Banning under-16s from social media does not reach what drove the riots: the harm was adult-made and amplified by design…

The government has made an age limit its answer to a system it says is failing children. But the social-media harm on display in Belfast over the past week was made by adults and amplified by platform design, and an age gate reaches neither. The danger is not that the ban is wrong, but that restricting who may use these platforms will be mistaken for regulating how they rank and spread content.

Keir Starmer has announced that under-16s will be banned from social media, with regulations the government hopes Parliament will approve before Christmas and the ban taking effect by spring 2027 if Parliament agrees. It will follow the Australian model, covering user-to-user platforms built around social interaction and posting, including Snapchat, TikTok, YouTube, Instagram, Facebook, and X, though not messaging services such as WhatsApp and Signal. The government says it will go beyond a blanket ban by switching off functions such as livestreaming and contact from strangers for under-16s across a wider range of services, including gaming, with those restrictions left on by default for sixteen- and seventeen-year-olds, and with curfews and breaks in infinite scrolling for under-18s still to be set out in July. The Prime Minister made his case in the language of childhood: social media is making children unhappy, is “designed to be addictive”, and is not a safe environment, so a full ban is, he said, “the right choice”. The consultation behind it drew more than 116,000 responses, with ninety per cent of parents backing a minimum age of sixteen.

I want to argue that, measured against the disorder of the past week, this is aimed at the wrong target. The problem is not that age restrictions address no real harm; children do face grooming, compulsive design and self-harm material, and a well-built age policy might reduce some of it. The problem is that the most serious public-order harms on this platform system are driven by adults, amplified by design, and left almost untouched by who is allowed to hold an account. The Prime Minister made his case entirely in the language of childhood and addiction, with not a word about the riots, the misinformation or the amplification that has run through Belfast over the past week. That disorder, the third such episode in two years, is the test this policy has to meet, and read against it, the ban changes who may use the platforms but leaves their amplifying design in place.

The people who drove it were adults

The most consequential amplification identified in the available reporting came from adults with large audiences. The current Belfast unrest followed a knife attack on the evening of 8 June, for which a Sudanese man was charged with attempted murder. Within hours, footage and a parallel stream of false claims, including that the victim was a child who had died, spread across Facebook, X and TikTok. Tommy Robinson posted a list of protest locations across Great Britain and Northern Ireland; Elon Musk reshared his call to protest; and Rupert Lowe pledged mass deportations and the return of the death penalty against those he called “barbarians”. Anonymous WhatsApp messages, on a service the ban explicitly leaves out, told men aged eighteen and over to be ready to fight, and a long list of home addresses said to house immigrants was circulated, which police said was putting lives at risk. The identifiable actors in this chain were adults, and the messages themselves were directed principally at adults. The ban nevertheless makes children the population directly regulated.

The 2025 Ballymena violence followed the same pattern. Loyalist Facebook pages shared protest details and false claims, rioters livestreamed arson, and a 32-year-old man was charged with sending menacing communications and with encouraging or assisting offences. One of the sharpest controversies of that episode concerned not a teenager but a minister: the Communities Minister, Gordon Lyons, was accused of inflaming tensions over a Facebook post about people moved to Larne Leisure Centre, which was later set alight, though he denied wrongdoing and said police had encouraged elected representatives to share the information. The merits of that dispute are for others to decide, but the basic point stands. When controversy can centre on the social-media conduct of an elected minister, excluding fifteen-year-olds from Instagram plainly does not reach the principal actors involved.

The origins of the present cycle make the same point. Amnesty International’s study of the 2024 riots records that the Southport murders were committed by a UK-born seventeen-year-old, and that false online claims recasting him as a Muslim asylum-seeker who had arrived by boat helped trigger and fuel the violence that followed. The precipitating harm was misinformation, generated and amplified principally by adults, with a minor at its centre, falsely described as the perpetrator rather than as a young user an age gate would catch.

