The youth-ban coalition is coming apart, and the compliance work is moving to design

A ban and a design duty are not the same build, and the difference is where the check lives. A hard age gate is one choke point at registration: one team owns it, and it either holds or it does not. Duties written to cover "all the online places where young people will be spending their time" land on defaults, contact permissions, recommender behaviour and reporting paths in every surface that has them, including assistant features nobody classed as social when the roadmap was drawn ^[9].

The circumvention finding is the part worth reading twice. If a large share of minors walk past the gate ^[4], a platform has paid the full verification bill and still has its under-18 population, now recorded as adults in every downstream system that reads the age field. That is worse than no gate, because the safety team loses sight of the cohort it is accountable for. Coutinho's point about why teens stay is the mechanism: removal does not hold when the product is the communication channel ^[3]. And the users with the fewest alternatives, including the isolated LGBTQ+ young people the free-speech objections name ^[6], are the ones who lose most where it does hold ^[4].

Count the jurisdictions in play. Of the five this reporting touches, exactly one has a ban actually operating ^[12], and the evidence cited from it is circumvention ^[8]. France's version went to the Constitutional Council and did not come back ^[1]. Nothing has advanced in the United States at either level ^[5]. Brazil's own online safety commissioner doubts the approach travels ^[2].

What regulators do win is narrower and easier to miss. Brazil got livestreaming switched off inside Discord after the death of a 13-year-old girl allegedly coerced by other users, a case Discord says was not broadcast live ^[10]. Techdirt's read is that pulling one feature from one platform, when the same technology sits on nearly every other, mostly relocates the next incident to somewhere nobody is watching ^[11]. Whatever that does for children, it tells you what enforcement looks like when a ban is unavailable: feature by feature, surface by surface, one product at a time. That is a design duty arriving by another route.

So the planning assumption that survives is age assurance as a graded signal feeding tiered experiences, rather than a boolean gate, plus obligations that follow minors into chatbots ^[9]. Techdirt's own position is blunter, that bans misdiagnose the problem and worsen it ^[13], but a product team does not need to settle that argument to notice that the ban-shaped roadmap has lost its legal footing in the two places it was furthest along ^[1][5]. The EU proposal expected this autumn is where the alternative gets specified ^[8].