Australia was the country everyone else was watching. Its under-16 social media ban took effect on 10 December 2025, and governments from London to Kuala Lumpur said they would judge their own plans by how it went. The first official three-month evaluation is now out, and the headline number is not the one supporters wanted: more than eight in ten teenagers are still on social media, and the regulator blames the platforms, not the children.
What the report actually found
The evaluation was run by eSafety, Australia's online safety regulator, comparing a baseline survey taken before the rules began with follow-up data three months in. Use of at least one age-restricted platform fell from 85.9% to 81.5%. Account ownership dropped more visibly, from 52.4% to 42.1%, a decline of 10.3 percentage points, with the clearest reductions on YouTube, Snapchat and TikTok.
So accounts were deleted in real numbers. Access was not. Most children in the 10 to 15 age group were using social media about as often in March as they were before the ban, and parents were less aware of what their children were doing than before.
How the teenagers stayed on
The breakdown of workarounds is the most useful part of the report, because it shows where the system leaks:
- Just over half said the platform never asked them to confirm their age at all.
- 37.1% simply stated they were 16 or older and were let through.
- 18.2% went through an age check that wrongly cleared them as over 16.
- 12% tried to fool face-based age estimation with makeup or a false ID, and 10.9% had help from a parent.
- 7.4% used a VPN.
eSafety Commissioner Julie Inman Grant put the cause plainly: the primary reason the ban has moved so little is ineffective implementation of age assurance by the platforms themselves.
Why this is not the VPN story it will be sold as
Expect the 7.4% figure to travel further than any other number in this report, because it is the first official government statistic on VPN use to get around an age restriction, and it lands in the middle of an active debate about restricting VPNs themselves.
Read in context, it argues the opposite. A tool used by 7.4% of children is not what kept 81.5% of them online. The gap between "no age check at all" and "used a VPN" is roughly seven to one. Blaming circumvention tools for the failure of a control that was never applied is a convenient way to avoid the harder conclusion in the report: the platforms did not do the checking.
What happens next
The government is not treating this as a defeat. Assistant Minister Andrew Leigh compared the law to a minimum drinking age and said nobody expected 100% compliance, pointing to the millions of accounts that were closed. Canberra says it will double down on enforcement and higher penalties for platforms that do not verify properly.
The report also shows early signs of migration to services outside the restricted list, with Reddit and Pinterest named. That is the pattern every blocking regime eventually meets: traffic does not stop, it moves to whatever was not on the list when the list was written.
What other countries take from this
Australia is now the reference case for a growing group of jurisdictions writing the same kind of law, and we have tracked each of them: Malaysia's under-16 ban with biometric verification, the UAE's rules for under-15s and New York's finalised age-check regime. The running list of countries and age thresholds is in our guide to social media bans for minors.
The honest reading for anyone in these countries is that age assurance and privacy are on a collision course. When self-declaration fails, the next step proposed is always stronger identification: document scans, face estimation, bank checks. Each of those creates a new database that ties a real identity to an account, which is exactly why encrypted connections and minimal data collection keep coming up in the same conversation. A VPN changes which network sees your traffic, it does not make an identity check disappear, and Australia's own numbers show it was never the main way around this one.
