Introduction
A child aged fifteen can use the internet to obtain information about a medical condition, maintain friendships across geographical distances, participate in political debate and seek support for an identity which may not be understood at home. The same child can be targeted by a sexual predator, exposed to content promoting self-harm, subjected to abuse or drawn into an apparently endless stream of attention-maximising recommendations. It is not intellectually coherent to describe the digital environment as either an indispensable realm of liberation or an intrinsically harmful space. It is both a medium through which children’s rights are exercised and an environment in which those rights may be infringed. That dual character presents the central challenge for legislation which seeks to prohibit social media companies from offering specified services to everyone under sixteen.
Law Teacher writing support
Writing a law essay?
We choose a writer who is an expert in your area of law and qualified to at least the academic level you select. Most of our writers hold a master’s degree or PhD, have teaching experience, or both.
Send us your question and, if you wish, any marking criteria. We’ll write an essay to your instructions.
On 15 June 2026 the United Kingdom Government announced that it would introduce an under-16 social media restriction, accompanied by age-assurance requirements and additional controls on harmful features across other online services. Its July response to the consultation, Growing up in the online world, set out further protections for sixteen- and seventeen-year-olds, including default overnight restrictions and controls on personalised feeds. The first regulations are intended to be laid before Parliament by the end of 2026 and to take effect in late spring 2027. At the date of this essay, however, the announced restriction is not yet an operative statutory prohibition. Parliament has enacted a regulation-making power, but the actual legal scope, exemptions, compliance duties and commencement will depend upon secondary legislation and its implementation.[1][2][3]
The constitutional question is not simply whether the state may protect children from demonstrable harm. It plainly may, and in some circumstances must. It is whether a broad service-level exclusion is a lawful, necessary and proportionate means of achieving protection when it also restricts expression, association, access to information and developing personal autonomy. The difficulties are unusually acute because the proposed prohibition is directed less at particular unlawful acts than at combinations of platform design features, and because enforcement may require widespread processing of personal data by both adults and children. There is an additional question of distributive justice: the child who most depends on online communities may also be the child least easily served by the assumption that offline alternatives are available.
This essay argues that a carefully delimited minimum-age rule could in principle be compatible with children’s rights, particularly where the regulator demonstrates serious, specific risks which cannot be reduced effectively through less restrictive design controls. But compatibility cannot be inferred from the protective intention of the policy, from majority parental support, or from the existence of a similar Australian scheme. The government bears the burden of justifying the reach of the interference, the age threshold, the selection of services, its evidential assumptions and the collateral costs of age assurance. The stronger and more categorical the restriction, the more demanding that justification should become. The preferable model is a rights-sensitive, graduated scheme in which exclusion is genuinely a last resort for a defined class of high-risk services, with accessible alternatives, effective review and transparent evaluation built into its operation.
What has Parliament enacted, and what remains only a proposal?
The distinction between an enabling power and a substantive ban is fundamental. Section 70 of the Children’s Wellbeing and Schools Act 2026, which came into force on 29 April 2026, inserted section 214A into the Online Safety Act 2023. The new section allows the Secretary of State, for the purpose of protecting children against risks of harm, to require providers of specified internet services to prevent or restrict children of prescribed ages from accessing services or particular features. It gives examples of possible controls, including limits on time spent using services, restrictions at certain times of day, controls on unsolicited communications, the identification of a child’s location and livestreaming. This is a flexible delegated power, not a freestanding statutory rule that every person under sixteen is already barred from social media.[4]
Section 214A is also more discriminating than the popular shorthand of a universal ban suggests. In making regulations the Secretary of State must have regard to differences between the features of services and to the ways in which those features may affect children at different ages. The power permits exemptions and detailed enforcement provision. It therefore contains an express legislative recognition that a child of seven and one of fifteen, or a livestreaming facility and a private messaging function, do not necessarily present equivalent risks. In legal terms, this matters because proportionality should be considered not only when regulations are challenged afterwards, but when their scope is designed in the first place. The breadth of a delegated power is no substitute for a reasoned exercise of it.[4]
Section 71 requires a statement on progress towards the first regulations. In the resulting June statement, the government said it intended to lay the first service-level age restrictions by the end of 2026; the instruments are subject to an affirmative parliamentary procedure, so approval by both Houses is required before they can be made. This provides political scrutiny, although it is not equivalent to full line-by-line amendment of a bill: Parliament must assess a package of detailed rules within the procedural constraints of delegated legislation. Since definitions and exceptions will largely determine how intrusive the scheme is, the availability of evidence and a sufficiently detailed statement of rights implications will be crucial to meaningful scrutiny.[5][6]
On 2 November 2026, during a Commons debate devoted to the proposed ban, the responsible minister confirmed an intention to lay the first regulations before the Christmas recess, followed by Ofcom guidance and implementation in late spring 2027. He also distinguished that initial restriction from further rules governing particular risky features, including stranger communication on gaming platforms, which he anticipated developing on a later timetable. The precise mechanism for allowing a platform to demonstrate that it had become sufficiently safe to regain eligibility for under-16s was not settled. The debate is important evidence of ministerial intention and outstanding design choices; it is not the text of operative regulations. A rigorous legal analysis must not collapse the sequence of announcement, delegated legislation, regulatory guidance and enforcement into one completed event.[3][7]
The announced regime is not a ban on the entire internet
The policy envisages restrictions on social media services whose combined features, including user-generated content and recommendation systems, create particular risks. Government materials identify major services such as Instagram, TikTok, Snapchat, Facebook, X and YouTube by way of analogy with the Australian model, while indicating that dedicated messaging services such as WhatsApp and Signal are not intended to be covered by the service-level ban. The government says children should retain access to educational material, news, games and communication with family and known friends, although giving practical effect to those assurances on mixed-purpose platforms is a significant unresolved problem. Parental permission is not, on the stated model, intended to be a general exception allowing a provider to disregard a statutory age restriction.[8]
For sixteen- and seventeen-year-olds, the July policy response proposes a different regulatory philosophy: social media access remains available, but notifications and access between midnight and 6 am, autoplay and personalised recommendation feeds would be restricted by default, with teenagers able to change these settings. Livestreaming and communication from strangers would also be disabled by default under the wider announced protections. Under-sixteens would face restrictions on high-risk functions on services outside the core social media category, including gaming; however, the minister’s November explanation indicates that the detailed feature-specific rules may follow the initial platform ban. An essay addressing the legality of the proposals must distinguish those proposed defaults, possible future regulations and the existing obligations of platforms under the Online Safety Act.[2][7][8]
The separate question of digital consent at thirteen
Section 72 of the 2026 Act permits regulations adjusting the age threshold in Article 8 of the UK GDPR for a child’s consent to the processing of data in relation to an information society service. The existing UK threshold is thirteen; the power permits the threshold to be set between thirteen and sixteen. It does not itself raise the age to sixteen, and in any event the data-protection threshold is not a general statutory minimum age for using social media. Article 8 is engaged only in its particular context where processing is founded on the child’s consent. A service may rely on a different lawful basis where the law permits. Equating a digital-consent threshold with a prohibition on account ownership would therefore misstate both the present data-protection regime and the proposed reforms.[9][10]
The starting point: children have rights as well as vulnerabilities
The United Nations Convention on the Rights of the Child (UNCRC) does not conceptualise childhood merely as a period during which the state or parents determine what is safe. Article 3(1) makes the child’s best interests a primary consideration in actions concerning children. Article 5 recognises the responsibilities of parents and others to provide direction and guidance consistently with the child’s evolving capacities. Article 12 gives children capable of forming views an opportunity to express those views and have them given due weight according to age and maturity. Articles 13, 15, 16 and 17 protect expression, association, privacy and access to information. Articles 19 and 34, meanwhile, require protection against violence, abuse and sexual exploitation; Article 31 protects play and recreation. These guarantees operate together rather than in an order which makes online safety automatically decisive in every case.[11]
Article 3’s description of best interests as a primary consideration must not be silently rewritten as an absolute paramountcy rule. A child’s safety is of the highest importance, but an objective that qualifies as protective does not render other Convention interests irrelevant. A prohibition may reduce exposure to predatory contact while also taking away a valuable means of communicating with peers. It may protect sleep while making access to a disability-support network more difficult. Those competing effects are not reasons to reject safeguarding; they are precisely why the best-interests inquiry must be substantive, informed and open to different circumstances. Furthermore, a population-wide age restriction must assess the interests of children as a group without treating all children’s needs as interchangeable.
