Protecting Children Without Turning the Internet into an Identity Checkpoint

• The Problem is Real, the Proposed Cure is not
• Children Should not be Treated as Tiny Adults in a Maximised-Attention Market
• Age Assurance is not Automatically Identity Verification - but it can become it
• A Phone App is not a Passport to Public Life
• Identity Gates Change the Relation Between People and Information
• The Record of “Only This Purpose” is not Reassuring
• A Confidential System is Still a System That Can Fail
• Hate, Extremism and the Difference Between Response and Amplification
• Why Anonymity Tools are a Democratic Safety Valve
• What a proportionate approach looks like
• Conclusion
• References

The Problem is Real, the Proposed Cure is not

There is a temptation to collapse two questions into one. The first is whether children and teenagers should be left, without guidance or limits, in the largest and most commercially optimised communication environment humanity has built. The second is whether states should respond by making ordinary access to information conditional on a verifiable identity, an age credential tied to identity or a real-name account. The answer to the first question can quite reasonably be no. The answer to the second should still be no.

Children need boundaries. This is neither puritanism nor a claim that the internet is uniquely evil. It is ordinary care. A parent would not normally tell an eleven-year-old to go alone to any bar, remote meeting place, or room full of unknown adults at night, then announce that this is safe. Yet a child with an unrestricted smartphone can be introduced, at any hour, to arbitrary adults, sexual material, violence, gambling, self-harm communities, manipulative advertising, harassment and an endless stream designed to hold attention rather than serve development. The medium is real social space, even when it is encountered through a small glowing rectangle.

It is also important not to make the inverse mistake. Children are not made safe by excluding them from knowledge, from peers, from queer communities, from mental-health support, from political information or from the wider world. Online spaces can provide friendship, creative work, education, help in a crisis and a way out of isolation. The relevant distinction is not “the internet” versus “real life”. It is between a child having structured, age-appropriate, supervised access and a child being placed alone in a global, opaque, incentive-driven environment.

That distinction matters because a genuine problem does not make every response legitimate. The present enthusiasm for mandatory age checks and real-name access risks building a general-purpose checkpoint between a person and information. Its first use may be restricted adult material. Its later uses are easy to imagine: gambling, “harmful” political speech, health information, protest organising, minority communities, discussion of war, labour organising or simply the views of a governments opponents. A system designed to decide who may see a category of information is a system that can later decide which categories count and who qualifies.

The better position is therefore demanding in two directions at once: take child safety seriously and refuse to turn that duty into a population-wide identity-and-surveillance infrastructure.

Children Should not be Treated as Tiny Adults in a Maximised-Attention Market

Some online content really is not suitable for children. This includes material made to sexually arouse adults; realistic violence; instructions and communities that promote self-harm, eating disorders, coercion or exploitation; gambling-like systems; content that deliberately frightens, humiliates or targets a young person. The fact that an older teenager may be able to discuss some of these subjects in an educational setting does not make a commercially packaged, frictionless, recommendation-driven feed appropriate for a younger child.

The danger is not only particular pages. A platform can convert a passing curiosity, a moment of distress or a single click into a repeated stream of similar material. Recommendation systems optimise measurable engagement - watch time, return visits, clicks, shares - not the long-term wellbeing of the person whose attention they capture. For minors, who are still developing impulse control, social judgement and a stable sense of self, this is a particularly poor environment in which to outsource the setting of boundaries.

The evidence on social media and mental health deserves accuracy rather than alarmist slogans. It does not establish that every hour online causes depression, or that social media has one uniform effect on every young person. Effects tend to be heterogeneous, often modest at population level and are entangled with prior distress, sleep, bullying, family circumstances, the kind of use, and the kind of content. An umbrella review by Valkenburg, Meier and Beyens found that much of the review literature described the association as weak or inconsistent, while also arguing that this does not make risks for particular adolescents unimportant[23]. A recent systematic review similarly found an association between problematic social-media use and symptoms of depression and anxiety, with sleep loss, social comparison and feedback-seeking among the relevant pathways; it explicitly calls for stronger longitudinal evidence[17].

