The Safety Ratchet

You see it on the late train before you read a single government PDF. A kid, maybe thirteen, maybe fifteen, phone lit under the chin, thumb still working the same apps the press release said were closed. That is a street signal, not a survey. It does not prove how many kids are still online. It proves the country is living inside a gap between the law’s promise and the carriage’s glow.

On 10 December 2025 Australia switched on the social media minimum age. Platforms had to take reasonable steps so under-16s could not hold accounts on age-restricted services. Children were not fined. Parents were not fined. The target was the account, not the glance over a mate’s shoulder. That distinction matters, because most of the public argument still collapses three different measures into one word: ban.

Account ownership is what the statute aims at. Platform use is what you can still do without your own login. Harm reduction is the moral warrant that sold the law. Those are not the same thing. Treat them as the same thing and every side gets a cheap win.

Three months in, the government’s own evaluation drew the line the politics did not want. eSafety’s early insights report, built with the Stanford Social Media Lab and an academic advisory group across more than four thousand children, put ownership of at least one age-restricted platform among 10-to-15-year-olds at 52.4 percent before the gate and 42.1 percent after. That is a real drop. It is not a collapse. Reported use moved from 85.9 percent to 81.5 percent. Daily-plus use barely budged. Among kids who still held accounts, about half said the platform had not age-checked them. Parent pulse numbers in the March 2026 compliance update were sharper on ownership and softer on the street: most parents whose child had an account before the gate said the child was still on Facebook, Instagram, Snapchat or TikTok, and the top reason was blunt. Nobody asked them to verify age.

Platforms also reported millions of accounts removed or restricted in the first weeks. Those are account events, not unique children, and eSafety was careful not to treat the raw count as proof of reasonable steps. Pureprofile’s Wave 2 commissioned survey, fielded around April 2026, put regular access to banned platforms by under-16s near four in five. Different method. Same direction. The gate moved some logins. The glow stayed.

What has not been proved is the thing the press conference needs most. The three-month evaluation said it was too early to measure long-term harm reduction. Broader wellbeing, offline activity and problematic-use measures showed little population-level change. Cyberbullying and image-based abuse complaints involving under-16 accounts did not show a notable drop in the early 2026 window against the prior year. That is not a verdict that the law failed. It is a verdict that Australia enlarged the enforcement machine before the outcome machine had results.

The government’s case still deserves the steelman. If the statutory duty is reasonable steps to keep under-16s off accounts, and major platforms are still carrying under-age logins because they have not built serious age assurance, then weak compliance is not a vibe. It is evidence. On 28 June 2026 the Prime Minister and Communications Minister Anika Wells announced stronger powers and doubled penalties, arguing platforms were not doing enough and eSafety needed more tools. Wells’s second-reading speech, as reproduced in the Senate committee record, said eSafety was investigating noncompliance by five major platforms and accused companies of dirty tricks and deliberate failure. You do not have to love the politics to accept the regulatory logic. When a licence condition is being gamed, the state usually reaches for information powers and higher fines before it reaches for patience.

Parliament then built the ratchet.

The Online Safety Amendment (Strengthening Enforcement for the Social Media Minimum Age) Act 2026 is Act No. 83 of 2026. Assent 11 September 2026. Whole Act commenced 12 September 2026. Schedule 1 doubles the major civil penalties for the minimum-age duties from 30,000 to 60,000 penalty units. At the 1 July 2026 penalty-unit rate of $364, that is about $21.84 million for an individual and about $109.2 million for a body corporate on the headline provider breaches. Older speeches that floated roughly $99 million were doing the sums on the previous $330 unit. The number moved. The deterrent story stayed.

Section 63G was rebuilt so eSafety can require information or documents from any person reasonably believed to hold material relevant to whether a service that is or may be an age-restricted social media platform is meeting the minimum-age duties. Not only the platform. Any person. Age-assurance vendors and app stores sit in the Explanatory Memorandum’s examples. The text is wider than the examples. A 63G notice cannot haul you in for questioning. That job moved to Schedule 2.

Schedule 2 is the part most explainers skim. It inserts an examination division. eSafety can summons a person believed capable of giving relevant information or evidence, require attendance, take answers on oath or affirmation, and run the examination in private with an adviser present. Fail the notice, refuse the oath, or refuse to answer and you face a civil penalty or a criminal offence carrying up to 12 months. There is a journalist-source carve-out with an evidential burden. Self-incrimination privilege is abrogated for individuals, with use and derivative-use immunity subject to listed carve-outs. High Court witness protections are picked up by reference. This is not a random street power. It is still a serious expansion of who can be pulled into a child-safety compliance fight that began as a platform account rule.

Australians are allowed to hold two ideas at once. Platforms that treat Australian law as optional invite escalation. Escalation that outruns proof invites a different kind of debt. Safeguards exist. Reasonable belief. Relevance. Private examinations. Advisers. Record copies. Journalist protection. Use immunity with holes. What the Act does not yet publish into the culture is a clean accountability dashboard: false-positive rates on age gates, adult collateral, privacy incidents, migration onto apps that were never the political target, and a definition of success strong enough to justify doubling the fine schedule before harm reduction shows up in the government’s own tables.

Then there is the money, which is where soft-merge does the most damage.

The MYEFO social media minimum-age package sits around $76.1 million over four years, with eSafety taking the largest share, plus departmental, public education and OAIC lines. Separately, Budget 2024–25 put $6.5 million in 2024–25 toward an age-assurance pilot. On 15 November 2024 the department announced that a consortium headed by the Age Check Certification Scheme, ACCS, had won the Age Assurance Technology Trial tender, ATM 10029696. That award is documented. The AusTender contract notice value for ACCS is not. After a serious public hunt, the named-beneficiary claim for the trial’s contract dollar figure remains unsubstantiated. Keep the $6.5 million as an appropriation envelope. Do not invent a vendor payday. If Australians want the contract price, that is Freedom of Information territory now, not another vibes round on the internet.

Beside this enforcement story sits a different machine that politics keeps trying to zip into the same suitcase. On 14 November 2024 Michelle Rowland announced Digital Duty of Care obligations for platforms. That was before the minimum-age law was even live, and long before the March and July 2026 compliance evidence. The 2026 exposure draft and the My Feed, My Way branding are about systemic safety duties, risk assessment, design features and user-empowerment tools for recommender systems. The draft’s investigation settings are built to sit apart from the Part 4A minimum-age regime. Duty of Care is a parallel architecture. It is not the legal child of a disappointing age gate. Mention it here only so nobody launders one into the other.

So what is the Australian Interest test?

A country can decide that childhood attention is worth hard law. That decision is not unserious. The street still has to ask whether we are building a proof engine or a permission engine. A proof engine would publish the outcome metrics with the same energy it publishes the penalty multiples. It would say what false positives it will tolerate, what privacy burns it will admit, what migration to messaging and gaming counts as success or failure, and what evidence would trigger winding power back. A permission engine doubles the fine, widens the summons class, and treats the next expansion as natural weather.

The train carriage does not settle the science. The evaluation does not settle the politics. The Act settles the sequence. Before Australia could show that the under-16 intervention reduced harm, Parliament enlarged the powers to enforce it. That may be justified. It is still a ratchet. Watch who gets asked next, who gets paid next, and which number the government is willing to put on the board when the next three-month report lands. That is how a safety story becomes a system.

Rico Holt  ·  ricoholt.com