The Committee on the Rights of the Child’s General Comment No 25 (2021) supplies the most directly relevant interpretative framework. It recognises both risks and opportunities in the digital environment and calls for children’s rights to be respected, protected and fulfilled online. In paragraphs 58–61, the Committee explains that digital environments facilitate young people’s expression, access to information and political views, including for children in disadvantaged circumstances. It states that restrictions on digital expression, including safety filters, must be lawful, necessary and proportionate, with their rationale transparently explained in age-appropriate terms. Paragraphs 64–65 similarly recognise online association and participation in communities, subject to legitimate and proportionate limitations. Such guidance does not mean the UN committee has already determined the legality of a particular UK age rule. It does mean that a categorical prohibition cannot be justified solely by rebranding children’s expression and association as ‘screen time’.[12]
General Comment No 25 also requires careful attention to businesses. A child’s rights can be impaired by private companies designing systems to maximise engagement, collecting data or amplifying dangerous material. The state has a legitimate protective role in requiring due diligence, safer products and access to effective remedies. The legal contest is therefore not between a rights-conscious approach which leaves technology companies alone and a protective approach which excludes children from their services. A third position is possible: require the companies to internalise the risks they create through mandatory design, age-appropriate functionality, governance and accountability, while retaining access wherever this can be achieved safely. Whether that is a sufficiently effective alternative to prohibition is an empirical and legal question, not a semantic one.[12]
The UNCRC has not been generally incorporated into United Kingdom law as a directly enforceable domestic code of rights. Its obligations remain important internationally and may inform the development and interpretation of domestic law, but the doctrinal route to judicial challenge ordinarily runs through the Human Rights Act 1998, the statutory powers being exercised, public-law principles and applicable data-protection requirements. The constitutional strength of a particular argument must not be exaggerated by assuming that every article of the Convention independently gives a child an immediately enforceable claim against a UK minister.[13]
The Human Rights Act: protection and access are both legally relevant
Article 8 of the European Convention on Human Rights (ECHR), incorporated through Schedule 1 to the Human Rights Act 1998 (HRA), protects respect for private and family life. For children it can encompass personal development, identity, relationships and aspects of communication. Article 10 protects freedom of expression, expressly including the freedom to receive and impart information and ideas. Article 11 protects peaceful assembly and association, and Article 14 prohibits unjustified discrimination in the enjoyment of Convention rights. The scope of a ban would therefore engage more than a putative ‘right to have a social media account’. The actual interferences would include the loss of particular channels for conversation, learning, self-expression and association, as well as the data processing demanded to determine eligibility.[14]
Articles 8 and 10 are qualified, rather than absolute, rights. Interference can be justified where it is prescribed by law, pursues a legitimate aim such as protection of health, morals or the rights of others, and is necessary in a democratic society. The protection of children against exploitation and serious harmful content readily satisfies the legitimate-aim stage in principle. The difficulty lies in whether the chosen restriction is sufficiently clear, evidence-based and proportionate. An intervention may pursue an overwhelmingly important objective and still extend beyond what the objective requires. The state must also acknowledge that, where age assurance affects adults, the rights of adults to access lawful information without intrusive checks become part of the balance.
Article 8 imposes positive as well as negative obligations. In K.U. v Finland, the European Court of Human Rights held that the state’s protective obligations were engaged where an online sexual advertisement involving a child exposed him to a serious threat to his private life, and inadequate legal mechanisms prevented the identification of the responsible person. The case supports the proposition that internet privacy is not preserved merely by abstaining from regulation: the state may need effective investigative and protective mechanisms against abuse. It does not, however, stand for the distinct proposition that children must be denied access to an entire class of lawful communications services. An obligation to protect from serious wrongs must be translated into an appropriately targeted means of doing so.[15]
At the other pole, Ahmet Yıldırım v Turkey and Cengiz and Others v Turkey emphasised the freedom-of-expression consequences of broad blocking measures affecting access to online material, including lawful information unrelated to the state concern which prompted the restrictions. These cases concern different factual and legal circumstances from a prospective child-specific minimum-age rule: they do not establish that any youth access restriction must fail. They nevertheless expose an important problem of overbreadth. A scheme aimed at high-risk interaction functions may suppress the educational, civic or expressive features of a mixed-purpose service unless it draws sufficiently precise boundaries.[16][17]
Conversely, the Grand Chamber’s judgment in Animal Defenders International v United Kingdom illustrates that a general legislative restriction may be upheld even where an individual application is difficult to defend in isolation. The Court gave weight to the carefully considered legislative rationale, the competing Convention interests and the quality of parliamentary scrutiny of a political-advertising restriction. The case concerns a different medium, objective and institutional history; it offers no automatic approval of social-media exclusion. Its significance is methodological: the strength of a general rule may depend upon rigorous consideration of alternatives, the scope of exceptions and the evidence underpinning the legislature’s judgment.[18]
Proportionality: applying the Bank Mellat framework
In Bank Mellat v HM Treasury (No 2) [2013] UKSC 39, the Supreme Court explained a structured approach to proportionality which asks, in substance, whether the objective is sufficiently important, whether the measure is rationally connected to it, whether a less intrusive but comparably effective measure could have been used, and whether a fair balance is struck between the interests of the individual and the community. The reasoning, notably Lord Sumption at paragraph 20 and Lord Reed at paragraph 74, is useful when assessing an under-16 ban. It does not allow a court to determine a future regulation’s compatibility in the abstract: the outcome will depend on the final statutory text, the evidence before the decision-maker, the design of safeguards and the applicable margin of judgment.[19]
First stage: is the objective sufficiently important?
There is little serious difficulty here. Protecting children from sexual exploitation, coercive contact, self-harm promotion, severe harassment and genuinely unsafe product design is a weighty public purpose. A legislative response can also seek to mitigate less immediate but consequential risks such as persistent sleep disruption or manipulative features engineered to prolong engagement. Yet the government should articulate which harms each measure targets. A ban adopted to address stranger grooming has a different causal justification from one adopted to address average daily screen time. If several rationales are placed together without distinguishing their evidential strength, it becomes too easy to invoke the clearest evidence for one harm to justify intervention against another. Judicial scrutiny, parliamentary oversight and later evaluation depend on a sufficiently precise account of the risks to be reduced.
Second stage: is excluding under-sixteens rationally connected to protection?
An age gate is capable of reducing direct exposure to particular platform features. If a service’s core functionality enables persistent recommendations, public visibility and unsolicited contact, preventing an underage account may reduce the opportunities for those interactions. That establishes a plausible rational connection, particularly in relation to risks whose occurrence depends on a child having access to a specific function. It does not establish that all online harms arise on the same platforms, that most children will cease to encounter the content without a personal account, or that use on an excluded service will necessarily be replaced by safer activity. Rationality is a lower threshold than demonstrated effectiveness, but the expected benefits still require a coherent account of mechanism and displacement.
There is a further temporal issue. Platforms can change rapidly: functions may be removed, recommender systems reconfigured, independent audit requirements imposed or new services created. A legislative model that permanently classifies all services of a particular brand as dangerous risks losing its factual foundation. A functional classification, capable of reassessment against specified criteria, is better connected to the actual source of harm. In the 2 November debate the minister indicated that platforms were not expected to be individually named in the initial regulations and that Ofcom would determine which met the criteria. This is potentially helpful, although legal certainty requires the criteria to be sufficiently clear to withstand strategic redesign and to avoid treating benign and high-risk services indiscriminately.[7]
Third stage: is there an equally effective, less intrusive measure?
