KOHEN AVOCATS FIRM · PARIS 16
France’s Under-15 Social Media Ban Is Blocked: What Global Platforms Must Fix Before Age Verification Returns

Hassan KOHEN acts in urgent matters, from the police station to the assize court. Initial strategic analysis: €80 incl. VAT, personal reply within 24 hours.

100% confidential · Legal professional privilege · No obligation

Article generated by artificial intelligence, through a process designed and supervised by the firm

Source: French Court of Cassation – Judilibre and Légifrance open data

Paris Bar Police custody, investigation, assize court CNB listing avocat.fr
Maître Hassan KOHEN, avocat au Barreau de Paris
Hassan KOHEN
Attorney at the Paris Bar

France’s Under-15 Social Media Ban Is Blocked: What Global Platforms Must Fix Before Age Verification Returns

Speak to a lawyer 06 89 11 34 45

France’s Constitutional Council has blocked the country’s planned blanket ban on social-media accounts for children under 15, but the decision is not a legal retreat from child protection. In Decision no. 2026-911 DC of 14 August 2026, the Council censured Article 1 of the statute because the measure restricted expression and privacy without building the safeguards needed to justify that restriction. The ruling is therefore a warning about legislative design: a state may pursue safer digital spaces, yet it must show why the chosen restriction is necessary, adapted to the risk and proportionate to its effects.

The ruling matters far beyond France. The proposed system would have required platforms serving the French market to verify the age of every user, close existing accounts held by under-15s and operate a verification process approved by the French regulator. The Associated Press report on the decision records the immediate political consequence: the Government and Parliament must now rethink the mechanism before trying again. For global platforms, counsel, parents and regulators, the practical question is no longer whether age assurance will return. It is whether it can return without turning child safety into a permanent identity checkpoint for the whole population.

This analysis explains the constitutional failure, the interaction with the Digital Services Act and the General Data Protection Regulation, and the compliance architecture a new French or European model will need. It also places the decision in the line of French case law on online speech, privacy, minors and age verification. The result is a cross-border lesson: the next enforceable system will have to prove its risk model, minimise the data it handles, give users a meaningful remedy and remain intelligible to courts in several jurisdictions at once.

I. Why did France’s under-15 social-media ban fail—and why does the decision reach global platforms?

A. Why the blanket ban failed the necessity and proportionality test

The law was designed as a direct access rule. A person below the age of 15 would not be permitted to open or keep an account on the covered social networks, while platforms would have to verify age rather than rely only on a user’s declaration. The political timetable was unusually tight: the system was intended to operate from the start of the September school year, with existing accounts closed within a short transition period. That timetable made the constitutional review especially important, because a rule affecting millions of users would have moved from legislative text to automated enforcement almost immediately.

The Council did not treat the objective as illegitimate. Protecting children from addiction, isolation, pornography, harassment, fraud and other harmful experiences can justify legislation. The constitutional problem was the fit between that objective and the legal instrument. The Constitutional Council’s Decision no. 2026-911 DC starts from Article 11 of the Declaration of the Rights of Man and of the Citizen, which protects communication of thoughts and opinions, and from the legislature’s power under Article 34 of the Constitution to establish rules governing public freedoms. The relevant provisions must be read with the official texts of Article 11 of the Declaration and Article 34 of the Constitution.

That framework requires more than a plausible policy concern. A restriction must be necessary, adapted to the objective and proportionate to the burden it places on rights. The Council found that the challenged measure applied across the category of social networks without sufficiently distinguishing between services, functions, content, age-related risks or protections already available. A private messaging feature, a professional community, a public-interest information forum and an algorithmic video service could fall within the same legal perimeter even though their risks and modes of access are materially different. A law that treats those situations as identical must explain why the difference does not matter. This statute did not do so convincingly enough.

The absence of individual or service-level assessment was decisive. The legislation attached the prohibition to the user’s age and the service’s classification, not to a demonstrated risk produced by a particular function. It also left no effective route by which a parent or legal representative could authorise, limit or tailor access in a way that responded to the child’s circumstances. The Council therefore saw a general ban that did not sufficiently account for the identity of the minor, the nature of the service or the safeguards that could reduce the risk. The decision does not say that Parliament can never set an age threshold. It says that a threshold cannot carry the whole constitutional burden by itself.

