EU KIDS Act: Brussels Seeks Tougher Rules for AI Chatbots Used by Children

The European Commission wants AI chatbots to exclude previous conversations with minors by default. It also wants to bar features that simulate a personal bond and are likely to make children emotionally dependent. The EU KIDS Act also proposes a minimum age of 15 for independent accounts on certain social media and video platforms.

Article 14 of the EU KIDS Act proposal covers general-purpose chatbots and AI companions that minors may use. An AI companion is a chatbot designed to provide company or conversation. The Commission is examining, among other things, how these systems can mimic human relationships. Providers would have to avoid features and behaviour likely to foster emotional dependence in this way.

The proposal also sets out a separate rule for chat history. By default, a chatbot would have to keep information from earlier conversations with a minor out of future chats. The only exception would be when that information is needed to protect the child. The Commission wants to prevent chatbots from accumulating sensitive information about children or reinforcing harmful conversational patterns over time.

Providers would have to assess the risks their chatbots pose to children before launch and take steps to mitigate them. The Commission would exempt micro and small businesses from the separate requirement to monitor risks after launch.

The Commission proposes a different access rule for general-purpose AI chatbots and AI companions. Children under 13 would be allowed to use a chatbot covered by the proposal only through tools that let a parent or guardian grant and manage access. The proposed design rules would still apply to minors aged 13 and over. OpenAI has already launched a separate ChatGPT experience for teens, with stricter default settings and parental controls.

Brussels is also looking at chatbots young people encounter inside other apps. If a social network, video platform or online game includes a chatbot, its provider would not be allowed to switch it on automatically for minors, give it prominent placement or actively promote it to them. If a young person turns the feature on, they must also be able to switch it off easily. The rules address a trend highlighted earlier by AI Wereld in a survey on Dutch teens asking AI personal questions.

The proposed age limit applies to social networks and video platforms with certain high-risk features. From the age of 15, young people would be able to create an account independently. For 13- and 14-year-olds, a parent or guardian could create a restricted account if the platform offers that option. Parental controls would remain on, parents could approve new contacts in advance, and daily use would be limited to one hour.

For children under 13, the proposal allows a limited exception for video platforms specifically designed for that age group. A parent could grant access through their own account, with a time limit and no personalised feed or search function. The child would not get an account of their own.

For new accounts on the platforms covered, entering a date of birth would not be enough. According to the European Commission’s explanation of the proposal, an independent, certified solution would have to confirm that the user meets the age requirement. The platform would receive only the result of the age check, not an identity document.

The European Commission presented the proposal on 17 September 2026. The European Parliament and the Council of the EU must still review it and agree on the text. The proposed age limits and chatbot rules are therefore not yet in force. The Commission is accepting feedback on the proposal until 3 December 2026.