This is the strongest challenge to a comprehensive service-level rule. Parliament already has statutory machinery under the Online Safety Act 2023 which requires regulated providers to assess and address children’s online safety risks. Ofcom’s child-safety regime has operated since July 2025. It includes duties relating to content harmful to children, the role of recommender systems and the use of effective age assurance in relevant circumstances. Further regulation might target unsolicited direct messages, livestreaming, recommendation of identified harmful content, public profiles, location sharing, late-night notifications, autoplay or algorithmic amplification without excluding a child from every lawful feature on the same service. The comparison is not between intervention and inaction; it is between regulatory designs capable of different degrees of risk reduction and interference.[20]
One counterargument has real force: these functions may be interconnected. Limiting the number of strangers who can message a child does not remove the incentives generated by public exposure; turning off recommendations may not solve harms created by direct contact; and a complex suite of controls may be difficult for young users, parents and regulators to monitor. Where the service is structurally organised around persistent recommendation, public visibility and high-pressure engagement, the government could reasonably conclude that restrictions imposed feature by feature are insufficient. Yet such a conclusion should be supported by identified failures in existing regulation and a credible comparison of alternative designs, rather than by treating the existence of some unsafe features as sufficient proof that no safer configuration is possible.
The government’s own treatment of older adolescents makes the necessity question harder. It proposes turning off autoplay and personalised feeds by default for sixteen- and seventeen-year-olds, allowing them to re-enable those features. It also cites an Ofcom experiment in which 95 per cent of participants across three age bands retained default safety settings on a simulated platform at the end of the study. This is not evidence that all real-world safeguards would have the same effect, especially when motivated users are exposed to a commercial interface. It does, however, provide a concrete reason to test whether enforceable, well-designed defaults for younger users could reduce substantial harm before ruling out access altogether. A legally convincing account of necessity should explain why the options considered adequate at sixteen are categorically inadequate at fifteen, taking account of the different capacities and risks on either side of the threshold.[2]
Fourth stage: does the measure strike a fair balance?
The fair-balance inquiry examines burdens that cannot be resolved by a simple finding that the intervention may work. A restriction can produce substantial protective benefit and still fall unfairly upon a minority who rely on a service for education, connection or support. It may also require adults to disclose identity information to access ordinary expression, create new risks through data breaches, or exclude lawful users wrongly identified as underage. These consequences should be weighed alongside benefits to children, families and society. An overall average benefit does not answer whether an avoidable, severe burden on a particular group is justified.
One must also distinguish the legal status of a proposed rule from its reviewability. Public authorities exercising statutory powers ordinarily must comply with Convention rights under section 6 HRA. Secondary legislation can be challenged on grounds including incompatibility with Convention rights, where the enabling statute does not require the incompatible provision, and on ordinary public-law grounds such as acting beyond statutory power. The HRA contains different remedial machinery for incompatible primary legislation, including declarations under section 4 rather than judicial invalidation of an Act of Parliament. The fact that the policy is expressed through affirmative regulations does not remove judicial supervision, but its statutory basis and possible remedies must be analysed accurately.[21]
What does the empirical evidence actually establish?
Research on children’s online lives has two quite different tasks. The first is to identify specific risks which require prevention, including grooming, harassment, exploitation and content encouraging dangerous behaviour. The second is to determine whether social media use as such causes deterioration in population-level wellbeing, and whether a particular restriction would improve outcomes. Evidence sufficient for the first task may be inadequate for the second. Conflating them risks creating a policy supported by serious accounts of individual injury but evaluated against broad measures such as happiness or screen time, neither of which captures whether the most dangerous encounters have declined.
The Department for Science, Innovation and Technology commissioned a research synthesis published in January 2026 which is unusually candid about the limitations of the existing evidence. It found a generally small association between more time spent on social media and poorer adolescent wellbeing, but cautioned against converting associations into causal claims. Only one of the seven reviews included in its evidence synthesis was assessed as high quality, and the underlying studies were frequently weak or inconsistent. The researchers found no relevant experimental studies in their supplementary review that had reduced social-media time among healthy adolescents to test the effect on their mental health. They stressed the importance of the activity undertaken, content encountered, circumstances of use and activities displaced, rather than treating time online as a uniform exposure.[22]
These findings support neither a confident prediction of universal harm nor an assertion that social media is harmless. Research uncertainty is not a moral licence to expose children to foreseeable sexual exploitation or coercion while awaiting an ideal randomised trial. Regulation of serious threats can be justified through evidence that is necessarily incomplete, particularly where the consequences are grave. The proper response is to keep distinct the evidence for acute identifiable harms, the evidence for more diffuse associations and the evidence that a particular intervention will reduce either. That separation enables a principled precautionary approach: act where risks are serious and plausible, preserve proportionality, and construct an evaluation capable of correcting mistakes.
For example, evidence that some children receive unwanted sexual messages establishes a serious problem of access, user identification, report handling and direct-contact controls. Evidence that high aggregate social-media use correlates with disturbed sleep may justify examining overnight design, notifications and persuasive features. Neither finding, without further reasoning, proves that a child should lose the ability to join an educational discussion forum or receive communications from a supervised youth organisation. The specificity of the identified risk should, wherever technically feasible, guide the specificity of the legal response. The contrary approach invites a mismatch between harms evidenced, restrictions enacted and outcomes measured.
What the UK pilot studies do—and do not—show
The government’s 2026 qualitative pilot involved 309 households with thirteen- to seventeen-year-olds, exploring interventions such as removal of social media apps, daily usage limits and overnight curfews over approximately four weeks. Many participants reported benefits in concentration, sleep, family interaction or calmness. Some also described social exclusion, frustration or ways around restrictions. The study yields useful evidence about acceptability, family adaptation and the practical operation of controls. Its design, however, does not establish that a statutory national ban would cause equivalent changes: participants’ allocation reflected preference rather than random assignment, the sample was small and the observation period short. Nor does a parent’s perception that a child seemed more present prove a sustained improvement in psychological health.[23]
Qualitative research can legitimately inform a regulator’s understanding of how a rule feels and how people respond. It should not be treated as an estimate of a population-wide treatment effect. There are also differences between a voluntary, negotiated household intervention and a binding restriction enforced by platforms against whole cohorts. Parents and children taking part in a short experiment may understand its temporary character, agree to try it or have access to other support. Under a compulsory ban, a child may be unable to negotiate an exception and may lose access permanently until sixteen. A robust evidence base needs longitudinal studies, variation across users and platforms, and attention to the outcomes of those who dislike, resist or cannot easily accommodate restrictions.
Public consultation, parental anxiety and the child’s own voice
The 2026 consultation attracted 116,211 responses, a substantial level of engagement which cannot be dismissed. The government’s June statement reported that nine out of ten parents who responded supported a minimum age of at least sixteen. Yet the same statement disclosed that only 19 per cent of the children who responded supported a minimum age of sixteen across all social media services; the responses from young people were more favourable to restrictions on some platforms or particular engagement features. The government subsequently emphasised that approximately two-thirds of young respondents supported age restrictions for at least some services. The three propositions are not equivalent. Support for controlling identified features does not logically translate into support for a universal platform-level ban.[5][6]
There is an additional methodological caveat. The official consultation methodology expressly states that consultations are not representative opinion polls: the sample is self-selecting, more motivated people are likely to participate and children and parents from lower-income backgrounds were underrepresented. The response total also includes organised campaign emails and different survey populations. Government commissioned a separate nationally representative survey, but figures from the self-selecting consultation cannot be presented as if they described the preferences of nine-tenths of all British parents, or all children. A decision-maker may legitimately attach weight to responses without turning the number of participants into a plebiscite on the precise instrument ultimately adopted.[24]
Under Article 12 UNCRC, consulting children is more than collecting comments before implementing an adult preference. It requires genuine consideration of their views, in light of their age and maturity, and a reasoned explanation when they are not followed. Children may understate harms, just as adults may underestimate the significance of online relationships; neither group’s views are conclusive. But a restriction made in children’s interests should identify the disagreement, evaluate it openly and explain why an objective risk judgment outweighs the expressed preference for less sweeping controls. Otherwise participation becomes procedural decoration rather than a rights-respecting method of policy formation.[11][12]
Harmful content, harmful contact and persuasive design are not the same legal problem
Three categories should remain conceptually distinct. Content harms arise when the material encountered promotes exploitation, suicide, self-injury, abuse or other harmful conduct. Contact harms arise from the behaviour of another person, including grooming, threatening communication or coercive demands for intimate images. Design harms arise from the system’s architecture: recommendations, visibility settings, location information, defaults, infinite scroll, notifications or incentives for prolonged use. These harms can intersect, but each has a different causal mechanism and potentially different effective intervention. A service-level ban captures them all only by removing access to the service, including its lawful and beneficial functions.