This is the right intellectual posture: take the risk seriously without pretending that correlation has solved every causal question. A 2020 meta-analysis found a small but statistically significant association between adolescent social-media use and depressive symptoms (pooled $r=0.11$); small effects can still matter when exposure is widespread and when a subset of children is especially vulnerable[14]. Reviews also connect excessive social-media use with poorer sleep, which is itself an important route by which unrestricted late-night use can damage wellbeing[1]. The policy implication is not “ban communication”. It is to create time boundaries, reduce persuasive and addictive design, require genuinely safe defaults for minors, provide responsive moderation and reporting and give parents and young people usable controls.

“Doomscrolling” is an especially useful word here because it describes a pattern rather than a medium: repetitive, difficult-to-stop consumption of alarming or negative material. It is not only a youth problem; adults do it too. Research is still emerging and much of it is cross-sectional, so it should not be described as proof of a single causal disease. But the signal is not imaginary. Satici et al. found doomscrolling associated with psychological distress and lower wellbeing in their validation studies[18], while a 2024 study in samples from Iran and the United States reported associations with existential anxiety and more pessimistic views of human nature[19]. This supports a mundane conclusion: a feed engineered to keep a person searching for the next threat is not neutral nutrition for a developing mind.

No machine can supply the role of an attentive adult. Age-appropriate devices, bedtime rules, family discussions, trusted contact routes, school support, limited public discoverability and the ability to say “show me what happened” are not glamorous policy, but they are real protection. Where parents are too overworked, absent, unwell, unsupported or themselves captured by the same attention economy, that is a social problem - of care, housing, working time, education, mental-health support and community - not merely an isolated defect of “the media”. A state that avoids those responsibilities but asks every citizen to present credentials before reading is choosing the administratively easy solution, not necessarily the humane or effective one.

Age Assurance is not Automatically Identity Verification - but it can become it

The technical distinction must be made plainly. It is possible, in principle, for a person to prove a limited statement - “I am over 18” - without telling a website their name, exact birthdate, address or a stable identifier. This can use a credential issuer, cryptographic signatures, unlinkable or limited-use tokens and a zero-knowledge proof. The European Commissions current age-verification blueprint expressly describes this aspiration: proof that a user is above a threshold without revealing other personal information[8].

That design is substantially better than uploading a passport to every adult site. It is not, however, a reason to stop asking hard questions - or to decide that any institution deserves the role it requires. No government, credential issuer, platform, wallet provider, browser vendor, app-store operator, network operator or contractor is sufficiently trustworthy to become the standing gatekeeper of a persons access to information. This is not an accusation that every employee is malicious. It is a recognition that institutions change, are pressured, are breached, make mistakes, follow changing laws, outsource functions and eventually acquire new administrators and political masters. A free society must be designed so that ordinary lawful reading and speech do not depend on continuing confidence in the discretion, competence or goodwill of any one of them.

Who initially binds the credential to a person? Which issuer is trusted? Can an issuer, wallet provider, browser, network operator, or government observe issuance, renewal, revocation or repeated use? Are proofs genuinely unlinkable across relying parties? What metadata is retained, for how long and under which legal power can it be demanded? Does the implementation match the published design, and can independent researchers audit both code and deployment? Can a teenager safely seek information about sexuality, abuse, mental health or politics without generating a durable record? The cryptographic protocol may reveal little to the final website while the surrounding system still creates a sensitive map. And keep in mind: Laws are changed quickly. Technical reality is not. A system that does not exist cannot be misused - a system that exists can be modified quickly.

“Privacy-preserving” is an engineering goal and a property to be demonstrated, not a spell that makes a political risk disappear. A central issuer can be designed not to learn where a credential is used. That is good cryptographic hygiene, but it does not make an access gate acceptable. It still places a state-recognised or state-regulated credential between a person and information, establishes an authority able to issue, deny, revoke or alter admission and makes access contingent on carrying an approved proof. A system may begin with pornography and gambling; it also creates a ready-made interface for any future law that says access to a category requires an eligibility proof. The dangerous change is not only a database leak. It is the normalisation of asking permission, through an accredited gatekeeper, before receiving information.

The conclusion of this essay is therefore deliberately stronger than a request for better implementation: a privacy-preserving proof of age must not become a prerequisite for receiving lawful information. If a proof is voluntary and used by a person to obtain a benefit from a particular private service, that raises a different question. A compulsory proof, however cleverly designed, changes a basic liberty into a conditional entitlement. The public should not have to obtain a digital token, persuade a credential issuer, possess an approved device or pass an eligibility check before it may read and participate in ordinary lawful communication. A state-managed identity requirement or a platform real-name rule is more intrusive still, and should be rejected.

