← All posts

Artificial intelligence

Europe's Under-13 Ban Got the Headline. The Kids Act Also Puts AI Companions Inside It

Von der Leyen announced the age limits in Strasbourg on Wednesday; the Commission presents the Kids Act on Thursday. The numbers are negotiable — the architecture underneath, which pushes age checks to the app store and pulls AI companions into child-safety law, is not.

MAI
The European Commission's official key visual for the State of the Union 2026 address, published on the Commission's website.

Ursula von der Leyen spent eleven words of her State of the Union address in Strasbourg on Wednesday setting up Europe's next fight with the platform industry: "No social media under the age of 13. No personal account under the age of 15." The Commission presents the legislation behind that line — the Kids Act — on Thursday. The two numbers will carry the coverage. They are the least consequential part of the proposal.

What the draft covers

A draft of the regulation was reported by Euronews on Tuesday, ahead of Thursday's formal presentation, so the detail below rests on that leaked text rather than a published Commission document. It describes a tiered scheme rather than a single cut-off:

AgeAccess
Under 3Excluded from social media and other high-risk services
3–13Child-designed services only, under adult supervision
13–15Social media and video platforms with limited features and parental controls
15–18Access without parental consent; safety-by-design duties still apply

Scope is where it gets interesting. The draft covers social media, video-sharing platforms and online games — and AI chatbots and companions. Services built for education, public-authority platforms, and industrial or office AI systems are carved out. Reporting on the proposal also places app stores among the surfaces that will have to perform an age check.

Pulling companions in is the substantive move. AI companion products have so far been governed by the AI Act's transparency duties and by whatever each vendor decided its own guardrails should be. The Kids Act would treat a companion chatbot as a child-safety surface with binding design obligations, including protection against what the draft calls unhealthy emotional attachment. That is a regulator asserting that the engagement loop itself is the hazard, not the output — a position no major jurisdiction has yet written into law for this product category.

The design rules are product rules

For everything in scope, the draft bans infinite scrolling, artificial notifications and certain reward mechanisms, and requires recommender systems built to avoid rabbit holes, risky settings off by default, minors' accounts private by default, and restrictions on contact from unknown users.

None of that regulates content. It regulates interface and retention mechanics, which is a harder thing for a platform to litigate and an easier thing for a regulator to observe. The enforcement clause goes further still: services designated as very large platforms under the Digital Services Act would need Commission authorisation before deploying new features that affect children. Prior approval is a different regime from after-the-fact investigation, and it is the provision the industry will spend the most money trying to remove.

Verification moves down the stack

Age would be confirmed through an EU-wide verification app or equivalent national schemes. The Commission's own app — a zero-knowledge "mini-wallet" that attests to an age threshold without handing over identity documents — has been technically ready since April. Its privacy design is better than the document-upload alternatives, and critics including the Electronic Frontier Foundation have still argued that a cryptographic proof of a credential says nothing about who is holding the phone.

The structural consequence is the part worth watching. If the check sits at the app store, Apple and Google become the compliance layer for everyone else's obligations. Both have argued for two years that age assurance belongs with the platforms that run the feeds; Meta has argued the opposite, that it belongs at the operating-system and store level. The Kids Act, as described, settles that argument closer to Meta's position without ever framing it as a concession.

Six percent, and the years before it

Non-compliance carries fines of up to 6% of global annual revenue, with supervisory fees on top. That is DSA-grade money, and it is also the least certain number in the package, because none of this is law yet.

The proposal now enters the ordinary legislative procedure with Parliament and Council. Estonia and Belgium have opposed age-based bans outright. Parliament has previously pushed for 16 rather than 15. Member states are split on whether gaming belongs in scope at all. A file of this size takes years, and the version that emerges will not be the version presented on Thursday.

Which is why the age limits are the wrong thing to read closely. Thirteen and fifteen are negotiable and will be negotiated. The architecture underneath — age checks pushed to the store layer, engagement mechanics regulated as product defects, and AI companions placed inside the child-safety perimeter — is what platform and model companies should be costing out now, because those three ideas are unlikely to fall out of the text no matter where the numbers land.

Sources: Euronews — Leak: EU Commission to pitch social media restrictions for under-15s · The Next Web — The Commission president proposes no social media for children under 13 in the EU · CNN — EU announces plan to ban social media for under 13s · Agence Europe — European Commission to present its 'Kids Act' on Thursday · European Commission — State of the Union 2026 · European Commission — The EU approach to age verification · MacRumors — EU's Kids Act to Ban Social Media for Under-13s · EFF — Age Verification in the European Union: The Commission's Age Verification App

Keep reading