This distinction is also relevant to institutional responsibility. The perpetrator of sexual extortion commits wrongdoing, but the platform may have independently failed to design safe contact controls or respond to reports. A user might voluntarily seek distressing information, while an algorithm may amplify it in ways the user did not anticipate. A child may remain online late because friends are communicating, because notifications encourage repeated checking or because autoplay removes natural stopping points. Regulatory duties should follow the nature of the risk, including platform design and governance, rather than assign the entire burden of self-protection to a minor or assume that all solutions must take the form of exclusion.
In principle a systemic design solution can be more effective than a nominally child-friendly setting which a platform quietly undermines. An example would be a requirement that accounts known or reasonably believed to be used by children cannot be contacted by unknown adults, supplemented by robust reporting and investigation procedures. This targets a specific dangerous interaction and puts responsibility on the firm controlling the interface. On the other hand, if a service cannot reliably separate lawful from dangerous unsolicited contact or its commercial model depends on exposing users to a global audience, an outright age restriction on that service might become proportionate. The necessary finding concerns the inadequacy of safeguards in that service architecture, not the assumed incapacity of all children under sixteen.
The classification problem: what counts as social media?
The phrase ‘social media’ has no sufficiently precise meaning simply because it is familiar in ordinary speech. A video-hosting service may offer publicly accessible educational lectures, privately maintained subscriptions, public comments, algorithmic feeds, live broadcasts and direct contact. A messaging application may be primarily interpersonal yet include large public channels. A game may be chiefly recreational while facilitating voice conversations, public broadcasts and recommendations. An online forum may consist predominantly of moderated exchanges about health or study, but technically permit strangers to communicate. Deciding that a whole service belongs inside or outside a prohibition requires more than identifying whether users can upload content.
At the 2 November Commons debate, the minister referred to a working approach aimed at user-to-user services enabling online interaction and user posting alongside an algorithm. He also acknowledged that apparently non-social services can develop features resembling social media, citing the possibility of stranger conversations on a service better known for music streaming. The government did not propose naming individual platforms directly in the regulations; it contemplated service criteria and Ofcom supervision. These indications narrow the conceptual problem but do not settle it. Almost any platform with search ranking, recommendations and interactive comments could potentially share some of the characteristics. If the threshold is too broad, innocent information services may be affected; if too narrow, companies may change interfaces to evade the rule while retaining the relevant dangers.[7]
Legal certainty, part of the requirement that Convention interferences be prescribed by law, requires a person to be able to understand the relevant legal rule and its consequences with adequate foreseeability. A regulation which simply directs Ofcom to suppress ‘addictive social media’ would leave central choices to discretionary enforcement without a sufficiently intelligible standard. Better drafting would define relevant functionalities, their prevalence and configuration, the provider’s intended purpose, exposure to strangers, and interaction between recommendations and user-generated content. It could then set objective indicators or thresholds and allow evidence-based exclusion or exemption. Classification needs to be adaptable, but flexibility must be secured through legally constrained judgment rather than unlimited discretion.
Access to a platform does not necessarily mean access to every function
Three levels of restriction should be kept analytically separate. A service-level rule prevents an underage user from accessing a designated platform in the specified manner. A feature-level rule might permit the platform but disable public posting, uninvited messages or livestreaming. A content-level rule removes or restricts defined categories of harmful material. They differ in their consequences for lawful speech and in their enforceability. A service-level rule is comparatively simple to state but may be excessively broad. Content moderation can preserve access but is susceptible to error and does little about some contact or design hazards. Feature restrictions often target risks most directly but can be technically difficult to verify and may become obsolete as platforms evolve.
The policy’s treatment of YouTube illustrates the difficulty. A fifteen-year-old watching a documentary without an account is not in the same position as a fifteen-year-old participating in an unmoderated live discussion, although both may be using the same service. Similarly, a child whose school uses a public video account to distribute revision materials may lose functionality if the platform applies a blunt log-in block. The government says it wants to preserve education and high-quality content, but it will need a legally workable distinction between access to lawful material and participation in the combinations of features said to cause harm. A vague assurance that ‘educational content’ remains available is not a substitute for a tested pathway through which the child can actually obtain it.[8][2]
Nor should all platforms be presumed equally capable of providing age-segregated spaces. A small community forum may lack the resources to run sophisticated verification, while a large multinational platform may integrate assurance into its account architecture. There is an arguable public interest in ensuring that a company cannot escape obligations by remaining small, particularly where severe contact harms occur. Yet compliance demands should be proportionate to risk and scale. Regulation that makes lawful small services commercially unviable while leaving large incumbents with feasible but intrusive systems may inadvertently concentrate online public discourse within precisely the large platforms whose design practices prompted concern.
Gaming, messaging and the danger of moving rather than reducing risk
Gaming makes the case against crude technological categories especially clear. Many games allow children to cooperate in teams, maintain friendships, learn skills, create virtual worlds and communicate through text or voice. The existence of communication does not make every game equivalent to a public social-media feed; however, open voice chat, private messaging by unknown adults and livestreaming can create some of the same exploitation and harassment opportunities. The government’s ‘Australia-plus’ approach intends to reach dangerous functions in gaming without prohibiting gaming generally. That is more readily defensible in principle than treating all networked play as social media, provided that the functional controls are effective and do not unintentionally disable ordinary interaction with known friends.[7]
There is a difficult boundary between a stranger and a friend in an online game. A child may join a game with peers never met face to face but genuinely known through a school or supervised club. A predator may use a seemingly harmless multiplayer encounter to migrate the child to an encrypted messaging service; a simple rule preventing strangers from sending private messages could reduce that opportunity. But a restriction preventing any cooperative communication with an unfamiliar player could substantially diminish play, social participation and access for children who cannot easily participate in physical activities. The government’s eventual rule must determine whether anonymous match-making can remain while private communications are disabled, what ‘known’ means and what methods of verification would be proportionate.
Ofcom’s 2026 research indicates that online gaming and interaction with people known only online are already ordinary parts of childhood. It would therefore be unsafe to assume that children excluded from mainstream social media will move exclusively into supervised offline clubs. Some may migrate to multiplayer games, smaller forums, encrypted messaging networks or less visible services which offer similar content or contact opportunities but have weaker reporting and moderation. Such displacement does not mean regulation is futile. It means the assessment must measure not only reduced use of named platforms, but exposure to the underlying harms across the services children actually use after the intervention.[25]
Messaging exemptions create a related tension. Preserving family and friendship communications is essential to the government’s claim that the ban does not deprive children of the benefits of connectivity. However, a dedicated messaging service can also support contact from strangers, public channels and the circulation of harmful material. The distinction between a ‘safe’ messaging service and an ‘unsafe’ social platform cannot rest on branding alone. If the principal concern is unsolicited communication, a risk-based rule applicable across services may be more principled than a platform distinction that leaves similar functionality untouched merely because the service is conventionally described as messaging rather than social media.
AI companions and adjacent platforms
Conversational AI introduces a further category which does not fit the traditional model of user-to-user social media. A chatbot may simulate friendship, offer personal advice or encourage prolonged interaction without another human user producing the message. The government has announced additional measures concerning sexualised chatbot functions for under-eighteens and breaks during use by children, as well as consideration of stronger steps against dangerous advice. These proposals should not be mistakenly presented as automatically encompassed by the under-16 social-media ban. Their relevance is comparative: if particular harm comes from simulated intimacy, dependency or generated sexual content, extending the design-based regulatory principle to those features may be more coherent than determining legal responsibility by the category of the service provider.[2]
There is nevertheless a risk that a function-focused regime fragments into overlapping rules which are hard for families and providers to understand. A teenager may move through an online game, a public streaming channel, a chatbot and an encrypted conversation in a single evening. A coherent regulatory architecture should therefore connect identification of harm, age-appropriate safeguards, reporting routes and enforcement duties across categories. The aim is not to eliminate every risk in one place only to reproduce it in another. The criteria should remain intelligible even as technological labels change.
