The sunset that won't set

On 4 September 2026, three of the most intrusive powers ever handed to Australian police were scheduled to die. Data disruption warrants let officers alter data on a device. Account takeover warrants let them assume control of an online identity. Network activity warrants let them monitor whole networks. Parliament attached a five year sunset when it created them in 2021. That date came and went, because with nine days to spare the extension received Royal Assent: the Act was assented on 26 August 2026, and the schedule moving the sunset to 4 September 2029 commenced the next day, eight days before the old deadline.

The sunset was not decoration. It was the safeguard: a forced reckoning, on a fixed date, at which the powers would have to justify themselves from scratch or lapse. The reckoning happened. The Independent National Security Legislation Monitor, the government's own reviewer, tabled a report on 1 September 2025 that made 21 recommendations and found that 'the main safeguard, the current system for issuing warrants, is not fit for purpose'. The Monitor said the powers 'should be allowed to sunset' if recommendations 6-8 were not implemented. They have not been implemented.

Instead of a lapse, there is a schedule inside a schedule: Part 3 of Schedule 1 of the Crimes and Other Legislation Amendment (Omnibus No. 1) Bill 2026, which moved the sunset three years down the road (now Act No. 70 of 2026). Here is the mechanism to watch, because it is the whole story. A sunset clause that moves whenever it approaches is not a safeguard. It is safeguard theatre, and the omnibus bill is its stage.

01A sunset only disciplines power if the sun can go down

The powers come from the Surveillance Legislation Amendment (Identify and Disrupt) Act 2021, the statute that gave Australian police what plain language calls hacking powers: altering data, taking over accounts, monitoring networks. Powers of that reach were made politically possible by an expiry date. The promise embedded in the 2021 Act was that the powers were provisional, that they would face a genuine test before becoming permanent furniture.

The operational record since is thin. The powers were used 26 times over 2021-2024. Attributed arrests: nil. For a time the Australian Federal Police claimed one arrest flowing from an account takeover warrant. Appendix D of its Annual Report 2024-25 corrected that figure to 'Nil'.

The Monitor read the low numbers two ways. The ACIC and the AFP describe their use of the powers as judicious, and the report accepts that restraint is part of it: 'This shows that restraint is being exercised when using these significant powers.' But it adds, in the same paragraph, that 'practical limitations arise because of the complex technology and specialised skills required to execute some of the warrants', and later lists those limits as 'cost, resourcing and technological capability'. Low use, on the watchdog's reading, is partly choice and partly capacity.

None of that settles whether the powers should exist. It settles something narrower and more important: whether the case for them was ever tested against results. Testing that case, on a fixed date, was the sunset's entire job.

02The reckoning arrived on time, and it was filed

The test the sunset was designed to force did in fact occur. The Independent National Security Legislation Monitor examined the regime and tabled its report on 1 September 2025, a year ahead of expiry. It made 21 recommendations. Its verdict on the machinery that issues these warrants was blunt.

The Monitor, Jake Blight, has since put the same finding in the first person in the UNSW Law Journal: 'Critically, I found that the current system for issuing warrants is not fit for purpose.' That is not outside scholarship confirming the report. It is the report's own author restating it, in an article written while the omnibus bill was still before Parliament.

The report is specific about what that means for one of the three powers. Network activity warrants, it found, 'allow surveillance of whole networks of people to gather intelligence "relevant" to crime, but without a requirement to have a reasonable suspicion that all those under surveillance are engaged in criminal activity.'

The Monitor did not recommend scrapping the powers outright. He recommended that the AFP retain all three and the ACIC two, not data disruption, on conditions. His fix for the warrant system has three parts: 'retired judges as the issuing authorities; public interest monitors; and access to independent technical advice.' And he named the consequence of ignoring the conditions: the powers 'should be allowed to sunset' if recommendations 6-8 were not implemented. As of today they have not been. On the watchdog's own stated condition, the sun should have set.