This does not leave child protection empty-handed. It directs regulation toward the party that sells and designs the risky environment: require safer defaults for accounts known or reasonably likely to belong to minors; restrict exploitative recommender systems and advertising; provide device- and family-level controls chosen by parents; remove criminal (not undesired) material; and investigate concrete abuse. The burden should not be shifted from platforms and parents onto every adults right to receive information.

A Phone App is not a Passport to Public Life

The proposal becomes even less acceptable when the credential is carried in a compulsory mobile-phone application. A phone is not an identity document, a democratic qualification or a condition of access to public discourse. Making an app the route through an age gate silently turns possession of a particular, current, networked consumer device into a prerequisite for reading. It excludes people who do not own a smartphone, cannot afford a supported model or data plan, use a shared device, have a disability that makes the app inaccessible, use a privacy-oriented operating system, are travelling or simply refuse to carry a personal tracking device.

It also concentrates risk. A mobile application is embedded in an operating system with a device identifier, app-store account, push-notification service, network connection, telemetry, crash reporting, analytics libraries, battery and location-related metadata and a continuous update channel. A protocol may promise that the final site sees only “over 18”, while the operating system, application provider, issuer, network or a future compulsory update can still create a rich correlation surface. A phone-based gate makes revocation and denial immediate: disable an account, demand a new app version, change a policy, or make the device fail attestation and a person can be excluded from information at once.

There is a deeper objection than security. A government should not get to establish a de facto rule that one must carry a state-compatible app in ones pocket to enter parts of the public conversation. Such a rule makes freedom of information dependent on private hardware vendors, app-store policies, battery life, network service and administrative approval. That is a total non-starter for a free society. The right response is to make services safer for children, not to turn the smartphone into a domestic border checkpoint.

Identity Gates Change the Relation Between People and Information

Anonymous and pseudonymous speech have a legitimate social function. A survivor of abuse may need to ask for help. A teenager may need to explore an identity before they can safely disclose it at home. A whistle-blower, a union organiser, a dissident, a patient, a journalists source or a person living in a small community may need to speak and read without making themselves immediately legible to an employer, local authority, political faction, family member, platform or state.

The argument is sometimes reduced to the straw man that anonymity means impunity. It does not. Serious crimes can be investigated with due process and evidence. Services can moderate conduct, throttle abuse and suspend accounts without collecting a government identity for every reader. Pseudonymity also allows reputation, continuity and community norms without forcing a legal name into every interaction. The relevant question is not whether identity is ever useful. It is whether everyone should have to disclose or certify identity before participating in ordinary communication. They should not.

The same point applies to companies. A platform does not need a passport scan to apply rate limits, block harassment, maintain an account reputation, provide child settings or comply with a court order directed at a specific serious offence. Giving every platform a verified identity or a reusable age credential that becomes a de facto identifier, expands the number of institutions capable of correlating a persons behaviour. Data that begin as “proof of age” have a predictable gravitational pull toward advertising, profiling, fraud systems, employment screening, insurance, political pressure and law-enforcement requests.

The state should not receive that map either. To know who reads, searches for, joins or discusses a topic is to know something about private life, association, belief and future action. Freedom to seek and receive information is a basic condition of personal autonomy and democratic self-government; it must begin from a presumption of open, uncredentialed access, not from the question whether a person has passed an app-mediated admission test. The European Convention on Human Rights protects not only speaking but receiving and imparting information and ideas without interference by public authority[3]. Article 19 of the International Covenant on Civil and Political Rights protects opinion without interference and protects the freedom to seek, receive and impart information[21].

These rights are not absolute. The same human-rights instruments permit carefully defined restrictions for interests such as the rights of others, public safety and prevention of crime, subject to legality, necessity and proportionality. That qualification is crucial: nobody has a civil-liberties right to sexually exploit a child, make credible threats, stalk someone, distribute abuse material or coordinate violence. But the qualification is not a blank cheque. A general identity gate imposed on everyone in order to deal with a minority of offenders is a particularly poor fit for necessity and proportionality. It burdens the innocent reader, chills lawful speech, establishes a reusable control point and still leaves determined offenders able to use stolen credentials, foreign services, automation, encryption, proxies or simply one anothers accounts.

