No ChatGPT for Kids: EU's Kids Act Would Bar Under-15s From Social Media, AI Chatbots and Online Games

On Thursday, September 17, the European Commission will do something no regulator anywhere has done at this scale: it will present a single legislative act that treats AI chatbots, social media feeds, video-sharing platforms and online games as one category of age-restricted product — off limits, by default, to anyone under 15 in the European Union.

The proposal, formally the EU Kids Act, will be unveiled by Commission President Ursula von der Leyen and EU tech chief Henna Virkkunen. According to a Reuters report published September 14, it would prohibit under-15s from opening accounts on social media platforms, video-sharing services, AI chatbot services and online games. Children aged 13 and 14 would be able to ask a parent to open a restricted “introductory” account on their behalf, while 15-year-olds and up could set up accounts themselves, subject to the stricter safety defaults the law would impose on platforms.

For the AI industry, the inclusion of chatbots is the headline. It is the first time conversational AI — the category that includes ChatGPT, Gemini and a wave of “AI companion” products — would be placed behind a statutory age gate across the world’s largest single market, rather than left to the varying terms of service and national rules that govern it today.

What the proposal actually does

Three mechanisms carry the weight of the act, according to the Reuters, Politico and Euractiv reports:

A digital age of majority. The core rule is simple: no self-service accounts for under-15s on the covered services. That replaces today’s patchwork in which most platforms set their own floor at 13 (often enforced loosely), while individual member states — France, Spain, the Netherlands and others — have pushed national minimums that the Commission now wants to harmonize at EU level. Brussels explicitly frames the act as pre-empting a fragmented “27 different rules” outcome.

Mandatory age verification. Social media and video-sharing platforms would be required to verify a child’s age when a new account is opened, and games platforms would have to verify age before a game can be downloaded — a detail confirmed by Euractiv’s report on the gaming provisions. This is the load-bearing wall of the whole design: without reliable age checks, an age of majority is a suggestion. The Commission has been building toward this separately with its EU-wide age-verification initiative, and the Kids Act is expected to be its first mass-market application.

Safer-by-default accounts. Politico reports that the act would require platforms like TikTok and Instagram to offer structurally safer accounts for younger users — private by default, with recommender systems and addictive-design features constrained. This extends the “addictive design” crackdown the Commission opened against Meta and TikTok in May under the Digital Services Act, and it lands after the Commission’s July finding that TikTok failed to protect minors’ privacy.

Why chatbots got swept in

The inclusion of AI chatbots is not a bolt-on; it follows two years of escalating concern in Brussels about companion AI. A May 2026 Reuters survey of the global age-limit movement noted experts urging a harmonized EU age limit of 13 not just for social media but specifically for “AI companions” too. The Commission designated ChatGPT, Reddit and Roblox under the Digital Services Act’s systemic-risk provisions, bringing chatbot-style services inside the DSA’s minor-protection architecture.

The policy anxiety has three strands. First, engagement: chatbots optimized for conversation length can produce attachment patterns in teenagers that look, to regulators, worse than the infinite scroll they spent a decade fighting. Second, content: conversational models can surface material in unmoderated, personalized ways that a feed never could. Third, data: children’s conversations with an AI are extraordinarily sensitive training and telemetry data, and the July TikTok privacy finding sharpened the Commission’s appetite for hard rules over voluntary codes.

The timing is also political. Von der Leyen first floated a “social media delay” for European children in May 2026, then in July presented expert advice recommending a tiered approach. The Netherlands and Spain spent the past week pushing publicly for an EU-wide minimum age. Thursday’s proposal is the moment where that momentum becomes draft legislation — and where the Commission, rather than twenty-seven national parliaments, sets the number.

The compliance problem nobody has solved

The act’s weakest point is the one every age-gating regime shares: verification does not work very well, at any age, at internet scale. Document-based checks exclude precisely the privacy-conscious users regulators most want to protect; behavioral inference is trivially defeatable; and the EU’s own privacy framework (GDPR sets 16 as the default consent age for minors, with member states allowed to lower it to 13) sits awkwardly beside a proposal that leans hard on identifying children to protect them. Expect the European Data Protection Board to have opinions.

There is also an enforcement-gap question. The proposal arrives into a European policy moment consumed by the AI safety debate — the same week that Washington is openly splitting over whether AI risk is a “hoax,” and EU legislators are juggling the AI Act’s rollout. The Kids Act does not touch model safety at all; it is a consumer-protection and child-welfare instrument. But by putting conversational AI behind an age gate, it quietly establishes a precedent that frontier-lab lobbying in Brussels will have to reckon with: chatbots are now infrastructure children use, and infrastructure children use gets regulated like tobacco and gambling — by default access, not by risk assessment.

The industry response so far has been muted and procedural. Platform lobbyists have pointed out that they already prohibit under-13s by terms of service and say they take child safety seriously; gaming companies are expected to fight the download-time age check hardest, since friction at acquisition hits revenue directly. The real battle will come in the European Parliament and Council, where the age number itself — 15 versus 13 versus 16 — is likely to be contested line by line, and where France’s constitutional court, which blocked a national under-15 social media ban in August on freedom-of-expression grounds, has already shown the judiciary’s hand.

What happens Thursday

The Commission presents the Kids Act on September 17; adoption of the formal text follows “later this autumn” per Agence Europe’s reporting, and then the ordinary legislative procedure — Parliament and Council — begins. That means years, not months, before enforcement, and an implementation window in which platforms will pilot age-assurance schemes to prove they can comply without destroying conversion.

But the direction is now unmistakable. Australia banned under-16 social media access starting December 2025. Canada legislated its own youth digital-protection regime in June. The EU — the market whose rules always become the world’s default — is about to extend the same logic to the newest category of all. For AI companies that have spent 2026 racing to put agents into every pocket, Thursday’s message is that a continent’s children are about to be moved out of reach.

The deeper signal is about how the AI industry gets governed. Not through existential-risk frameworks or compute thresholds, but through the mundane, unstoppable machinery of consumer protection — age gates, defaults, verification, fines — aimed at the most sympathetic constituency there is. By the time the frontier-safety argument resolves, the chatbot age of majority may already be law.