Age assurance: an essential mechanism with its own rights cost
Age assurance is not a technical footnote to the proposed policy. It is the mechanism through which the restriction will become effective, and the mechanism may itself substantially interfere with privacy and lawful expression. A rule prohibiting providers from serving under-sixteens is easy to announce but difficult to enforce if companies can rely on an unverified date of birth supplied at registration. Conversely, a system that compels every user to submit identity documents to multiple platforms may create a new infrastructure of personal-data exposure and exclude people unable or unwilling to comply. The legality of the scheme must therefore encompass both the restriction and the particular method used to identify those to whom it applies.
On 29 October 2026, Ofcom published a rapid assessment of age-assurance methods capable of being ‘highly effective’ at determining whether a user is at least sixteen. It identified four approaches: comparison with photographic identity documents, digital identity services, banking-based checks and facial age estimation. Each has benefits and limitations. The report is a technical assessment to inform future regulations and guidance, not a declaration that any one method is compulsory. It also explains that the UK’s proposed assurance standard is intended to be more demanding than the Australian model’s obligation to take reasonable steps. Ofcom recognises that circumvention cannot be eliminated and that its assessment does not settle all implementation choices.[26]
Photographic identity and digital credentials
Document-based checks can establish age with considerable certainty where the document is genuine and belongs to the user. Their weakness is inclusivity and data exposure. Ofcom estimated that 94 per cent of people aged sixteen or over have a passport or driving licence, but that approximately 150,000 sixteen- and seventeen-year-olds possess neither. If documentary proof becomes the practical default, a legal right to access a platform at sixteen may be frustrated for some young people by inability to prove eligibility promptly. The problem is especially serious where a service supplies educational information, essential peer support or opportunities for political participation which cannot easily be replicated elsewhere.[26]
A digital identity provider can reduce the burden of repeatedly producing an original document: once the age is checked, the provider may issue a reusable proof of a relevant attribute. Whether this meaningfully improves privacy depends on its design. A reusable credential which establishes only ‘over sixteen’ and prevents different platforms from correlating usage may minimise unnecessary disclosure. A centralised system which records every request, the identity of each relying platform and the user’s activity may create a surveillance risk quite different from a one-off age check. The appropriate requirement is to prove the minimum fact necessary for lawful access rather than construct a general log of where citizens seek information.
Bank-based checks and the exclusion problem
Banking-based verification may confirm an age already established during account opening. Ofcom cites evidence that approximately 90 per cent of sixteen- and seventeen-year-olds hold a bank account. This makes the method useful but not universal; the remaining group is not negligible, and a person may have a bank account which is unsuitable for a third-party verification service. It would be hard to justify a scheme that turns a commercial banking relationship into a de facto condition for lawful expression, especially for young people with precarious family circumstances, limited documentation or barriers to financial services. Multiple verification routes and accessible challenge mechanisms are therefore integral to proportionality, not optional customer-service enhancements.[26]
Facial age estimation and borderline ages
Facial age estimation uses images and statistical inference to approximate a person’s age. It is attractive to providers because it may avoid a document check for some users. But the nearer the user is to the threshold, the more significant an estimation error becomes. Ofcom acknowledges that some technologies have margins of error of several years around age sixteen and envisages a ‘challenge age’ approach, with a second method used where necessary. Accuracy also needs examination across demographic groups, device quality, disability and differing conditions of presentation. A system which wrongly rejects a sixteen-year-old does not become fair merely because its average accuracy across all users is high. Equally, the failure to reject a fifteen-year-old affects the degree of protection afforded to children.[26]
The data-protection classification of facial processing requires care. A facial photograph is personal data where an individual can be identified. Certain biometric data processed for the purpose of uniquely identifying a natural person attract special-category protection under Article 9 UK GDPR. It would be inaccurate to assert that every facial age-estimation exercise automatically constitutes the same category of biometric identification: the purpose and technical operation matter. However, even where Article 9 is not triggered, facial images and inferred age information can be highly sensitive in practice. Their collection, retention, transmission and security must be justified and limited to what is necessary for the relevant assurance purpose.[27]
Ofcom further distinguishes age inference based on a user’s behaviour from an initial access check. Behavioural inference presupposes that the person has already used the platform, so it cannot alone operate as the gate which prevents an underage user entering. It might help to identify accounts that bypassed a previous check. That potential use creates a further privacy issue: if services routinely analyse the behaviour, social connections and communications of users to estimate whether they are underage, effective enforcement may depend on profiling that is itself intrusive. Lawmakers need to specify what surveillance is permitted in the name of detecting circumvention, and how it will be constrained to avoid generalised monitoring.[26]
The UK GDPR and the Children’s Code: more than a request for consent
The UK GDPR requires processing to have a lawful basis and comply with principles including fairness, transparency, purpose limitation, data minimisation, accuracy, storage limitation and security. The collection of identification documents or facial images is not rendered lawful simply because Parliament approves a social-media restriction. Providers and independent age-assurance vendors will need to identify their roles and lawful bases, explain processing in a way children can understand, limit onward disclosure, avoid unnecessary retention and undertake data protection impact assessments where the statutory conditions for high-risk processing are met. Where a platform genuinely requires only an answer to whether a user meets an age threshold, disclosure of a full name, date of birth or identity document to the platform may be difficult to reconcile with data minimisation unless separately justified.[10][27]
The Information Commissioner’s Age Appropriate Design Code, often called the Children’s Code, supplies an existing framework for online services likely to be accessed by children. Its approach is risk-based and grounded in the child’s best interests, privacy by default and proportionate age assurance. The point of legal significance is that a safety intervention may create an additional data-processing risk: checking age may protect a child against unsafe material while simultaneously exposing the child or their family to overcollection, tracking or discrimination. These are not unrelated objections raised by privacy advocates after the policy is devised; they are aspects of the statutory design question. A rights-respecting age gate should be judged by the amount of sensitive information it requires, the number of entities that receive it, the possibility of linking separate checks and the availability of deletion or effective redress.[28]
The law governing significant solely automated decisions also needs to be stated in its current form. Section 80 of the Data (Use and Access) Act 2025, brought into force in February 2026, replaced the previous UK GDPR Article 22 framework with Articles 22A–22D. These provisions address significant decisions based entirely on automated processing, including applicable restrictions and safeguards. Consequently, a contemporary account of wrongful age-gate refusals should not repeat the earlier formulation of Article 22 as though it were unchanged. Whether a particular decision falls within the statutory scheme depends on its legal or similarly significant effects, whether meaningful human involvement exists, the data used and other conditions. Regardless of that classification, an accessible review process, meaningful explanation and correction of errors are strong requirements of fair regulatory design.[29][30]
Privacy risks also arise for adults. Ofcom has indicated that some existing adult account holders may not need an additional check where reliable information is already available, but many adults are likely to be asked to demonstrate age. If lawful access to political debate, journalism or support forums is conditioned on obtaining a digital credential, the interference is not restricted to children. Articles 8 and 10 ECHR protect adults as well. A data breach or a record linking identity to particular services can have consequences for journalists, whistleblowers, domestic-abuse survivors and others using sensitive channels. The policy must account for this wider audience; the assertion that adults remain legally eligible does not by itself answer the privacy cost of proving eligibility.[26]
Equality, disability and the unequal value of online access
The argument that children can simply replace social media with offline friendship is least persuasive when examined against unequal circumstances. A child living in an isolated rural community, a disabled young person for whom travel is difficult, a child experiencing bullying at school or one seeking confidential information about sexuality may have fewer realistic alternatives to online connection than a child with extensive local opportunities. A child with a long-term medical condition may rely on moderated patient communities; another may seek advice without revealing a sensitive concern to family members. These examples do not establish an unqualified right to every platform. They demonstrate why a universal rule must be tested against the actual distribution of its burdens rather than an assumed typical childhood.