The Record of “Only This Purpose” is not Reassuring

History does not prove that every child-protection measure becomes tyranny. It does show however why powers built in response to fear should be judged by their structure, not merely by the intention announced at enactment. The recurring pattern is familiar:

• a narrow emergency is presented
• a broad technical capability is acquired
• exceptions and data-sharing arrangements follow
• independent courts or investigators later discover that safeguards were weaker than promised

After 11 September 2001, the United States expanded surveillance powers in the name of counter-terrorism. Section 215 of the USA PATRIOT Act was used for the NSAs bulk telephone-records programme. The independent Privacy and Civil Liberties Oversight Board later concluded that the programme lacked a viable legal foundation, raised serious constitutional and civil-liberties concerns and had shown only limited value; the USA FREEDOM Act ended that bulk programme in 2015[16]. This is not an argument that terrorism was invented or that intelligence work is never needed. It is an argument that “terrorism” is not a sufficient answer to the question “may the state assemble a record of ordinary peoples associations?”

Europe supplies its own lessons. The Court of Justice of the European Union annulled the Data Retention Directive after it required broad retention of communications metadata for serious-crime purposes, finding the interference with privacy and data protection disproportionate[4]. In Big Brother Watch and Others v. United Kingdom, the European Court of Human Rights found violations of Articles 8 and 10 in the United Kingdoms historic bulk interception and communications-data regimes; it identified inadequate independent authorisation, insufficient safeguards and insufficient protection for confidential journalistic material[10]. The Court did not say that all bulk interception is inherently unlawful. It said the power required genuine end-to-end safeguards, necessity and proportionality. That is a high bar for good reason.

Other examples make the same point in different vocabulary. Anti-terror legislation has been used to facilitate broad communications collection; “extremism” definitions in authoritarian states routinely encompass political opposition and minority groups; and proposals framed as stopping child sexual abuse have sometimes been designed in ways that would weaken private communications or introduce client-side scanning. The UN Special Rapporteur and other mandate holders warned in 2020 that the proposed US EARN IT Act, though directed at child sexual-abuse material, raised concerns for privacy, freedom of expression and encryption[22]. The lesson is not to ignore child exploitation. It is to insist that the investigation of specific offences target offenders and evidence, rather than converting private communication into a permanently inspectable space.

The speed-control point is especially important. Censorship is not limited to deleting a text. A state can make reading slower, more costly, riskier, less private, or conditional on a credential. It can force a service to identify readers, delay dissemination, lower reach, filter search results, remove circumvention tools or create self-censorship through the possibility of later identification. When people believe that their reading list and associations can be reconstructed, many will avoid lawful but sensitive subjects. That chilling effect is not a side issue; it is how a formally free information space becomes practically narrower.

The United Kingdom provides a concrete warning about overbroad speech policing without requiring exaggerated claims that “police arrest people for every Facebook post”. In Miller vs College of Policing, a police officer visited Harry Millers workplace over tweets that were recorded as a non-crime hate incident, warned him about escalation and advised him to stop posting gender-critical messages. The Court of Appeal held that the guidance’s operation unlawfully interfered with Article 10 rights and had a chilling effect in a public-interest debate[7]. This does not mean harassment or incitement should be tolerated; British prosecutorial guidance itself distinguishes communications that are merely offensive, shocking or rude - which are generally protected - from conduct crossing the high threshold for an offence[5]. It shows instead how a laudable desire to oppose hatred can produce an apparatus that pressures lawful speech when concepts are broad and consequences occur before a court ever considers guilt.

A Confidential System is Still a System That Can Fail

Claims that an identity or age-verification mechanism is anonymous, double-blind or secure should be tested, not ridiculed. Good cryptography can materially reduce disclosure. But no one can guarantee that every implementation, vendor integration, logging setting, network path, support process, lawful-access request, future update, or data-sharing agreement will preserve that ideal. Systems fail through code defects, bad architecture, insider access, misconfigured spreadsheets, phishing, procurement shortcuts and political decisions as well as through broken mathematics.

Governments are not uniquely malicious, but neither are they uniquely competent custodians of a permanent access map. In 2015, the US Office of Personnel Management reported breaches affecting 21.5 million peoples sensitive background-investigation records, involving current, former and prospective federal employees and contractors[12]. In 2023, the Police Service of Northern Ireland disclosed a spreadsheet in response to a freedom-of-information request; hidden data exposed the surnames, initials, ranks and roles of all 9.483 officers and staff. The UK Information Commissioner later fined the PSNI 750.000 pounds[13], but this does not erase the leaked data. These events do not imply that a particular European age-proof system will necessarily leak. They establish the more modest and decisive point: institutions that possess sensitive identity data are fallible and a database or log cannot be compromised if the system never needed to create it.