The comparison with the earlier age-verification dispute concerning online pornography is useful but limited. In its 5 January 2023 decision, no. 22-40.017, the First Civil Chamber considered a challenge involving Article 227-24 of the Criminal Code and legislation addressing minors’ access to pornographic content. That statutory context targeted a defined category of serious harm. The corresponding Article 227-24 of the Criminal Code also contains a specific rule stating that the offence may exist even where a minor’s access resulted from a simple declaration of being at least 18. The Court held, in that context, that a more reliable age check could be necessary, adapted and proportionate.

That precedent does not automatically validate a universal social-media checkpoint. Pornographic content and social-media access are not the same legal object. The first involves a narrowly identified category of material that the legislature can define and regulate. The second includes political speech, education, peer communication, journalism, art, commerce and private relationships. A platform-wide prohibition can reach lawful speech before any harmful content appears. The government therefore needs a stronger demonstration of why each affected service creates the relevant risk and why less intrusive measures would not work.

The Court’s approach also matters for enforcement design. A regulator cannot repair a disproportionate statute through informal guidance after the statute has already imposed the restriction. The legislature must define the perimeter, the trigger, the duration, the exceptions and the remedies with enough precision for users and platforms to predict the consequences. A decree can provide technical details; it cannot create the missing constitutional safeguards. If the next bill simply changes the age threshold or gives the same blanket rule a new name, the underlying defect will remain.

For an international platform, this is a concrete warning against a single global switch. A French rule may be territorially limited in law, yet its technical implementation can affect account creation, identity services, parental controls, advertising eligibility and data flows across the entire product. If the platform uses one age-assurance vendor for Europe, a French obligation can become an architectural constraint for users in other countries. The company must separate what French law requires from what its global product chooses to standardise, and it must be able to show the difference in an audit or court record.

The political response confirms that the issue is not closed. The President has indicated that a revised measure remains possible during the current term. The constitutional decision is thus better understood as a drafting deadline than as a final answer. The next proposal must supply an evidentiary record: which services create which harms, why the chosen intervention addresses those harms, how it protects lawful communication, and how the system works when a user is travelling, lacks a national identity document or challenges the classification of a service. Those questions are cross-border by nature, even when the original statute is French.

B. Why age verification turned a child-protection bill into a privacy problem for every user

The second constitutional weakness was the privacy impact of requiring proof of age from everyone. An under-15 ban is enforced at the front door. Unless the platform can reliably identify the people who fall below the threshold, it must ask all users for information or a token capable of establishing age. That changes the legal subject of the rule. The child is no longer the only person affected; adults, visitors, journalists, whistle-blowers, lawyers and users of public-interest services also become participants in an identity-screening system.

The Council treated that consequence under the protection of private life. Its reasoning identified the need for legal safeguards governing the conditions and limits of age proof. A statute that says “prove your age” but does not set the permitted source, retention period, access controls, security duties, separation of identity from age status or rules for deletion leaves too much of the rights balance to private operators. The problem is not simply that data might leak. It is that the law has not defined the boundaries within which collecting the data is constitutionally acceptable.

That concern has a French private-law dimension as well. Article 9 of the Civil Code states the right to respect for private life and gives judges powers to prevent or stop an infringement. The provision is not an age-verification code, but it shows why the design cannot be evaluated only through the lens of platform convenience. A database that links a real identity to a sensitive account, a parental relationship, a health-related vulnerability or a political interest can create a new privacy risk even if the original objective was protective.

French online-speech case law has repeatedly required a balance rather than a one-way priority. In First Civil Chamber decision no. 18-14.675 of 5 June 2019, the Court dealt with the relationship between personal data, online visibility and the public interest. In decision no. 15-17.729 of 12 May 2016, it addressed the treatment of information in online press archives and the conditions in which access or indexing may be challenged. The common lesson is practical: the operator must identify the interest served by a restriction, the data or speech affected, the duration of the impact and the less intrusive alternatives.

The same balance appears in First Civil Chamber decision no. 13-12.464 of 10 September 2014, which concerned the use of a person’s name and online indexing features, and in decision no. 17-22.381 of 11 July 2018, involving the relationship between online expression, personal rights and responsibility for published material. These decisions do not decide the 2026 statute, but they make one point difficult to ignore: a system that identifies, filters or limits users must be designed around a documented rights balance, not around an assumption that a platform can collect whatever data makes enforcement easiest.