General Comment No 25 expressly recognises that digital interaction can be particularly important for disadvantaged or vulnerable children wishing to communicate with others who share their experiences. It also warns against digital inequalities which exacerbate pre-existing exclusion. A ban which removes access from everyone may appear formally equal, yet produce different substantive effects according to disability, family resources, location or the availability of alternative support. The relevant comparison is therefore not simply between equal treatment and parental discretion. It is between the proposed rule and an accessible design that provides comparable protection while maintaining meaningful opportunities for children whose circumstances make access unusually important.[12]
Article 14 ECHR may become relevant where the interference with Article 8 or 10 rights has a discriminatory effect which lacks sufficient justification. Domestic equality law also requires precision: section 28 of the Equality Act 2010 excludes the protected characteristic of age in relation to persons under eighteen from the ordinary services provisions in Part 3, so the general services age-discrimination regime is not a simple substitute for a children’s human-rights challenge. But disability discrimination and duties concerning reasonable adjustments are different matters and can be relevant to the design and accessibility of services. A statutory prohibition cannot necessarily be avoided through an individual adjustment which would directly contradict it; this makes it all the more important for regulation-makers to address predictable disadvantage at the design stage.[31]
A child could also be formally protected yet less safe in practice if access to moderated mainstream communities is lost and the child turns to obscure or unregulated groups. The government should therefore consider whether carefully specified exemptions, purpose-limited access, verified youth organisations or youth-safe service models are legally and operationally viable. Exceptions must not be so open-ended that they defeat the restriction or require children to disclose intimate vulnerabilities to gain entry. An exception which demanded proof of a child’s medical diagnosis or sexual identity in order to access a support service would introduce privacy risks of its own. The better answer may be certified safe functions or providers rather than a personal exemption based on disclosure of sensitive attributes.
Australia as a natural experiment, not a conclusive precedent
Australia’s social-media minimum-age regime began operating on 10 December 2025. It generally places obligations upon designated age-restricted social-media providers to take reasonable steps to prevent Australians under sixteen from holding accounts, rather than imposing offences on children or their parents. Dedicated messaging and online gaming services are excluded from its central account restriction, which is one reason the United Kingdom describes its wider feature-specific proposals as ‘Australia-plus’. The Australian experience is valuable because it demonstrates the difficulties of compliance, account identification and displacement under an actual age-based scheme. It does not determine what the ECHR requires in Britain, and the UK proposes a stricter standard of age assurance than Australia’s reasonable-steps test.[32][26]
Australia’s eSafety Commissioner published initial evaluation findings in July 2026 after a three-month follow-up. Among children aged ten to fifteen, reported ownership of an account on at least one restricted platform fell from approximately 52 per cent before implementation to 42 per cent. Yet the proportion who reported using one or more restricted platforms fell more modestly, from almost 86 per cent to over 81 per cent. These figures reveal an essential distinction between account possession and actual exposure. A rule can reduce the number of underage accounts without eliminating access to public videos, shared accounts or services which do not require registration. Evaluations which report only account closures may substantially overstate the reduction in children’s engagement with the underlying platforms.[33]
In the same evaluation, the regulator found that many children continued to hold accounts because services had not yet checked their age. Approximately half of those retaining accounts identified that reason. More sophisticated workarounds, such as virtual private networks or use of another person’s documents, were reported by relatively small proportions of children at that early stage. This finding is important because the familiar claim that children will simply outsmart all restrictions is too crude. Early non-compliance may be chiefly a question of incomplete industry implementation, not the irresistible ingenuity of teenagers. The UK could improve compliance through effective assurance and auditing. That does not answer whether the resulting level of exclusion is proportionate, but it prevents a simplistic inference that any minimum-age rule is inherently unenforceable.[33]
The early Australian evidence also raises concerns about parents’ knowledge of children’s use and substitution into other services. The evaluation describes an increase in the share of parents unaware that their child was using a restricted service, and some movement towards messaging and gaming. Those observations do not establish that the policy caused all displacement or that every substitute is more dangerous. They do suggest that the regulatory objective should be assessed by exposure to specified harms and the quality of protective relationships, not merely by changes in visible account numbers. A reduction in officially registered social-media users accompanied by greater use of obscure services would be a very different result from a reduction in grooming, bullying and sleep disturbance accompanied by maintained access to educational content.
Nor is a three-month follow-up sufficient to establish long-term mental-health outcomes, the permanence of behavioural changes or the consequences for social development. The Australian study gives a useful initial indication of platform implementation and family adaptation, not an experimental estimate proving that banning accounts improves population wellbeing. For UK purposes the lesson is institutional as much as empirical: an enforcement regime needs reliable baseline measures, a distinction between account and platform use, independent access to relevant data and published measures of both benefits and unintended consequences. The government’s planned reforms should be judged after implementation against that broader set of criteria, rather than against the success of launching the restriction by a politically promised date.[33][34]
Parental responsibility and the child’s developing autonomy
One possible objection to state intervention is that parents, not ministers, should decide when their children can use social media. That argument has intuitive force in ordinary family life but cannot be an absolute constitutional rule. The state already regulates products, education, employment and activities where children’s vulnerability or third-party conduct creates significant risks. Article 5 UNCRC recognises parental direction in the context of evolving capacities; it does not authorise the state to ignore abusive commercial systems because a parent can theoretically supervise the child’s account. Families differ markedly in available time, digital knowledge, resources and their capacity to negotiate with platform design. A system which relies only on parental monitoring may leave unprotected precisely those children whose carers are least able to provide it.
The opposite claim—that a single statutory age threshold should replace all parental judgment—is also difficult to defend without qualification. Children under sixteen vary greatly in maturity and life circumstances, and parents may have reasoned views about access to a particular moderated or educational service. Removing parental consent as a route to access would place responsibility on the provider and reduce the pressure on families to bargain with powerful companies; it would also disallow informed individual decisions even where a service is demonstrably safe for the child’s use. The proportionality inquiry should ask whether this loss of flexibility is necessary to avoid under-enforcement, social pressure or uncertainty, rather than presuppose that parental discretion must either dominate or disappear.
Age thresholds are not automatically arbitrary. Law routinely uses them to draw administrable boundaries even though a birthday does not transform personal maturity overnight. There can be good reasons to prefer a clear rule to a discretionary maturity assessment that is intrusive, inconsistent or costly. Yet the choice of sixteen rather than fourteen or fifteen requires a reason connected to evidence about risk, capacity and regulatory practicability. The government’s own acknowledgment that sixteen- and seventeen-year-olds should retain control over default settings demonstrates that autonomy develops progressively; it does not explain by itself why a fifteen-year-old must be excluded from every covered service while a person just over sixteen may enable high-risk features. Graduated rights and safety protections are more coherent with evolving capacities than a sharp binary without a reasoned transition.
A carefully constructed transitional regime could preserve some supervised or limited access before sixteen, use secure default configurations at sixteen and progressively expand control thereafter. But that, too, presents challenges: supervised access may exclude children whose parents are unavailable or unsafe, and discretionary access based on apparent maturity may require sensitive profiling. The legal task is not to idealise personalised exceptions. It is to compare the real administrative and rights costs of a bright line with those of workable functional alternatives, and to explain openly why the adopted model provides the fairest overall balance.
What legal obligations will technology companies retain?