The case against a general identity gate is therefore not based on distrust of one current government alone. Laws, administrations, contractors and political majorities change. A credential issued under a liberal government can be inherited by an illiberal one. Security guarantees are never stronger than the weakest component and the future legal environment. The appropriate design principle is data minimisation: do not collect a map of lawful private reading, association or expression merely because it might be convenient later.

Hate, Extremism and the Difference Between Response and Amplification

There is no need to romanticise hateful speech. National Socialism, racism, antisemitism, misogyny, homophobia, threats and campaigns to hurt people are ugly, destructive and sometimes criminal. People targeted by such speech deserve support and protection. The question is how a free society denies it power without giving the state a general licence to decide what citizens may read, say or discuss.

One indispensable mechanism is social: do not give destructive movements a permanent resonator. Do not quote-tweet every provocateur into prominence; do not turn every small groups bait into a national spectacle; build better institutions, local ties, economic security and persuasive counter-speech; and enforce clear laws against threats, stalking, coercion, violence and targeted abuse. A society that treats every odious claim as irresistible content for outrage-driven amplification often helps the claim travel further.

That is not a magic doctrine of “ignore all extremism”. Direct threats, recruitment into violence, coordinated harassment, child exploitation and incitement require serious, lawful response. Nor should vulnerable people be told to endure abuse silently. The point is narrower: broad censorship and mass identification cannot supply the social stability, trust and resilience that make destructive ideas unattractive. If a society has become so brittle that an anonymous troll or fringe slogan can destabilise it, the long-term remedy is to repair the social conditions, not to record the identity of every reader.

The media itself illustrates why attention should be allocated carefully. Research on European immigration coverage finds recurring patterns in which migrants are under-represented and commonly framed through delinquency, criminality, negativity and conflict; repeated exposure can shape attitudes and even vote choice[6]. In another well-known study, Boykoff and Boykoff showed how the journalistic norm of presenting “both sides” of climate science produced coverage that diverged sharply from the scientific evidentiary balance[2]. The point is not that all journalists lie or that all coverage is a conspiracy. Framing is often produced by selection pressures, deadlines, editorial habits, commercial incentives, access to official sources and the need for a dramatic conflict.

There are also institutional admissions of error. In 2004, The New York Times publicly reviewed shortcomings in parts of its pre-war reporting on alleged Iraqi weapons of mass destruction[25]. That example is useful precisely because it comes from a major outlet with correction mechanisms: large established media can produce excellent reporting and can still transmit mistaken premises, selective sourcing or political framing at consequential moments.

Artificial intelligence makes source literacy more necessary, not less. Synthetic images, generated text, coordinated persona networks and cheap translation can make an item look more polished than it is. But the corrective cannot be “trust official media, distrust everything else”, nor “trust a blue tick, an ID card or a government-approved source”. A resilient public compares independent sources, follows primary documents, checks dates and incentives, distinguishes evidence from commentary, asks what is missing and remains able to inspect dissenting accounts. The ability to do that privately matters. A state that controls which sources may be reached, and knows who sought an unpopular source, has made verification dependent on its own permission.

Why Anonymity Tools are a Democratic Safety Valve

Tor is useful here not because it is mystical or because it prevents all investigation. It is a privacy and censorship-circumvention tool that routes traffic through a network so that a destination does not simply see the users ordinary network address. It can reduce linkability between a persons location and their reading or publication. Its protection has limits: a user can still reveal themselves by logging into an identified account, by sharing distinctive information, by malware compromise or through poorly configured applications. It is not a substitute for careful operational security.

Nevertheless, the reaction of censoring states is revealing. Chinas Great Firewall has long tried to block Tor and discover Tor bridges; Winter and Lindskogs technical study documented the mechanism and potential countermeasures[24]. Russia began blocking access to the Tor network and Tor Project website on some networks in 2021, as measured by the Open Observatory of Network Interference (OONI), and later efforts have targeted bridges and circumvention transports[15]. The Tor Project reported further Russian blocking pressure in 2024 while noting that bridges and pluggable transports remained important means of access[20]. In China, many users have relied on circumvention tools to access and share information during politically sensitive events, despite extensive blocking[11].