The current French framework already illustrates the difference between a legal objective and a technical method. Article 6-7 of the Law on confidence in the digital economy, introduced in the context of minors’ access to certain services, uses parental authorisation, information duties, suspension requests and technical solutions referenced by the communications regulator after consultation with the data-protection authority. Whether that mechanism is sufficient for social media is a separate question. Its structure at least makes the actors, procedure and oversight visible. A new general rule must be at least as precise when it extends across lawful communications.

For a platform operating in several countries, the most sensitive issue will be purpose limitation. Age status may be needed to determine access to one feature, but it does not follow that the platform may retain the user’s identity, date of birth or government-document image for advertising, recommendation, account recovery or law-enforcement requests. The compliance design should produce the smallest possible result: for example, an independent provider confirms that the user is above or below the relevant threshold, then returns a short-lived attribute or cryptographic proof without disclosing the underlying identity. That architecture is not automatically lawful, but it gives the legal analysis a defensible starting point.

There is also a risk of function creep. A service that stores age information for French access control may later be asked to share it with another regulator, use it to rank content or apply it to users outside France. The next law must state whether such reuse is forbidden, when it can be ordered by a judge, how long an audit trail may be kept and who bears responsibility for a supplier’s failure. Without those boundaries, the system’s privacy cost grows over time and the original child-protection rationale becomes a general identity infrastructure.

That is why the decision has global reach. The issue is not whether a French website can read a passport. It is whether an international platform can operate a trustworthy proof-of-age layer without making identity a condition of ordinary speech. The answer will be examined under French constitutional law, the GDPR, the DSA, contractual data-processing rules and, in practice, the laws of the jurisdictions where the proof provider and platform store or process information. A compliant launch therefore requires a cross-border data map before a product rollout, not a privacy notice after it.

II. What must platforms and lawmakers do before France and the EU try again?

A. How the DSA and the EU age-verification framework change the compliance baseline

France is not legislating on an empty field. The Digital Services Act already provides a European baseline for services accessible to minors. Article 28 of Regulation (EU) 2022/2065 requires platforms accessible to minors to adopt measures that protect privacy, safety and security. The provision uses a deliberately qualified standard: “Providers of online platforms accessible to minors shall put in place appropriate and proportionate measures to ensure a high level of privacy, safety, and security of minors, on their service.” The same article says compliance does not require a platform to process additional personal data solely to assess whether a recipient is a minor.

That formulation does two things. It confirms that child safety is a binding compliance subject rather than a voluntary promise, and it prevents safety from becoming a blank cheque for data collection. A platform must assess its audience and service design, but it must choose measures that fit the risk. The DSA does not itself impose a uniform under-15 ban. It creates duties of safety, privacy and proportionality that national measures must respect. Very large online platforms also face additional systemic-risk and mitigation duties under Articles 34 and 35, including the need to evaluate and reduce risks linked to the design and use of the service.

The European Commission’s age-verification work points in the same direction. Its EU age-verification policy page describes a privacy-preserving blueprint intended to let a person prove an age threshold without disclosing identity or browsing history to the service. The Commission’s technical model is compatible with the European Digital Identity Wallet and can use a trusted proof provider. The objective is not to make every platform hold a copy of an identity document. It is to create a separable proof that answers a narrow question and leaves unrelated personal information out of the transaction.

The Commission’s Recommendation (EU) 2026/1035 adds a policy timetable: Member States are encouraged to make privacy-enhancing proof-of-age solutions available by the end of 2026, either through a European Digital Identity Wallet integration or a standalone solution. A recommendation is not the same as a directly applicable regulation, and it cannot cure a defective French statute. It does, however, give legislators a ready-made direction. A revised bill that ignores a European privacy-preserving architecture will need to explain why it is imposing more intrusive processing than the solution being developed at Union level.

The GDPR adds its own limits. Article 5(1)(c) of Regulation (EU) 2016/679 expresses data minimisation; Article 8 governs a child’s consent in relation to information-society services; Article 25 requires data protection by design and by default; Article 32 requires security appropriate to the risk; and Article 35 may require a data-protection impact assessment where the processing is likely to create a high risk. The practical effect is cumulative. A national age rule does not eliminate the controller’s obligation to identify a lawful basis, define the purpose, minimise the data, protect it, inform the user and demonstrate why the chosen design is justified.