The government's response, dated November 2025 and tabled in the House on 10 February 2026 and the Senate on 2 March 2026, agreed recommendations 1-3 and extends the powers 'in the first instance' pending the long promised electronic surveillance reform. By the Monitor's own count, the government agreed 4 of his 21 recommendations, agreed 2 in part and noted the other 15. The Monitor replied in a submission dated 7 April 2026, saying it was 'disappointing' that electronic surveillance reform was taking so long, and that the office was 'concerned'. Those are the watchdog's verbatim words about the fate of its own review.

Correction, 25 September 2026. This article previously said the Monitor called the government's 'approach' disappointing. The Monitor's word 'disappointing' was about how long electronic surveillance reform is taking.

The reckoning arrived on schedule. The deadline is what moved.

Correction, 25 September 2026. This article previously sourced the 'not fit for purpose' finding to 'Dr Blight' in the UNSW Law Journal, as if that article documented the Monitor's finding from outside. The finding is in the Monitor's own report, which we now quote directly. Jake Blight is the Monitor; the INSLM's own website calls him Mr Jake Blight, and his journal article is his first-person account of his own review.

03An omnibus is where a hard argument goes to be brief

Now look at the vehicle. The extension does not travel as a bill about surveillance. It travels as Schedule 1 Part 3 of the Crimes and Other Legislation Amendment (Omnibus No. 1) Bill 2026, introduced on 11 March 2026 and passed by the House on 25 March 2026. The Senate agreed the second reading on 12 August 2026 and worked through Committee of the Whole across 12-13 August. When we first published this piece on 14 August 2026 the bill was still before the Senate, and we would not report its passage as fact. It is fact now. The Senate agreed the third reading on 20 August 2026, the House accepted the Senate's amendments the same day, and the bill received Royal Assent on 26 August 2026 as Act No. 70 of 2026. Schedule 1, which holds the extension, commenced on 27 August 2026, eight days before the old sunset. Two Opposition amendments were agreed on 17 August, one of them as amended by the Government. An amendment that would have brought the sunset forward to 1 July 2027 was not agreed. The powers now run to 4 September 2029, and the reckoning this sunset was built to force has been deferred again.

What the schedule does is precise. It extends the sunset for data disruption warrants, network activity warrants and account takeover warrants, and the related emergency authorisations, from 4 September 2026 to 4 September 2029. There is one genuine narrowing inside it: the ACIC loses its data disruption warrant power, a change the Monitor himself supported. Note that, and note where it happened, inside a package rather than inside a debate about the powers themselves.

The government's case for the extension is in its own Explanatory Memorandum. Its Statement of Compatibility with Human Rights says the powers carry 'a range of safeguards, stringent thresholds, proportionality tests, and clear specifications regarding the actions permitted under each warrant', and that the schedule, 'while extending the sunsetting date for these powers, still preserves all existing safeguards and thresholds in the current framework.' The first safeguard on its list is issuance by 'a Judge or nominated Administrative Review Tribunal member' for data disruption and network activity warrants (a magistrate issues account takeover warrants). That is the machinery the Monitor found not fit for purpose. Every warrant examined in his review had been issued by Administrative Review Tribunal members, an arrangement he called 'inappropriate and unsustainable'. Preserving all existing safeguards, in other words, preserves the one the watchdog said does not work.

That placement is the mechanism. A standalone bill titled to extend police hacking powers against the advice of the government's own reviewer would draw a standalone fight. An omnibus crimes bill draws a queue. The extension becomes one part of one schedule among many, priced into a package vote, and a surveillance extension that never has to survive a debate of its own is an extension that will almost always survive.

04The standing excuse is a reform that never arrives

The stated justification for extending rather than reforming is that a bigger fix is coming. The electronic surveillance reform project flows from the Richardson Comprehensive Review. The project commenced in 2020, with a bill originally due in 2023 on the Richardson review's own timeline. It has not been delivered.

The government states that rationale in terms. Its Explanatory Memorandum says 'the extension of the sunsetting date is designed to ensure that the power remains in place until after the electronic surveillance reform is completed.' It is a coherent argument on its face. Its weakness is the one this section describes: it holds only for as long as the reform is actually coming.

This is how 'in the first instance' becomes a permanent tense. The powers are extended because the mega reform is pending. The mega reform stays pending. Each time a sunset approaches, the pending reform is the reason not to let it fire, and the failure to deliver reform converts itself into the argument for extension. Delay is not the obstacle to the fix. Delay is the operating fuel.