This does not prove that every user of Tor is virtuous. Tools that preserve privacy can be used by people with good intentions and by people with bad ones, just as roads, cash, locks and encryption can. The question for a free society is whether the ordinary person may have private access to lawful information and a means to speak without becoming a target. In China and Russia, the answer is visibly important because the state has already made access politically dangerous. In Europe, the same capacity is a safeguard against future excess, against coercive profiling and against the temptation to treat inconvenient citizens as a security problem.

It also exposes a practical weakness in identity-gate policy. A determined criminal, determined abuser or determined political actor can seek technical workarounds, stolen credentials, offshore providers, coercion or private networks. The group most reliably burdened by universal checks is the broad ordinary public: adults who want lawful privacy, small services without compliance departments, young people seeking sensitive help, minorities, researchers and people who simply do not want their reading habits connected to a legal identity. When a policy is easy to bypass for the determined and highly intrusive for everyone else, it should not be casually called effective child protection.

What a proportionate approach looks like

The choice is not between abandoning children and constructing a universal speech passport. A serious approach starts from the different responsibilities of families, platforms, schools, service providers and the state.

Parents and guardians should provide age-appropriate devices and accounts, establish sleep and screen-time boundaries, talk about sexual material and manipulation before a platform does, use tools that give them visibility without humiliating surveillance and remain a safe person to contact when something goes wrong. Society should make this feasible by supporting parents with time, education, counselling and functioning child-protection services.

Platforms should be expected to design for minors rather than merely write terms of service. That means safe defaults, low discoverability by strangers, no targeted behavioural advertising to children, clear reporting routes, rapid action on grooming and abuse, meaningful researcher access, limits on addictive recommender patterns, age-appropriate privacy settings and transparent moderation procedures. The European Commissions guidance under Article 28 of the Digital Services Act itself identifies risks including grooming, harmful content, addictive behaviour, cyberbullying and harmful commercial practices; age assurance is only one possible measure in a wider set[9].

The state should fund education, mental-health care, child-protection investigation and specialised law enforcement; prosecute actual abuse, threats, stalking and exploitation through due process; protect encryption and private communication; require strong security and transparency from platforms and preserve judicial, parliamentary and independent oversight. It should avoid mandatory real-name rules, compulsory general identity credentials for ordinary online participation, universal scanning, and retention systems that make everyones reading and communication available for later inspection.

No age proof - whether identity-revealing, “anonymous”, double-blind, wallet-based or zero-knowledge - should be compulsory merely to read, search, join a discussion or receive information. Nor should a mobile-phone app be required as the key to any part of ordinary public discourse. Where adults choose to use voluntary controls for a particular service, they should be free to do so; but law and platform policy must not convert those controls into a general condition of access. The burden belongs on service design, targeted enforcement, child care and parental tools, not on every person to authenticate their right to learn.

Conclusion

Children deserve more than an unfiltered encounter with the whole internet. They deserve adults who pay attention, platforms that do not exploit immaturity, schools that teach judgement and a society that provides real support when family care fails. Some content is unsuitable for minors. Excessive and compulsive engagement is harmful for many people, including adults. These are good reasons for boundaries, better design and targeted protection.

They are not good reasons to place a government-approved identity gate between the public and information. A privacy-preserving proof may disclose less than a passport scan, but it still makes the right to receive information conditional on a credential, issuer and technical infrastructure; it is therefore not an acceptable solution. A mandatory mobile app is worse again: it makes a personal networked device the key to public life and adds exclusion, correlation and remote-control risks. Such gates create a technical and legal lever that can later be reused to sort readers, slow information, profile association and suppress unpopular communities or opposition. Claims of anonymity and privacy are worthwhile only when the entire system - not only its brochure or protocol- can demonstrate them over time. The historical record of emergency surveillance, overbroad data retention and public-sector data breaches gives no reason to be casual about that trust.

A free and stable society does not become safer by making every citizen legible before they may look, learn or speak. It becomes safer when children are cared for, harmful conduct is addressed precisely, platforms are held to account, people learn how to judge information, and privacy remains a normal condition of lawful life rather than a privilege one must justify.