International platforms should therefore distinguish age assurance from identity verification. Age assurance may use an identity document, a bank or public-authority attribute, a telecom signal, a trusted third-party estimate or a device-level parental control. Each method has different error rates, exclusion risks, security consequences and legal bases. A controller should not present them as equivalent. The product file should record why the chosen method is reliable for the threshold, how it treats people without the relevant document, how it handles false positives and false negatives, and whether a human review or appeal is available.

The DSA also changes the relationship between a national regulator and a platform’s risk report. A company cannot satisfy Article 28 with a generic statement that children are protected. It must connect the measure to the service’s actual design: recommender systems, direct messaging, livestreaming, advertising, geolocation, contact discovery, content moderation and account recovery. For a very large platform, that assessment should sit beside systemic-risk analysis, independent testing, incident records and mitigation decisions. The French constitutional file will ask a similar question from the other direction: what evidence shows that a universal ban is needed for the identified risks?

Data governance must also be transnational. A French platform subsidiary may determine the purpose of age checks while a parent company chooses the vendor and a provider in another Member State operates the proof service. The controller and processor agreements must identify who can access raw documents, who can issue the age token, who can revoke it, where logs are kept and what happens after a user deletes an account. The security design must cover not only the final platform database but also the identity provider, software development kit, analytics layer, customer-support interface and regulator-access process.

The most credible architecture will be one in which the platform learns less. A proof provider can return a yes-or-no result for “over 15” with a limited validity period. The platform can record that a threshold was satisfied for a particular action without storing the user’s date of birth. The proof provider can keep the identity relationship separate and subject to its own legal duties. If the user contests an error, the provider can offer a correction route without disclosing more information to the platform than is necessary. This model still needs independent testing, but it aligns the technical answer with the DSA’s proportionality and data-minimisation logic.

B. What a legally defensible French rewrite must prove about risk, parents, data and remedies

A new French bill should begin with a service taxonomy, not a slogan. It should identify the functions that generate the risk and explain why an age threshold is needed for those functions. A platform that hosts public short-form video, permits algorithmic recommendations to unknown users and enables direct contact may require stronger controls than a closed school community or a professional network used for a regulated activity. The distinction need not be perfect, but the law should expose the criteria so that Parliament, regulators, courts and platforms can test them.

The supporting impact assessment should be evidence-led. It should state the harms measured, the age groups affected, the studies relied on, the expected benefit of the proposed intervention and the costs imposed on lawful communication. It should compare less intrusive tools: default private accounts, limits on direct contact, recommender-system changes, advertising restrictions, time controls, reporting routes, parental dashboards and targeted exclusion of high-risk functions. If a blanket ban remains the preferred option, the government must explain why those alternatives cannot achieve the objective and why the same answer is appropriate for every covered service.

The age-assurance chapter should then establish strict boundaries. The statute should name the permitted categories of proof or authorise a regulator to approve methods against public criteria. It should prohibit the platform from retaining the underlying identity data when a threshold token is sufficient. It should require purpose limitation, short retention, encryption, access logging, supplier audits and deletion. It should state whether a proof may be reused across services and countries. It should also address the person who has no passport, no national digital identity, no bank relationship or no compatible device. A right that depends on an unavailable document is not a neutral right.

Privacy by design cannot mean privacy by aspiration. Before deployment, the controller should complete a documented impact assessment covering minors, adults, vulnerable users and people whose identity data may expose them to danger. The assessment should test whether a platform can operate a separate proof channel, whether age tokens can be linked back to accounts, whether support staff can see raw identity data and whether a supplier can infer a user’s activity. The result should be available to the competent regulator and sufficiently explained to users. Where the risk cannot be reduced, the platform should not launch the feature until the risk is resolved or the service is redesigned.

The parental role requires equal care. A parent should be able to understand the rule, configure a child’s access where the law permits it, request suspension and contest an incorrect classification. But parental authorisation cannot be treated as a substitute for every other safeguard. It does not remove the child’s privacy interests, justify indefinite monitoring or allow a platform to collect more data than needed. The law should separate parental control from parental surveillance and provide a procedure where the interests of a child and an adult representative diverge.

Due process is part of the product. A user wrongly classified as under 15 must receive a reason, a usable appeal channel and a decision within a defined time. The appeal should not require the user to disclose the same sensitive document repeatedly to different contractors. A platform should preserve access to emergency, educational and public-interest information while the dispute is pending, unless a specific and evidenced risk requires an interim restriction. A regulator or court must be able to review the decision, the evidence and the technical model rather than receiving only a conclusion that the automated tool is accurate.

