Under Australia's child protection law, Meta used artificial intelligence to estimate user age and took action on roughly 750,000 accounts assessed as belonging to users under 16. The figures cover the first six months of the world's first ban on social media use by under-16s.
Meta said the action was intended to block minors' access to the platforms and strengthen legal compliance and user safety.
What makes this significant is how it was done. The enforcement did not run on reports or ID verification — it ran on AI-based age estimation at scale.
Two implications follow.
1. Self-declared age loses its authority. Platform age policies have historically leaned on the birth date entered at signup. Once behavioral inference is used for actual enforcement, entering a different birth year stops working.
2. False positives become the operational problem. Probabilistic estimation means adult accounts get deactivated incorrectly. The appeals process — and how fast it resolves — determines the real quality of the system.
This is one country, but regulatory movement is running in parallel across several markets.
Read alongside the social media addiction lawsuits allowed to proceed against Meta — see Meta's Social Media Addiction Lawsuits Can Proceed — the direction is not ambiguous.
Australia is the test bed. Regulators in other markets are reading these six months of data, and so are the platforms.
Approximately 756,000 in six months — about 460,000 on Instagram and 290,000 on Facebook.
Meta used artificial intelligence to estimate user age, identifying suspected under-16 accounts through inference rather than the birth date entered at signup.
It bans social media use by anyone under 16 — the first such law in the world.
Teen audience reach contracts, age inference models can change what age-based targeting actually captures, and creators with heavily teen audiences lose reach at the base.
To apply what you just read to your own site, start with a free audit of where things are now.
A strategist replies within 24 hours on business days.
YouTube has improved AI likeness claim alerts and simplified disputes. The copyright menu consolidates into 'Claims,' and creators get four grounds to push back.
Meta's Edits app adds project folder reordering, still-image-to-video conversion, text style bookmarks, and richer templates. A video editor is quietly becoming a production workflow tool.
More than a million people used LinkedIn's AI slop reporting option in its first two weeks, and views on slop-classified content fell 40%. But reporting is feed control, not enforcement.