The EU just wrote down the thing parents have been worried about
There's a line in the European Commission's new proposal that stopped me. AI chatbots and companions, it says, will no longer be able to act in ways that make children emotionally dependent on them.
That's a regulator putting into law the exact thing parents have been trying to describe for a year. Not "my kid sees bad content". My kid has got attached to it.
What was actually proposed
On the 17th September the Commission proposed the EU KIDS Act. It's a proposal, not law, and it now goes to the European Parliament and the Council before anything happens.
Two separate things in it get muddled in the headlines, so it's worth pulling them apart.
The social media rules work by age. Under 13, no social media. From 13 to 15, a mini account managed by a parent or guardian, limited features, one hour a day. From 15, they can open their own.
The AI rules are different. They don't work by age band, they apply to everyone under 18. Chatbots and companions have to be off by default, and they can't be designed to create emotional dependency. Alongside that, under-18 profiles would be private by default with geolocation, camera and microphone switched off, and things like infinite scroll, reward tricks and notifications during sleeping hours would go.
Why companions got their own rules
The Act defines an AI companion as a system that provides sustained, personalised interaction or companionship, simulating a social, emotional or interpersonal relationship.
Read that again, because it's not describing homework help. It's describing something that remembers you, picks up where you left off, and behaves like it's invested in your life. That's the bit teenagers get attached to, and that's the bit the Commission has gone after.
Enforcement for AI companions and chatbots would run through the EU AI Act rather than the rules covering social platforms, which tells you they're being treated as their own category rather than more of the same.
What this means if you're in the UK
Nothing, today. We're not in the EU, this isn't law yet, and your child can still use any of these apps this afternoon.
What's changed is the direction. Regulators have stopped treating "AI chatbot that acts like a friend" as ordinary internet content, and started treating it as its own risk. Australia has already restricted social media for under-16s. This is the next step, and it's aimed squarely at emotional design.
Whether any of it lands, and in what form, is months away at best.
Where that leaves you now
None of this helps you tonight. If you want to know what your child is actually talking to an AI about, that's still on you, and most parental controls can't see any of it.
That's what we built Halo Aware for. Our extension covers Character.AI along with the other platforms we support, and gives you a plain summary of what's been discussed rather than a pile of messages to read.