References

• [1] Alonzo, R. et al. “Interplay Between Social Media Use, Sleep Quality, and Mental Health in Youth: A Systematic Review.” Sleep Medicine Reviews 56 (2021): 101414. https://doi.org/10.1016/j.smrv.2020.101414
• [2] Boykoff, M. T., and J. M. Boykoff. “Balance as Bias: Global Warming and the US Prestige Press.” Global Environmental Change 14, no. 2 (2004): 125–136. https://doi.org/10.1016/j.gloenvcha.2003.10.001
• [3] . Council of Europe. European Convention on Human Rights, Article 10. Official text
• [4] Court of Justice of the European Union. Digital Rights Ireland and Seitlinger and Others, Joined Cases C-293/12 and C-594/12, judgment of 8 April 2014. Judgment
• [5] . Crown Prosecution Service. “Responding to Intimidating Behaviour in Elections and Public Office.” Guidance
• [6] Eberl, J.-M. et al. “The European Media Discourse on Immigration and its Effects: A Literature Review.” Annals of the International Communication Association 42, no. 3 (2018): 207–223. https://doi.org/10.1093/anncom/wqy002
• [7] England and Wales Court of Appeal. R (Miller) v College of Policing 2021 EWCA Civ 1926. Judgment
• [8] European Commission. “Commission Makes Available Age-Verification Blueprint.” 14 July 2025. Official release
• [9] European Commission. “Commission Publishes Guidelines on the Protection of Minors.” 14 July 2025, updated 22 April 2026. Guidelines page
• [10] European Court of Human Rights. Big Brother Watch and Others v. the United Kingdom GC, 25 May 2021. Court factsheet and case summary
• [11] Freedom House. China: Freedom on the Net 2024. Report
• [12] Government Accountability Office. OPM Has Improved Controls, but Further Efforts Are Needed. GAO-17-614, 2017. Report
• [13] Information Commissioner’s Office. “What Price Privacy? Poor PSNI Procedures Culminate in £750k Fine.” 3 October 2024. Decision summary
• [14] Ivie, E. J. et al. “A Meta-Analysis of the Association Between Adolescent Social Media Use and Depressive Symptoms.” Clinical Psychological Science 8, no. 6 (2020): 1102–1116. https://doi.org/10.1177/2167702620923842
• [15] Open Observatory of Network Interference. “Russia Started Blocking Tor.” 17 December 2021. Measurement report
• [16] Privacy and Civil Liberties Oversight Board. Recommendations Assessment Report. 2016. Report
• [17] Saleem, N. et al. “Exploring the Relationship Between Social Media Use and Symptoms of Depression and Anxiety Among Children and Adolescents: A Systematic Narrative Review.” Cyberpsychology, Behavior, and Social Networking 27, no. 11 (2024): 771–797. https://doi.org/10.1089/cyber.2023.0456
• [18] Satici, S. A., E. Gocet Tekin, M. E. Deniz, and B. Satici. “Doomscrolling Scale: Its Association with Personality Traits, Psychological Distress, Social Media Use, and Wellbeing.” Applied Research in Quality of Life 18, no. 2 (2023): 833–847. https://doi.org/10.1007/s11482-022-10110-7
• [19] Shabahang, R. et al. “Doomscrolling Evokes Existential Anxiety and Fosters Pessimism About Human Nature? Evidence From Iran and the United States.” Computers in Human Behavior Reports 15 (2024): 100438. https://doi.org/10.1016/j.chbr.2024.100438
• [20] Tor Project. “Tor in Russia: A Call for More WebTunnel Bridges.” 28 November 2024. Technical update
• [21] United Nations Office of the High Commissioner for Human Rights. International Covenant on Civil and Political Rights, Article 19. Official text
• [22] United Nations Special Procedures. “Communication Concerning the Proposed EARN IT Act.” 19 March 2020. Communication
• [23] Valkenburg, P. M., A. Meier, and I. Beyens. “Social Media Use and its Impact on Adolescent Mental Health: An Umbrella Review of the Evidence.” Current Opinion in Psychology 44 (2022): 58–68. https://doi.org/10.1016/j.copsyc.2021.08.017
• [24] Winter, P., and S. Lindskog. “How China Is Blocking Tor.” USENIX FOCI, 2012. Paper
• [25] PBS NewsHour. “The New York Times WMD Coverage.” 26 May 2004. Report on the newspaper’s self-critique

This article is tagged: Opinion, Internet, Society, Privacy, Experience, Social, Security