A six-year-old is holding her mother's phone. If the proposal passes, there will be a rule for what she can do with it: children's videos, through the parent's account, up to one hour a day.
That is the easiest part to understand. The bigger part sits underneath.
Fact
The facts: on 17 September 2026 the European Commission adopted a proposal for the EU KIDS Act. Under the proposal, children under 13 have no access to social media; from 3 to 13 they can use dedicated child-friendly video services through an account managed by a parent, up to one hour a day. From 13 to 15 - mini accounts under parental control, with limited contacts and up to one hour of screen time a day. An account of their own - from 15. For everyone under 18, social media, video sharing, online games, AI companions and chatbots must remove addictive features and profiling-based recommendations, infinite scroll without stopping points, reward tricks and notifications during sleeping hours, as well as unsolicited contact from strangers. AI companions and chatbots are off by default and may not simulate a personal relationship in a way that creates emotional dependency. Minors' profiles are private by default, with location, camera and microphone off. Online services and app stores must use age assurance tools, for example the EU age verification app, which keeps no documents or biometrics; when a new social media account is opened, verification is mandatory. Very large platforms carry the burden of proving their services are safe by design, with a compliance plan and an independent auditor, and under the expedited procedures the Commission must conclude investigations within 90 days. Enforcement builds on the structures of the Digital Services Act and the Artificial Intelligence Act. The proposal has been sent to the European Parliament and the Council.
The AI companion line
The proposal builds on the report of an expert panel delivered in July. The most interesting part for me is the line on AI companions: a chatbot that stays off until someone consciously switches it on, and that has no right to simulate a close relationship in a way that makes the child dependent on it.
The machine may help the child. It may not bind the child to itself.
Reversing the burden of proof is the other big thing. Until now the regulator had to prove that a platform does harm, as in the TikTok case in July. Under the proposal a very large platform will have to prove that it does no harm, with a compliance plan, and an independent auditor will assess the new service or feature.
None of this is law yet. Parliament and the Council have yet to examine it, and the Commission writes that it wants swift adoption. If you have a child under 15, watch for when the text reaches a vote, because it will apply across the whole EU, which means here too.
The visual is generated code-art, no third-party images.