The mechanism was amplification, not access

Amnesty’s analysis sets out how this works, using X’s publicly released 2023 ranking model. The system promotes posts based on predicted engagement rather than accuracy, rewarding replies and sustained argument far more than passive approval, while its main check, a user report, applies only after someone reports a post that may already have spread. Premium subscribers receive an additional ranking advantage, and Amnesty notes that several of the prominent accounts it examines, including those of Musk, Robinson, and Tate, carry paid verification badges. Amnesty is careful about the limits of this: the weights describe the 2023 public code rather than the live system across 2024 to 2026, and it accepts that crisis-period activity and ordinary resharing, not the algorithm alone, drove some of the reach. Even with those caveats, the reach it documents extended far beyond ordinary follower networks, with false posts about the Southport attacker’s identity gathering tens of millions of impressions within a day.

The case of Lucy Connolly shows the gap between this harm and the proposed remedy. On the evening of the Southport murders, she posted a message calling for hotels housing asylum-seekers to be set on fire. She deleted it herself after about three and a half hours, by which point, according to the Court of Appeal’s findings, it had been viewed 310,000 times and reposted 940 times from an account with around 9,000 followers. She was convicted of inciting racial hatred under section 19 of the Public Order Act 1986 and sentenced to thirty-one months, a sentence the Court of Appeal upheld in May 2025. Age-gating the platform’s youngest users would have changed none of this.

The authorities’ own conduct shows which instrument fits. After 2024, Ofcom’s chief executive, Dame Melanie Dawes, concluded in an open letter to the Secretary of State that such content “appears to have contributed to the significant violent disorder” after Southport, and pointed to the part played by virality and algorithmic recommendation in a crisis. The Commons Science, Innovation and Technology Committee reported in July 2025 that social-media business models incentivise the spread of dangerous content in a way that endangers public safety in the days after the murders, that the online safety regime has major holes, and that it is the companies and their recommendation algorithms that curate and amplify what users see. During the present unrest, Ofcom said some of the disorder appeared to have been incited online and that it was monitoring the situation and contacting providers under the Online Safety Act. In each emergency, the instrument chosen has been the design-facing one, backed by the prosecution of those who incite. An age threshold features nowhere in the response to the actual violence.

The case for the ban, and its limits

The serious argument for the ban is not that it would stop the riots. It is that adult production of harmful content and children’s exposure to it are distinct problems, and that a policy may sensibly reduce the second without solving the first. Removing children from particular recommendation systems could cut one audience for inflammatory or radicalising material, and children were not bystanders in this disorder: a thirteen-year-old was charged with rioting in Ballymena, around thirty youths set a bin alight in Ardoyne, and the Education Minister, Paul Givan, called it alarming that some schoolchildren had attacked the homes of their peers, while attendance collapsed at nearby schools.

That benefit is worth conceding, but it is narrow and uncertain. It does not reach the adults who create and amplify the content, nor does it change the systems that rank and distribute it to everyone who remains. Its scale and durability are unproven, and it is easily displaced to less visible or less regulated spaces, including Telegram and the private messaging that the ban excludes altogether, where some of the adult organising already runs. It is also worth being clear about what the law currently does and does not say. There is no statutory minimum age for using social media in the UK; the age of thirteen that is often cited is the age at which a child can consent on their own behalf to a service processing their data, where consent is the legal basis, not an access ban, and services can rely on other bases. The real weakness in existing controls is how easily they are evaded: Ofcom’s 2025 research estimated that 34 per cent of children aged eight to fifteen had their own profile on an online service reporting a user age of at least sixteen. The government says that keeping those functionality restrictions on by default for sixteen- and seventeen-year-olds will prevent a cliff edge at sixteen, but it eases the transition in features rather than in access. The coalition of child-protection charities, online-safety groups, academics and bereaved families that opposed a blanket ban also warned of a cliff edge at sixteen, when young people could encounter high-risk platforms without acquired resilience. That concern has particular force where political radicalisation is involved. The clearest point of overlap between the package and the disorder is livestreaming: Ballymena rioters used it to broadcast arson. But that restriction is also age-limited and therefore misses the adults who did the broadcasting.