The proposed reforms allocate the front-line burden to providers rather than criminalising children. That is a significant and generally defensible allocation of responsibility. The company controls registration, ranking systems, public visibility, messaging architecture, user reporting and many aspects of data processing. Parents and children may influence choices, but they cannot individually redesign the service or audit its internal safety systems. The Online Safety Act already makes regulated firms responsible for duties concerning illegal content and the protection of children. The anticipated regulations would add specific access and functionality restrictions, enforced through the wider regulatory framework. They would not turn a child who circumvented an age check into an offender merely for accessing the service.[20][32]
Legal accountability requires more than age-verification screens. Providers should be able to demonstrate how they classify accounts and features, measure false acceptance and rejection rates, remove or disable underage access, prevent re-entry where required, preserve relevant evidence and resolve disputes. Regulatory guidance will need to avoid creating incentives for indiscriminate collection of sensitive documents or covert monitoring. The duty should be to operate an effective, lawful and auditable system, not merely to purchase an assurance product and declare compliance. Outsourcing a check to a specialist vendor does not remove the need to consider privacy, accuracy, security and the provider’s wider statutory obligations.
There is a further question about smaller services and online marketplaces. A platform may host several distinct functions through third-party applications, public channels, embedded video, private messages or live interactions. The regulations need to specify who is responsible for controls when access is mediated through another business and what evidence demonstrates compliance. An application store or device manufacturer may assist age assurance, but Ofcom’s October analysis notes that app stores are not presently regulated under the Online Safety Act simply in that capacity. Extending obligations to them would require a distinct legal basis and a careful allocation of responsibilities. No court should be asked to infer a new statutory duty merely because a technology actor could contribute practically to a safer system.[26]
Enforcement also presents a risk of regulatory asymmetry. Well-resourced companies may be better able to implement sophisticated checking and challenge regulatory decisions, while smaller moderated communities may struggle with compliance despite posing lower levels of risk. Conversely, a platform with a large youth audience and particularly risky design should not evade obligations because age-checking is expensive. A fair enforcement strategy should align costs and duties with the risk of particular functions, demand evidence of actual effectiveness and preserve the possibility that a service can change its design and thereby cease to satisfy the criteria for a prohibition.
Five practical situations demonstrate why safeguards matter
A fifteen-year-old seeking confidential support
Consider a fifteen-year-old who has limited safe support at home and participates in a moderated online community dealing with a sensitive health or identity issue. A blanket exclusion from a large platform might terminate an important relationship with peers or support organisations even if the particular community has strong safeguards. A lawful policy need not grant this child unrestricted access to public recommendations, livestreams or strangers’ messages. It should nevertheless examine whether the support can remain accessible through a safer interface, independent moderated site or well-designed exemption. The seriousness of the child’s need should influence the fair-balance inquiry, even if no special individual exception is ultimately justified.
A fourteen-year-old contacted through a game
A fourteen-year-old uses a multiplayer game and receives an unsolicited private message from an adult. A prohibition confined to conventional social networks may provide little direct protection; a targeted restriction on messaging by unknown adults could address the risk. Yet a ban on all stranger interaction in the game might remove the cooperative dimension which makes it valuable. The regulatory question is whether access can be redesigned to retain the child’s play and known-peer contact while preventing the specific avenue for grooming. This example supports the government’s recognition of gaming features but also shows why delayed feature regulation should be assessed alongside the initial social-media prohibition.
A sixteen-year-old wrongly rejected by an age check
A person who has just turned sixteen is refused access when a facial estimator gives an age of fourteen. They have no passport, driving licence or suitable bank verification route. Legally the proposed age threshold would permit access, but the technical system denies it. An internal appeal requiring another unobtainable identity document would be illusory. Fair implementation demands independent verification alternatives, timely correction, accessible explanations and a means to challenge repeated erroneous rejection without surrendering more personal data than necessary. Ofcom’s published estimate of teenagers lacking photo identification makes this a foreseeable class of cases rather than an exceptional hypothetical.[26]
An adult seeking anonymous political information
An adult wishes to view a lawful political discussion but does not want a platform to link their legal identity to their browsing. If an age gate requires a passport scan retained by the service, the ban has an additional Article 8 and 10 consequence for the adult. A privacy-preserving proof that the visitor is above the threshold may satisfy the protective objective with substantially less interference. The example demonstrates why a policy ostensibly directed only at children can raise a general freedom-of-expression question through its enforcement architecture.
A provider removes the dangerous feature
Suppose a service formerly allowed public live broadcasts and algorithmic recommendations to strangers but redesigns itself to offer chronological educational videos and restricted interaction with verified youth groups. If the law continues to bar all under-sixteens solely because the service once qualified as social media, the rational connection between risk and prohibition weakens. A mechanism for reassessment on objective evidence would improve both proportionality and incentives: providers would have a commercial reason to design safer services rather than accept that young users must remain excluded regardless of improvement.
How could the scheme be made more compatible with children’s rights?
The case for regulating unsafe online environments is strong. The case for treating service-level exclusion as the only serious form of regulation is not. A credible scheme should incorporate the following principles from the outset, through legislation, Ofcom guidance and transparent supervision, rather than treating them as reforms to be considered only after harm or litigation occurs.
First, publish a sufficiently precise and reviewable definition of covered services. The criteria should identify the combination of functionalities and structural risks which justify restriction, distinguishing a hazardous recommender-based interaction platform from a moderated service or a repository of educational material. Providers should be able to obtain reasoned classification decisions, and there should be a pathway for reassessment when risky functions are removed. The government should explain the relationships between the first regulations and later feature-specific controls so that neither a regulatory gap nor an unnecessary overlap is concealed by the timetable.
Secondly, establish the evidence for necessity rather than rely exclusively on parental preference. An explanatory assessment should distinguish criminal exploitation, unwanted contact, harmful content, sleep and attention outcomes, and general mental-health claims. For each category it should state what the existing Online Safety Act regime has achieved, what gaps remain, which narrower interventions were considered and why they were deemed insufficient. Where the evidence is uncertain, the analysis should say so, including the possibility that a precautionary restriction is justified by severity despite incomplete causal estimates. Accuracy about uncertainty strengthens the legitimacy of public decision-making.
Thirdly, design age assurance around minimal disclosure and accessible proof. Verification should ordinarily communicate only the fact needed for eligibility, with narrowly limited retention and third-party access. The system should use multiple effective methods, make accessible alternatives available to users without photographic ID or bank accounts, test demographic error rates and provide a rapid correction mechanism. It should account for persons whose safety depends on maintaining practical anonymity. A choice between unsafe access and intrusive surveillance would undermine the policy’s claim to protect children’s rights.
Fourthly, maintain genuine pathways to beneficial content and participation. Government assurances about education, news, health support and communication with known friends need operational form. This could involve functionality-specific access, genuinely moderated age-appropriate services, support for youth organisations and clear rules for public material viewable without an account. Exemptions should attach wherever possible to objectively safe services or functions rather than force children to disclose sensitive medical, family or identity information to qualify. The interests of disabled, isolated and vulnerable children should be considered explicitly in the assessment of discriminatory impact.
Fifthly, hear children and respond to disagreement. Article 12 UNCRC and the principles reflected in General Comment No 25 require more than a headline about consultation numbers. The government and Ofcom should continue structured engagement across ages, disabilities and socio-economic backgrounds, publish the rationale for rejecting less restrictive approaches where relevant, and explain new rules in language children understand. Particular care is required where consultation responses indicate support for feature restrictions but opposition to excluding every platform.
Sixthly, mandate independent evaluation and genuine opportunities to revise the regime. The implementation should begin with publicly specified outcome indicators: prevalence of sexual exploitation and grooming, unwanted contacts, harmful-content encounters, reported sleep disturbance, access to support and education, erroneous age-gate decisions, data breaches, workarounds and migration to other services. Measurements should distinguish actual use from account ownership. Findings should be broken down where lawful and feasible by age, disability and other relevant characteristics without requiring unnecessary personal profiling. A review date, publication obligation and mechanism to narrow or expand rules in response to robust evidence would turn the policy into a monitored intervention rather than a permanent assertion of certainty.
These safeguards do not eliminate the possibility that broad age restrictions will be necessary for platforms which remain structurally unsafe. Instead they help establish the conditions under which that conclusion would be defensible. They also recognise a basic institutional lesson from the Australian evaluation: if success is defined only as removing accounts, the system may miss continued exposure, displacement and secondary harms. A rights-sensitive regulator should measure whether children are safer and able to participate meaningfully, not simply whether an age gate can produce a high rejection rate.