To keep this honest: the only status we can verify for the reform project is that it has not been delivered. The responsible department's project page could not be checked before publication, and we will not guess at its contents.

05While Parliament argues about the sunset, the ledger keeps filling

The extension debate concerns 26 uses of exotic powers. The routine surveillance machine runs at another scale entirely, and its annual reports land with almost no coverage. Our researchers read the primary PDFs and re-summed the tables.

Correction, 24 September 2026. This article previously gave this figure as 364,868. The TIA Act Annual Report 2024-25 gives 364,868 as the total authorisations for existing telecommunications data made by 21 agencies; of those, 357,864 were made to enforce the criminal law under s178.

Correction, 25 September 2026. The image above previously showed this figure as 364,868, and its kicker described the bill as before the Senate. The graphic has now been corrected to 357,864, and to Act No. 70 of 2026, Schedule 1 in force 27 August 2026.

Victoria Police made 136,155 of those authorisations and NSW Police 126,775, leaving Victoria the largest user in the country for at least the second consecutive year. In 2023-24 the split was 129,561 to 124,079. The lead widened from about 5,500 to about 9,400.

Correction, 25 September 2026. This article previously said the 2024-25 journalist information warrant was issued 'to identify a journalist's source'. The annual report records its purpose as the enforcement of the criminal law; identifying a journalist's source is what the warrant scheme exists to control, not a purpose the report states for this warrant.

The same annual report also covers Part 15 of the Telecommunications Act, the industry assistance scheme created by the 2018 TOLA laws, and it fills in that picture: 58 voluntary Technical Assistance Requests in 2024-25, down two from 60 the year before, 43 of them from NSW Police, five each from the ACIC and Victoria Police, four from the AFP and one from South Australia Police. Compulsory notices: zero Technical Assistance Notices, which the report records as 'a decrease of two from 2023-24' (both of the previous year's notices were the AFP's), and zero Technical Capability Notices. In 2024-25 every recorded piece of assistance flowed through the voluntary channel; the compulsory notice powers went unused.

Correction, 25 September 2026. This article previously attributed the TOLA industry assistance figures to a 'Surveillance Devices Act Annual Report 2024-25'. They are in the combined Telecommunications (Interception and Access) Act 1979 and Part 15 of the Telecommunications Act 1997 Annual Report 2024-25, the same report cited above for the s178 figures. The figures themselves were correct.

06The same play is running next door

The pattern is not confined to one bill. The ASIO Amendment Bill (No. 2) 2025 passed the House on 12 February 2026 and is before the Senate, where the last recorded debate was on 14 May 2026 and the Law Council of Australia's most recent update is dated 28 May 2026. The government has since proposed an amendment retaining the sunset on ASIO's compulsory questioning powers. The usage ledger there is thinner still: 20 questioning warrants sought since 2003, and five since 2006, per the Australian Strategic Policy Institute. Rarely used powers, and a debate over their sunset running in a chamber most people are not watching.

So name the case plainly. THE SURVEILLANCE RORT is not a person and it is not a plot, and no misconduct is implied of any individual in this story. It is three interlocking mechanisms. A sunset clause that supplies the appearance of a deadline without its force. An omnibus vehicle that lets an extension pass without ever surviving a standalone debate. And a perpetually unfinished mega reform that converts delay itself into the argument for extension. Each part is lawful. Together they make expiry practically impossible, which means the safeguard sold to the public in 2021 was never load bearing.

Watch the date, not the rhetoric. The new sunset is 4 September 2029, now law. The reform that justifies the extension was due in 2023. The watchdog's conditions for letting the powers continue remain unmet; its own words were that the pace of reform is 'disappointing', and that it is 'concerned'. When 2029 approaches, the test will be simple: does the sun finally go down, or does the calendar move again. Everything you need to know about this safeguard is in that answer.

Update, 25 September 2026. The 4 September 2026 sunset has passed and the extension commenced on 27 August 2026. This article previously still described the new sunset as applying 'if the Senate agrees', and spoke of the old date as arriving this week.

If it's a rort, we cover it.