Recent French decisions reinforce the importance of that reviewable balance. First Civil Chamber decision no. 25-14.228 of 3 June 2026 illustrates the Court’s continuing attention to the competing interests involved in the online preservation and dissemination of information. Decision no. 19-24.780 of 17 February 2021 addresses the significance of information made available online and the expectations that can reasonably attach to its use. The message for a new age-control system is not that every error creates liability. It is that a restriction must have a transparent reason, a bounded effect and a route to correction.

The protection of minors must be operational too. Criminal Chamber decision no. 20-86.318 of 17 March 2021 sits in a body of case law concerned with harmful communications involving minors and the evidentiary question of what a service or defendant knew. That perspective supports targeted prevention, reporting and preservation duties. It does not prove that every minor must be excluded from every social network. The next French framework should combine age assurance with design obligations that address the actual mechanism of harm instead of treating age alone as the complete risk assessment.

Institutional competence must be clear. The law should tell the data-protection authority which privacy questions it must approve, the communications regulator which technical or service obligations it supervises, and the courts which remedies they may grant. It should define how a European Commission or Digital Services Coordinator request interacts with a French order. It should set out notice, observations, appeal and judicial-review periods. Article 6-7 of the existing digital-economy statute shows the value of identifying a regulator, a consultation process and a route to injunction. A revised cross-platform system should build those steps into the statute rather than relying on informal coordination.

For platforms, preparation can begin before the next bill is introduced. The legal and product teams should map every age-sensitive feature, list the data fields currently collected, identify all processors and test a proof-of-age token that cannot be converted into a full identity profile. They should run error and exclusion tests across travellers, dual nationals, refugees, people with disabilities and users whose names or documents do not match automated records. They should model a French-only rule, an EU-wide rule and a global product standard separately. A decision log should record why one option was chosen and what legal assumption would require a redesign.

They should also keep the enforcement record ready. If the service is challenged, the company will need to show its risk assessment, privacy impact assessment, security testing, supplier due diligence, age-assurance accuracy, complaint statistics and remediation decisions. A public transparency report can explain the categories of proof, retention period, appeals and government requests without disclosing security-sensitive details. The evidence must be coherent across the French regulator, a European Digital Services Coordinator and a data-protection authority in another Member State. Inconsistent descriptions are likely to be treated as a governance failure rather than a translation problem.

The cross-border allocation of responsibility should be contractual as well as technical. The platform should define who is the controller for the age decision, who is the processor for the proof, who answers a data-subject request and who notifies a security incident. It should audit subcontracting and prohibit the provider from using age checks to build advertising or behavioural profiles. It should determine how evidence is preserved when a user challenges a suspension, while ensuring that litigation holds do not become an excuse to keep raw identity documents indefinitely. Counsel should review the data-flow diagram with the same discipline applied to a regulated financial or sanctions system.

Finally, the law should contain a review clause. Technology, error rates and the European framework will change. A permanent restriction should not be based on a one-time assumption about platform design. Parliament should require periodic reporting on effectiveness, false positives, exclusion, privacy incidents, child-safety outcomes and litigation. The review should allow a measure to be narrowed when a less intrusive tool becomes available. That is not administrative hesitation; it is how proportionality remains real over time.

Conclusion

Decision no. 2026-911 DC is a constitutional rejection of an overbroad mechanism, not a rejection of the objective of protecting children online. France can legislate again, and the European framework already expects platforms to address minors’ safety. But the next rule must connect the harm to the service function, explain why the restriction is necessary, protect lawful communication and build privacy into the proof of age from the first line of code.

For global platforms, the immediate lesson is to prepare for a narrow, auditable age-assurance layer rather than a universal identity database. For lawmakers, the lesson is to define evidence, oversight, parental involvement, remedies and retention limits in the statute itself. For advisers handling cross-border matters, the issue is both constitutional and operational: a French requirement can alter a product used across the Union, while a European proof provider can determine the privacy consequences of a French access rule.

The next proposal will be judged by whether it can protect minors without making ordinary participation conditional on surrendering identity. A system that minimises data, separates proof from identity, tests its risk assumptions and gives users a genuine route to correction has a chance of surviving both French constitutional review and European enforcement. A system that simply asks every person to identify themselves and leaves the safeguards to a platform’s terms will face the same objection in a different form.

📄 Official circular