Conclusion
There is no necessary contradiction between children’s rights and a statutory restriction on access to dangerous online services. The state may have positive duties to protect children from serious abuse and is entitled to require technology companies to address the dangers created by their products. Children do not possess an unqualified Convention entitlement to use every social-media platform regardless of age or risk. Equally, their vulnerability does not extinguish their rights to expression, association, private life, access to information and progressively greater autonomy. Those rights must be included in the legal justification, not considered only after a protective decision has been taken.
The United Kingdom’s proposal remains legally unfinished as at 6 November 2026. Section 214A of the Online Safety Act supplies a substantial regulation-making power, but the operative scheme will depend on the regulations, definitions, exceptions and standards still to be settled. On the available material, the strongest justification concerns serious, specific harms from particular combinations of contact, content and design features. The weakest is an undifferentiated inference that every kind of social-media use causes poor mental health and that excluding every fifteen-year-old is necessarily the least intrusive effective response. The government’s scientific review, qualitative pilots, consultation results and Australia’s early experience do not support that simple inference, although they provide real reasons for urgent and enforceable safeguards.
The decisive legal test will therefore be the substance rather than the label of the reform. A proportionate regime must demonstrate why the selected services are covered, why the threshold of sixteen is justified, why less restrictive safety design will not suffice in the affected settings, how lawful and beneficial access will be preserved, and how the costs of surveillance and mistaken exclusion will be controlled. The strongest version of the legislation would treat a service-level ban as a justified last resort for demonstrably high-risk configurations, while imposing robust, auditable controls on dangerous features across the broader digital environment. That approach gives practical meaning to the proposition that protecting children and respecting their rights are complementary duties. A scheme which merely removes children from visible platforms, without proving that it reduces the underlying harm or preserving meaningful participation, risks sacrificing one set of children’s rights in the name of another.
References and primary sources
Numbered links lead to legislation, court judgments, intergovernmental materials, regulators’ publications and official government research. Policy announcements and consultations are identified as such; they are not treated as enacted restrictions. This essay reflects the position on 6 November 2026.
- Department for Science, Innovation and Technology, Growing up in the online world: progress statement (15 June 2026), including the announced minimum age, its proposed rationale, the children and parent consultation response figures, the planned affirmative regulations and expected spring 2027 implementation.
- Department for Science, Innovation and Technology, Growing up in the online world: government response (15 July 2026), particularly the proposed 16–17 defaults, wider functional restrictions, AI chatbot measures, consultation evidence and discussion of the UK intervention pilots.
- House of Commons Library, Proposals to ban social media for children (research briefing updated 3 November 2026), for the legislative chronology and distinction between announced measures and forthcoming regulations.
- Children’s Wellbeing and Schools Act 2026, s 70, inserting Online Safety Act 2023, s 214A (power to prevent or restrict children’s access to internet services and particular features, including the requirement to consider differences in features and children’s ages).
- Children’s Wellbeing and Schools Act 2026, s 71 (ministerial progress statement in relation to first regulations).
- Online Safety Act 2023, s 225(1)(ea), as amended, and the June 2026 progress statement (affirmative parliamentary procedure and proposed timetable).
- Hansard, House of Commons, Social media: ban for under-16s (2 November 2026), especially the responsible minister’s explanation of platform criteria, Christmas-recess regulations, proposed Ofcom implementation, staging of gaming-feature measures and the unresolved question of re-entry for redesigned safer services.
- UK Government, Fact sheet: new rules to protect children online (updated 18 September 2026), for illustrative platforms, exclusion of dedicated messaging services, planned gaming protections, access to educational material and proposed changes for older teenagers.
- Children’s Wellbeing and Schools Act 2026, s 72 (delegated power concerning the age applicable to consent for the processing of personal data in relation to information society services).
- Data Protection Act 2018, s 9, together with the official explanatory notes (thirteen-year threshold where an information society service relies on the child’s consent under Article 8 UK GDPR; distinct from the general minimum age for use of a service).
- United Nations, Convention on the Rights of the Child (1989), especially arts 3, 5, 12, 13, 15, 16, 17, 19, 31 and 34.
- UN Committee on the Rights of the Child, General comment No 25 (2021) on children’s rights in relation to the digital environment, UN Doc CRC/C/GC/25, particularly paras 9–13, 17–23, 35–42, 55–65 and those addressing participation, proportionality, privacy and business responsibility.
- Human Rights Act 1998 (domestic Convention-rights framework); the UNCRC is a distinct international treaty and has not been enacted as a generally directly enforceable statutory code throughout the UK.
- Human Rights Act 1998, Sch 1, ECHR arts 8, 10, 11 and 14 (private and family life, expression and receipt of information, association and non-discrimination).
- K.U. v Finland, App no 2872/02 (ECtHR, 2 December 2008), on the positive obligation to provide effective protection of a child’s private life against a serious threat arising from an internet advertisement.
- Ahmet Yıldırım v Turkey, App no 3111/10 (ECtHR, 18 December 2012), on disproportionate restrictions affecting access to online expression as a consequence of broad website blocking.
- Cengiz and Others v Turkey, App nos 48226/10 and 14027/11 (ECtHR, 1 December 2015), on the Article 10 implications of wholesale blocking of YouTube.
- Animal Defenders International v United Kingdom, App no 48876/08 (ECtHR Grand Chamber, 22 April 2013), on the proportionality of a general legislative restriction after close consideration of competing interests and parliamentary safeguards.
- Bank Mellat v HM Treasury (No 2) [2013] UKSC 39, especially Lord Sumption at [20] and Lord Reed at [74] (structured proportionality analysis).
- Online Safety Act 2023, especially existing child-safety provisions; Ofcom, Protection of children duties under the Online Safety Act (current regulatory guidance).
- Human Rights Act 1998, s 6 (public authorities and Convention rights) and s 4 (declarations of incompatibility in specified circumstances).
- Department for Science, Innovation and Technology, Understanding the impact of smartphones and social media on children and young people: executive summary (January 2026), especially the quality of systematic reviews, limits of causal inference and research priorities.
- Department for Science, Innovation and Technology, Social media intervention research 2026: qualitative pilots with 13- to 17-year-olds (July 2026), including study allocation, duration, perceived benefits, isolation and circumvention.
- Department for Science, Innovation and Technology, Summary of evidence, methodology, and organisations who responded to the consultation (July 2026), specifically the limits of self-selecting consultation responses and the distinction from representative survey evidence.
- Ofcom, Younger phone owners, the rise of AI, and consumption over creation: our latest look at UK children’s media lives (2026), supported by its children and parents media-use research.
- Ofcom, How age checks at age 16 can work to underpin UK’s social media ban (29 October 2026), and the underlying rapid assessment (four proposed effective methods, coverage, error and privacy considerations).
- Information Commissioner’s Office, Age assurance for the Children’s Code: expectations for age assurance and data-protection compliance (guidance on fairness, data minimisation, privacy and proportionality).
- Information Commissioner’s Office, Age appropriate design: a code of practice for online services, particularly the age-appropriate application standard and the risk-based approach to children’s personal data.
- Data (Use and Access) Act 2025, s 80 (amending UK GDPR provisions on automated decisions, including Articles 22A–22D); see the commencement instrument, SI 2026/82 (relevant commencement on 5 February 2026).
- Information Commissioner’s Office, The Data (Use and Access) Act 2025: summary of changes to data-protection law (updated June 2026; guidance confirming data-protection provisions now in force).
- Equality Act 2010, s 28 and s 29; see the official explanatory notes on the under-eighteen age exclusion from Part 3.
- Australian eSafety Commissioner, Social media age restrictions (implementation from 10 December 2025, providers’ reasonable-steps responsibilities and exceptions for messaging and gaming).
- Australian eSafety Commissioner, Early days, early insights: understanding experiences of social media age restrictions at the three-month follow-up (July 2026), especially account ownership, actual platform usage, reported reasons for non-compliance and the short evaluation horizon.
- Australian eSafety Commissioner, Social media minimum age: compliance update (March 2026), concerning early provider implementation and remaining account access.