcom’s trials show that safer starting settings can remain in place for many users. DSIT’s pilot shows that families found an overnight restriction more workable than severe time limits or deleting apps. Neither result guarantees that the proposed midnight rule will change real-world behaviour at national scale.

The defensible verdict is narrower. Britain has selected a plausible behavioural instrument and attached claims to it that the available evidence cannot yet prove. The policy earns credibility only if the final regulations prevent manipulative opt-out design, Ofcom publishes retention and displacement data, and government separates account-setting compliance from actual wellbeing outcomes.

If override rates are high, use migrates elsewhere, or sleep and harm indicators do not improve, the intervention should be revised rather than defended as a success. If the safer defaults persist without blocking legitimate access, the “for show” charge will have been too severe. The result must be measured; the announcement cannot decide it.