Age assurance and the risk of substitution

Enforcement will fall on the platforms rather than on children or parents, under the Australian-style model the government proposes: companies would have to take reasonable steps to keep under-16s off, using age-assurance methods such as photo-ID matching, facial age estimation or digital identity services. Self-declaration alone would not be sufficient. Each of those methods processes personal data, and the risks of exclusion, intrusive biometrics, and insecure identity data depend heavily on the architecture chosen. The Information Commissioner’s Office has called for privacy-preserving systems and warned against collecting more than is needed. Those concerns carry particular weight in a society where address lists were circulated during the present disorder, and where, during the 2025 Ballymena violence, Paul Reilly recorded minority-background residents placing Union flags on their doors to avoid being targeted. The point is not that age assurance resembles an ethnicity register, which it does not, but that the design of any mandatory age-assurance system deserves close scrutiny here rather than after the fact.

The bigger risk is political. The principal analyses considered here, from Amnesty, Ofcom, and the select committee, locate the remedy in platform design, transparency, and enforcement rather than in the age of users. The Molly Rose Foundation makes the same argument from the child-safety side, warning that a ban leaves the harder work on algorithms undone and offers parents a false sense of safety. The danger is that a visible, popular age measure is treated as though the structural problem has been dealt with, when the duties that bear most directly on amplification already sit within the Online Safety Act and depend on sustained, sufficiently robust enforcement.

Restricting access or regulating amplification

The riots are an adult-driven, design-amplified public-order problem, with a minor falsely recast online as an asylum-seeker at the origin of the 2024 episode rather than as a young user at its end. The ban the Prime Minister has now announced may do something for children on its own terms, and it should be judged on those terms rather than dismissed out of hand. But it does not reach the people who make and spread this material, it does not alter the systems that rank and distribute it, and its real risk is that it will be mistaken for the structural reform the evidence keeps pointing to. The question Belfast poses is not whether sixteen is the right age. It is whether the government is willing to regulate how these platforms amplify, or only who is allowed to use them.

Sources: Amnesty International (2025) UK: Technical Explainer on X’s Recommender System and the 2024 Racist Riots, EUR 45/0618/2025; Committee on the Administration of Justice and Rabble Cooperative (2025) ‘Inciting a Pogrom?’ Social Media and the Racist Disorder in Ballymena and beyond during summer 2025; House of Commons Science, Innovation and Technology Committee (2025) Social Media, Misinformation and Harmful Algorithms, Second Report of Session 2024–25, HC 442; House of Commons Library (2026) Proposals to ban social media for children, CBP 10468; Lucy Connolly v R [2025] EWCA Crim 657; Ofcom (2024) Letter from Dame Melanie Dawes to the Secretary of State, 22 October 2024; Ofcom (2025) Children’s User Age, Wave 4 (cited in CBP 10468); Reilly, P. (2025) ‘Ballymena violence fuelled by mainstreaming of far-right, not just social media’, University of Glasgow; Reilly, P. (2026) ‘Belfast unrest shows the power of social media as far-right views on immigration enter the mainstream’, The Conversation, 10 June 2026; BBC News (2026) ‘Under-16s to be banned from social media, Starmer announces’, 15 June 2026, and UK Government statement on the under-16s social media ban, 15 June 2026; BBC News NI (2026) ‘What we know about how Northern Ireland’s riots were organised’, 12 June 2026; ITV/UTV News (2025), reporting on arrests and school attendance during the Ballymena disorder; Molly Rose Foundation (2026) Joint statement from children’s and online safety organisations, experts and bereaved families on a social media ban for under-16s; Public Order Act 1986, s.19.