RN
In 2021 this was written as a novelty item — a strange thing happening at the edge of a children's game, filed somewhere between gossip and a shrug. Five years of reporting and litigation later, the framing looks badly wrong, and the useful version of the story is not "look what people made" but "why did a platform designed around child users prove so hard to police."
What the underlying problem actually is
The platform is a toolkit. Users build experiences and publish them, and the volume of that pipeline is the product. Everything that follows is a consequence of that design rather than an accident of it.
Sexual content on a service with strict word filters and a young user base does not appear as explicit uploads. It appears as private-server experiences, coded titles that evade keyword matching, and rendezvous arrangements that move to another service. Enforcement removes instances; the pipeline replaces them. That is not a claim about negligence, it is the structural reason moderation at this scale is a treadmill.
The genuinely difficult part is identity. A service built for children deliberately collects almost nothing: no real name, no verified email, no phone number. Safety staff quoted in reporting have described this as the specific reason pursuing offenders there is harder than elsewhere. The anonymity that protects a nine-year-old also protects the adult talking to them, and there is no version of that trade-off that is simply solvable.
What changed after the 2021 coverage
Three things, in order.
July 2024 — investigative reporting. Bloomberg Businessweek published an investigation into grooming on the platform, reporting that US police had arrested at least two dozen suspects since 2018 in cases involving children met there, and describing how the account model frustrates investigation.
October 2024 — a short-seller enters. Hindenburg Research published a report combining an allegation of inflated engagement metrics with a description of sexual content and grooming attempts encountered by its own test accounts. That the same document argued a financial thesis and a child-safety thesis is worth noting rather than dismissing: it made the safety question a market question, which is when it started moving.
2025 — the legal system arrives. Several US state attorneys general filed suit during the year, with Louisiana and Kentucky earlier and Texas and Florida following. In December 2025 the Judicial Panel on Multidistrict Litigation consolidated private cases, finding they shared common questions about parental controls, age and identity verification, adult-to-minor communication, and moderation.
The response, and what it costs
On 18 November 2025 the company announced that facial age estimation would become mandatory for access to any chat feature — first in Australia, New Zealand and the Netherlands from December, then worldwide in January 2026. Users are sorted into age bands — under 9, 9–12, 13–15, 16–17, 18–20, 21 and over — and messaging is constrained to your own band or adjacent ones. The estimation technology had been introduced earlier in 2025 for a vetted-contacts feature before being made a condition of communication generally.
This is a real structural change rather than a policy statement, and it is the first time a major gaming platform has made a biometric check the price of talking to anyone. It also converts one problem into another. A company that previously held almost no identifying data about its child users now runs face scans on all of them, and age estimation is a statistical inference rather than a fact — it has an error rate, and the errors fall on real accounts in both directions.
Whether it works is an empirical question that will be answered by whether prosecutions and reports fall, not by the announcement.
Where the video platforms fit
Recordings and commentary migrate to general video services, which is what the original headline was reacting to. The dynamic there is ordinary: explicit material violates the terms and gets removed when detected, while material that is suggestive, satirical, or simply about the phenomenon sits in the enormous grey zone that keyword search surfaces and automated review does not clear.
The important distinction, and the one the 2021 article failed to draw, is between commentary and content. Reporting that a moderation failure exists is legitimate and necessary. Material depicting sexualised child avatars is not on a spectrum with it — it is child sexual exploitation material in a different rendering engine, it is criminal to produce or possess in most of the world, and no editorial framing changes that.
What is actually worth taking from this
For anyone with a child on the platform, the practical controls are the ones that matter: an accurate birth date on the account, chat restricted through the parental settings, and the knowledge that the risky step is almost always a move to a private server or another app where none of these settings apply.
For everyone else, the value of the case is as evidence. It is now the best-documented example available of a platform whose core design — anonymous accounts, user-generated everything, a young user base, and growth measured in engagement — produced a safety failure that took an investigation, a short-seller, and a dozen attorneys general to force a structural response. The lesson generalises well beyond one company, and it is a considerably more interesting story than the one this article used to tell.