This summer produced an unusual pairing of events. In Canberra, Australia, regulators are still trying to make a world-first under-16 social media restriction bite. In Ankara, platforms are preparing for Nov. 1, when Türkiye’s 2026 amendments will bar social-network services for children under 15. In London, the government has promised a similar age floor for 2027. And in Oakland, California, Meta agreed in late August to a settlement worth up to $18 billion with nearly all U.S. states, plus default time limits, night blocks and quieter school hours for teenagers on Instagram and Facebook.
The coincidence is easy to miss if the debate stays fixed on a single question. Should a child have an account? That is a legitimate policy question. A protective age threshold can be a legitimate first line of defense. But governments are now regulating the doorway while litigation, and a growing body of research, keep pointing to the room behind it. The harder problem is what a platform does to a young user after login.
In the 1980 essay “Do Artifacts Have Politics?” political theorist Langdon Winner asked whether technical items carry political qualities of their own. The piece became one of the turning points of science and technology studies and a forerunner of later work on technopolitics. Winner was pushing against the comfortable idea that machines and systems are neutral tools whose effects depend only on the people who use them. Design and arrangement, he argued, can themselves embody particular forms of power. Choices that look flexible at the beginning harden once they sink into infrastructure, capital and habit. The point was that technical architecture is already a political fact.
Social media makes that argument concrete. The politics of a platform do not live only in the posts that travel through it. They live in the machinery that organizes attention. This machinery consists of recommendation systems, infinite scroll, autoplay, push notifications, visible like counts, nighttime pings and the quiet optimization of time-on-app. Those features are not afterthoughts. They are the product. A 15-year-old and a 25-year-old may see different content. They often meet the same engine of engagement.
The scientific case for public intervention is now strong enough to take seriously, and still too mixed to support slogans. A 2026 systematic review in JAMA Pediatrics, pooling 153 longitudinal studies and roughly 360,000 children and adolescents, found social media use associated with higher rates of depression, behavioral difficulties, self-injurious thoughts and substance use, and with lower academic achievement and self-perception. The associations were consistent. They were also mostly modest. That combination is inconvenient for both panic and complacency.
Other work sharpens the picture without flattening it. A 2025 meta-analysis of digital-addiction patterns across more than 1.5 million young people linked compulsive use with elevated odds of depression, anxiety and sleep problems. Research from the Adolescent Brain Cognitive Development study has associated rising social media time in early adolescence with weaker performance on some reading, memory and vocabulary tests. Nighttime notifications and sleep disruption appear again and again as a plausible pathway. None of this proves that every hour online produces the same injury in every child.
The relationship can even bend. A large Australian school cohort followed for three years found a U-shaped pattern. Moderate after-school use was linked with better well-being than either abstinence or the heaviest use, and the pattern shifted with age and sex. Separate longitudinal work has associated more constructive uses of social media – connection, support, a sense of identity – with later life satisfaction and prosocial outcomes, even as raw time spent online pointed the other way. Intensity and purpose matter. So does the difference between messaging a friend and disappearing into a feed designed never to end.
This is why a policy of total access versus total prohibition is too crude. If harm concentrates in particular designs and particular styles of use, regulation has to reach those designs. Leaving the burden mainly on parents and children asks families to out-engineer companies whose business model rewards the opposite outcome.
Australia remains the most important live test of an access-first strategy. Since Dec. 10, 2025, designated platforms have been required to take reasonable steps to stop under-16s from holding accounts. Millions of accounts were removed, restricted, or deactivated. That is nothing. An early observational study published in The BMJ in June, however, found little evidence of a substantial drop in actual use three months on. More than 85% of surveyed adolescents under 16 still reported using at least one restricted platform. Many were still on their own accounts. Age checks often amounted to a self-declared birthday or a selfie. Fake profiles and borrowed logins filled the rest of the gap.
The lesson is narrower, and more useful, than a headline about failure. Legal exclusion and behavioral exclusion are not the same thing. Account removal is not automatically exposure reduction. A rule that is easy to walk around, and socially costly to obey when friends remain inside, will leak. Australia’s experiment still deserves patient evaluation. Habits and enforcement can change. What it already shows is that an age gate is a layer, not a destination.
The Meta settlement points in another direction. Settlements are not scientific proof, and they are not a confession of every allegation made in court. What they reveal is the emerging theory of harm that governments are now willing to write into enforceable terms. The money is large. The more telling clauses are architectural. Teen accounts are to face a default two-hour daily cap across Facebook and Instagram, interruptions during long scrolling sessions, a night mode from midnight to 6 a.m., and fewer notifications during school hours, with parents holding the keys to those defaults. About $5 billion of the headline figure is even conditioned on rivals such as TikTok and YouTube adopting comparable protections. That is an admission, in commercial language, that one firm cannot easily disarm an engagement race by itself.
For years, platform governance was mostly a debate about speech. What to leave up, what to take down, how to police lies and cruelty ... Child-safety regulation is dragging the argument toward the product. How is the service built? Which behaviors does it reward? When may it tap a child on the shoulder? Should the same recommendation logic be sold to a ninth-grader and an adult? Those are design questions rather than content questions.
France’s Constitutional Council offered a different warning on Aug. 14, when it struck down a general ban on social media for under-15s as a disproportionate limit on freedom of expression and communication, while also flagging weak privacy safeguards around age proof. Child protection can justify limits. A blunt lock on a whole class of communication services, applied without regard to the child or the service, may not. That ruling suggests that durable rules will have to be more precise than a nationwide off switch.
Türkiye now enters the same argument with its own timetable. Law No. 7578, published in the Official Gazette on May 1, amends the internet law so that, from Nov. 1, social network providers may not offer services to children under 15 and must take measures, including age verification, to that end. They must also offer differentiated services for children who have turned 15, publish those measures, and provide parental tools that can reach account settings, paid transactions and time spent. The law is not a carbon copy of Australia’s, and it is not yet a finished design code. It is an opening.
After November, the useful test will not be only how many underage accounts were blocked. Did actual exposure fall, or did children move to false accounts and neighboring apps? Did nighttime use change? Did compulsive patterns ease? Did recommendation environments for adolescents look different from those built for adults? Did parental tools work in ordinary households, or only in the compliance brochure? Those questions decide whether an age rule becomes a ritual or a policy.
Winner’s old caution about timing belongs here. The first months of a new rule still look negotiable. Once age gates, verification vendors, app-store settings and teenage workarounds settle into daily infrastructure, the political choice hardens around whatever design was left untouched. Türkiye can still treat verification as the beginning of age-appropriate product regulation rather than the end of the argument. Younger children can face a firm access line. Adolescents can meet a protected environment with slower defaults, quieter nights, less social scoring, tighter recommender settings around self-harm and appearance content, and measurable duties on the firms that profit from their attention. Adults can keep an adult internet. Children should enter a different environment. They should not only pass a different gate.
The first wave of child online-safety law built fences around platforms. The next wave will have to inspect the machinery inside. Age limits can keep some children off the premises. They cannot decide what kind of premises remain legal for the children who still get in.
