Brussels Draws the Line at Fifteen

The document the European Commission has drafted and Reuters has seen does something no AI regulation has done before: it files the chatbot under the same heading as the feed. A proposal to be put to member states and Parliament would ban social media, video-sharing platforms, AI chatbots, and online games for anyone under fifteen — tiered down to parent-opened introductory accounts for thirteen- and fourteen-year-olds, with age verification required and a possible supervisory fee on companies to fund the enforcement itself. The AI industry’s consumer interface has just been assigned its first child-safety regime, drafted by people who remember what social media did.

The tiering is the tell. This is not a blunt under-16 switch like Australia’s; it is a ladder — fifteen and up get accounts of their own, thirteen and fourteen get what the draft calls introductory accounts opened by a parent, with limits, and below that the access narrows toward nothing. Ladders mean the drafters have been reading the research: harm is a function of age, exposure, and the absence of an adult in the loop, not of the platform class. What is genuinely new is which platform class got named. AI chatbots appear in the same sentence as TikTok-class video apps — the first time a major regulator has put companion-style conversation products inside a child-safety perimeter rather than beside it.

The politics arrived faster than the law. On September 8, President Macron wrote to von der Leyen urging exactly this floor; the Netherlands and Spain have been pushing an EU-wide minimum age since last week; Australia’s under-16 ban in August gave the bloc a working precedent and a talking point that only the largest market can make verification workable. The proposal still needs member states and Parliament, and the draft could change before it is even published. But the direction of travel is unmistakable: the age at which Europe thinks a child should meet an algorithm has become a matter for legislation, not terms of service.

The Chatbot Clause

The chatbot clause deserves its own reading, because it arrives at the exact moment the consumer-AI industry is betting its growth on minors’ older siblings — students, first-jobbers, teenagers a year or two over the line. Companion products have spent two years optimizing for engagement metrics that look, from a regulator’s distance, uncomfortably like the ones social media optimized for a decade ago. Brussels’ answer is to treat the conversation itself as the risk surface: not data collection, not recommendation, but the relationship. If the Kids Act passes in this shape, every AI assistant sold in Europe will need an age gate as rigorous as anything social media ever built — and the supervisory fee means the regulator will be standing behind it, funded by the platforms it watches.

What happens next is procedural but consequential: the Commission proposal goes to member states and Parliament, where the age line itself will be the fight — fifteen here, sixteen there, thirteen everywhere in industry’s counterproposals. The verification burden will be the second fight, and the fee the third. But the Overton window has already moved a full stop: the question is no longer whether AI chatbots will be regulated for children, but at what age the conversation becomes legal. An industry that spent two years shipping companions now has a date to circle — and it is not their own launch calendar that set it.