Australia keeps regulating technology after the fact: the pattern

Photo: rawpixel via rawpixel (CC0 1.0)
Four laws in nine years, each of them a world first or close to it, each passed quickly, each argued over for years afterwards. The mandatory metadata retention scheme of 2015. The encryption-access laws of 2018. The News Media Bargaining Code of 2021. The social media minimum age law of late 2024, which came into effect at the end of 2025. Add the earlier internet filter debate and the later privacy reforms and you have a country that regulates technology often, fast, and in a pattern that’s become predictable enough to describe. I’ve covered all of them from the press gallery side and this is the shape of it.
The trigger is a story, not a study #
Each of the four began with a public event rather than a policy process. The data retention scheme followed a run of security concerns and the argument that carriers were deleting records investigators needed. The encryption laws, formally the Telecommunications and Other Legislation Amendment (Assistance and Access) Act 2018, followed a period of terrorism-related prosecutions in which encrypted messaging featured, and were passed in the last sitting week of the year. The bargaining code followed the collapse in newspaper advertising revenue and an ACCC inquiry that named the platforms. The social media age law followed a public campaign, a state premier’s announcement and a federal election year.
None of that makes the laws wrong. It does mean they arrive with their conclusions already drafted, and the consultation, which in each case was real, tends to shape the details rather than the direction. Submissions to the encryption bill’s inquiry ran into the hundreds and the bill passed within weeks of the inquiry reporting, with a promise of amendments later. Some of those amendments took years. Some never came.

The mechanism is a bespoke instrument #
Where other countries have tended to extend existing law, Australia builds a new instrument each time. Data retention got its own scheme with its own dataset and its own oversight. The encryption law invented three new kinds of notice, technical assistance requests, technical assistance notices and technical capability notices, with different thresholds for each. The bargaining code created a designation power and an arbitration mechanism that exist nowhere else. The age law created a duty on platforms, enforced by the eSafety Commissioner, with a definition of “age-restricted social media platform” that the regulator has had to interpret through guidance.
The advantage is precision; each instrument fits its target. The cost is complexity, and a regulatory landscape that the businesses subject to it, particularly the smaller ones, find hard to read. The eSafety Commissioner’s office has been candid in its own reporting that the industry codes and standards under the Online Safety Act 2021, which sit alongside the age law, took years to negotiate and are still being revised.
The enforcement is slower than the passage #
This is the part of the pattern that gets the least attention. The bargaining code has never formally designated a platform; its effect came through the deals the platforms struck to avoid designation, and when one of them walked away from those deals in 2024 the government’s response was a new levy proposal rather than the designation the code was built around. The encryption notices have been used far less often than either side of the 2018 debate predicted, and the oversight reports say so in dry annual tables. The data retention scheme’s dataset was accessed heavily, and the review that followed found the access rules had drifted well beyond the scheme’s original purpose, which is a different kind of enforcement problem.
The age law is the live test. Its enforcement began in December 2025 and the early months have been about platforms’ age-assurance approaches, the regulator’s guidance and the inevitable question of what happens to a service that says it complies and plainly doesn’t. I’d expect that pattern to hold: a strong law, cautiously used, with the first real enforcement action arriving a year or more after passage.
The industry’s role in the pattern #
It would be convenient to tell this as a story of governments acting and companies suffering, and the industry tells it that way. The record is less tidy. The platforms’ response to the bargaining code included, at one point, a decision by one of them to switch off news for every Australian user for several days, which converted a technical argument about link value into a political one about corporate power and guaranteed the code’s passage. The carriers’ response to data retention was to argue about cost recovery rather than scope, and the scope is where the later problems were. And the technology industry’s submissions on the encryption bill were, by the account of several of the people who wrote them, drafted in the assumption that the bill would be delayed, which it wasn’t.
The lesson the more experienced government-affairs people took from those years is that the window for shaping an Australian technology law is short, early and mostly closed by the time the exposure draft appears. The ones who engage with the department during the long, dull review phase get changes. The ones who wait for the bill get a hearing and a press release.
What the pattern costs #
Three things, mostly. Compliance cost that falls hardest on the businesses least able to interpret a bespoke instrument, which is why the small end of the telecommunications and online sectors has spent a decade paying lawyers to read laws aimed at companies a thousand times their size. Regulatory drift, where a scheme built for one purpose is used for another because the drafting left room; the data retention review documented this in detail and the recommendations to tighten it took years to act on. And a credibility cost with the rest of the world, where Australian laws are cited, with some justification, as examples of what a well-meaning parliament does in a hurry. The encryption law in particular is still raised in overseas standards bodies as the reason certain products won’t be offered here in certain forms.
The exceptions that show the rule #
Two areas have gone differently, and they’re instructive. The Privacy Act reforms of 2022 and 2024, which raised penalties sharply and then introduced a statutory tort for serious invasions of privacy alongside a range of other changes, came out of a multi-year review by the Attorney-General’s Department rather than a single event, and they arrived in stages rather than in one late-year sitting. The privacy regulator’s enforcement has followed with a consistency the other regimes lack. And the Consumer Data Right, whatever its adoption problems, was designed through a Treasury process over several years and built as a framework rather than a reaction.
The difference, in both cases, is that the department owned the reform and the minister announced it, rather than the reverse. That’s the whole pattern in one sentence.
A fourth cost is harder to measure and I’ll state it as an opinion: each of these laws was drafted with a particular large company in mind, and each has been applied, in practice, mostly to that company. The bargaining code was about two platforms. The age law is about a handful. A framework that names its target in everything but the text is not really a framework; it’s a negotiation with the statute book as leverage, and the next company to arrive in the market inherits rules that were written for someone else’s behaviour.
What it means for the next one #
There will be a next one. Artificial intelligence is the obvious candidate: the proposals paper on mandatory guardrails for high-risk AI has been through consultation, the voluntary standard exists, and the government has said it prefers a targeted approach over an omnibus law. That’s the departmental-process route and I’d like it to hold. The risk is the story: an AI incident of the kind that generates a week of headlines, followed by a bill drafted in a fortnight, passed in a sitting week and amended for the rest of the decade.
My opinion, held for some years now and worth exactly what you paid for it, is that Australia’s technology laws would be better if they were slower and duller. The bespoke instruments passed in a hurry have mostly worked less well than the reforms that came through the long, tedious departmental route, and the industry’s complaints about “regulatory uncertainty” are, more often than not, complaints about laws whose enforcement nobody can predict because their drafting was never finished. A parliament that took eighteen months over an AI law would be criticised for dithering. It would also, on the evidence of the last nine years, produce a better law.
For the record, I filed the encryption bill story in December 2018 from a Senate corridor at 11pm, and I’ve re-read the piece since. It got the mechanism right and the enforcement wrong; I thought the notices would be common. They weren’t. That’s the pattern too: the people who write about these laws, me included, tend to overestimate how much they’ll be used and underestimate how long they’ll be argued about.
— Marcus Tully, Editor, Technology Spectator