France's top court kills the under-15 social media ban

France's Constitutional Council struck down the country's ban on social media for under-15s this afternoon, seventeen days before platforms were due to start enforcing it. Decision n° 2026-911 DC censures Article 1 of the law outright. The reasoning is the part worth reading anywhere age-assurance law is being drafted: a rule that applies only to children cannot be enforced without asking every adult to prove they are not one, and the French Parliament never specified how that proof was supposed to work.

The referral came from Socialist and La France insoumise deputies in late July, days after the text cleared Parliament. Two saisines carrying sixty signatures were enough to put Article 1 in front of the Council. It took the full month it is allowed, then killed the article.

RelatedChrome will warn on every HTTP site by default in 2026

What exactly did the Council censure?

The law, visant à protéger les mineurs des risques auxquels les expose l'utilisation des réseaux sociaux, would have barred anyone under 15 from holding a social media account. New accounts were blocked from September 1, 2026. Accounts already held by under-15s had to be closed by January 1, 2027. Enforcement sat with the platforms, which were told to verify age and given nothing else: no approved method, no technical standard, no named regulator in the operative text.

That gap is what broke it. The Council held that lawmakers could not impose a prohibition of general scope, stripping minors of access to a broad set of online communication services, without regard for the individual minor's situation or for the specific risks attached to each service. Restrictions on liberty have to be nécessaires, adaptées et proportionnées to the objective. A flat cutoff at 15, applied identically to a photo-sharing app and an algorithmic video feed, met none of those tests.

Then came the finding with reach beyond France. By prohibiting under-15s from certain services, the Council wrote, the law inherently requires every person, adults included, to prove their age before accessing them. Having created that obligation, the legislature failed to set the conditions and limits under which age would be established, and so provided no legal guarantees for the privacy and expression rights it was putting at risk. Article 1 fell on both grounds at once.

• The ban itself is dead. September 1 arrives with nothing to enforce, and the January 2027 account-closure deadline goes with it.
• The rest of the law stands. The extension of the school phone ban up to lycées was not challenged and survives intact.
• A rewrite is already commissioned. Emmanuel Macron has asked Prime Minister Sébastien Lecornu to prepare a new text addressing the Council's objections.
• France is no longer first in the EU. Australia's under-16 ban, live since December, remains the only national law of this kind actually in force anywhere.

Why does a ban on children become an ID check for adults?

This is not a French quirk, and it is the bit most coverage glides past. A service cannot apply a rule to under-15s unless it can tell who is under 15. Everyone arrives as an unknown. There is no way to filter the minority without interrogating the majority, so a 34-year-old who was never in scope of the ban still has to hand something over to demonstrate that fact.

What that something is decides whether the design is livable. Uploading a passport photo to a short-video app is one answer, and a bad one: it turns every consumer app into a repository of government identity documents. A cryptographic attestation issued once by a wallet or a device, then presented as a bare over-15 claim that reveals nothing else, is a very different answer. The distance between those two designs is the whole privacy question, and the French text left it blank for platforms to fill in however they liked.

Read carefully, the Council did not rule age assurance unconstitutional. Its complaint is procedural in the sharpest possible way: if you are going to make forty million adults identify themselves to read a feed, the conditions and the limits belong in the statute, not in a platform's product roadmap.

1. Jan 26, 2026Assemblée nationale adopts the bill at first reading 130 for, 21 against
2. Jul 21, 2026Parliament adopts the joint-committee text definitively Sénat 243 to 2 with 100 abstentions, then Assemblée 279 to 81
3. Late Jul 2026Socialist and LFI deputies refer Article 1 to the Conseil constitutionnel two saisines, sixty signatures
4. Aug 14, 2026Decision 2026-911 DC censures Article 1 disproportionate restriction on freedom of expression
5. Sep 1, 2026The date new under-15 accounts would have been blocked now moot
6. Jan 1, 2027The date existing under-15 accounts would have been closed now moot

How does France compare with the countries that already tried this?

The striking thing about the parliamentary record is how little opposition the ban faced. The Sénat passed it 243 to 2. The Assemblée passed it 279 to 81. A measure with that kind of cross-party support died anyway, because political consensus about protecting children says nothing about whether the drafting survives a proportionality test. Other countries have written the same idea very differently.

Denmark is the closest to following France's original path, with its own under-15 proposal advancing. The United Kingdom and Canada have both floated comparable bans without legislating one. The EU has taken the opposite tack: rather than a statutory age floor, it has been building the plumbing, with Commission guidelines under Article 28 of the Digital Services Act and a reference age-verification app whose specification mandates hardware-bound attestation. That sequencing now looks a lot smarter than France's.

RelatedAI Browsers Are Coming for Chrome's 90% Grip

What should platforms building age assurance take from this?

Three things, and none of them is that age checks are going away. First, the legal risk has moved from whether you check to how you check. A vague statutory duty is not a shield; it is now demonstrably a liability, because a court can strike the duty itself for being underspecified. Second, data-minimising designs are the ones that will survive review. A system that learns only that a user is over some threshold, and retains nothing, is far easier to defend as proportionate than one that collects documents. Third, per-service risk assessment is coming whether platforms want it or not: the Council specifically faulted the law for treating every service as equally dangerous, which cuts against blanket rules in both directions.

There is also a practical point for anyone shipping in the EU this autumn. Product teams that built French age-gating against the September 1 date now have that work stranded, but not wasted. Lecornu's replacement text is coming, and the Council has effectively written its specification: name the method, bound the data, differentiate by service.

Our take

France did the popular thing and skipped the hard part. Banning under-15s from social media polls well and drafts easily; specifying a privacy-preserving way to establish that someone is 15 is genuinely difficult engineering and politics, so Parliament passed the first and left the second to companies it does not trust. The Council caught it. Good.

What is frustrating is the timing. The technical answer to this problem is further along than the legislation suggests: attestation-based age proofs that leak nothing about identity are being specified right now at EU level. Had the French text pointed at that machinery instead of writing ‘verify age’ and walking away, the ban would probably have survived. Lecornu gets a second attempt with the Council's marking scheme in hand. There is no excuse for the next version to be this vague.

Original analysis by GenZTech, built from the Reuters report on the ruling and the Conseil constitutionnel's own decision.