Australia Abolished the Innovation Patent in 2021. New Ones Are Still Being Filed — and the Law Fixes the Last Day Any Can Exist: 25 August 2029.

In a report delivered in September 2016 and released that December, the Productivity Commission made a recommendation that needed no subclauses: “The Australian Government should abolish the innovation patent system.” The report’s overview put its verdict under a plain heading — “The ‘second-tier’ patent experiment has failed” — and its objection was strategic use: patent attorneys, the Commission wrote, “openly advertise ways in which users can game the system, including to improve their bargaining position in patent disputes and to frustrate entry by competitors.” The Government agreed in August 2017, Parliament legislated in February 2020, and on 25 August 2021 the innovation patent — an eight-year, five-claim right granted without substantive examination — closed to new filings.

What happened next is not news to patent attorneys: Mark Summerfield’s Patentology blog described the surviving divisional and conversion routes in May 2022, when post-close volumes were running only about 10% below the previous year’s, and the 2026 Chambers practice guide describes the same routes today. What the register now allows is a count of what those routes produced, and an end date. Since the closing day, 508 further innovation patent applications have been filed. Sixteen arrived in the first half of 2026; the newest carries a filing date of 12 June 2026, four years and nine months after abolition. Nothing irregular is happening. Every one of those filings is the statute working exactly as Parliament drafted it — and the same drafting choice that lets them in also guarantees that no Australian innovation patent can exist beyond 25 August 2029.

One sentence, and the door it leaves open#

The Act that killed the system — the Intellectual Property Laws Amendment (Productivity Commission Response Part 2 and Other Measures) Act 2020 — is spare on the subject. Schedule 1 Part 2, the whole abolition, runs to eight items, and unlike Schedules 2 and 3 of the same Act, which each carry an express Part headed “Application and transitional provisions”, the innovation-patent Part carries no saving provision and no end date. The move that decides what can still be filed is a single sentence bolted onto the formalities check as s 52(3) of the Patents Act: “It is a requirement of the formalities check that the date of the patent (if granted) would be a date before the day this subsection commences.” Two companion items add a second test at examination and certification — every claim must have a priority date before the same day (ss 101B(2)(ha), 101E(1)(a)(viiia)) — so the route stays open only to families whose claims already predate the close.

The load-bearing phrase is “the date of the patent”. Parliament chose to test that date, not the date of filing — and the two are not the same thing. Under reg 6.3 of the Patents Regulations 1991, a divisional application — a spin-off carved from an earlier “parent” application — takes its parent’s date as its own date of patent. An application physically filed in 2026 can therefore wear a date from 2021 or earlier, satisfy the test, and proceed. Divisional filing is ordinary patent practice, not a manoeuvre — every patent system has an equivalent, and the divisional provisions themselves, ss 79B and 79C, were left untouched by the abolition.

IP Australia’s examiners’ manual describes the result plainly: “Innovation applications must have a date of patent before 26 August 2021”, which can be achieved by filing a divisional off a parent with a pre-cutoff date, dividing a granted innovation patent, or converting an old standard application. The register shows the consequence: of the 508 post-close filings, 504 carry a divisional link to one of 409 distinct parents. The taps stay open because the regulations give a three-month divisional window each time an old standard application reaches acceptance, and a one-month window each time an innovation patent is certified — so as pre-2021 applications continue to work through examination, they continue to throw off innovation divisionals, years after “abolition”.

The channel is not any one actor’s instrument. The 508 filings stand in the names of 390 distinct applicants; the most frequent appears on 19 of them, 3.7% of the cohort, and the ten most frequent together account for 17.7%. About 63% of the applications have an Australian applicant. And the flow is drying up on its own — 196 filings in 2022, 107 in 2023, 50 in 2024, 24 in 2025 — as the stock of pre-cutoff parents that can still spawn divisionals works its way through the system.

What a patent from a dead system buys#

An innovation patent is granted after a formalities check only. IP Australia’s manual is candid that grant “does not bestow any enforceable rights and privileges upon the patentee”. The right becomes a weapon only on certification — an optional, separately-requested substantive examination — because s 120(1A) forbids infringement proceedings on an uncertified innovation patent, and s 129A makes even threatening proceedings before certification an unjustified threat.

As at the 27 July 2026 extract, 190 of the 508 post-close filings — 37.4% — had been certified. That figure sits in the normal range for what these filings are. Divisional innovation filings have always certified at rates far above the system-wide average: across the comparison windows in the published data, divisionals certify at 25–45% — 25.2% for divisionals inside the 2021 rush, 29.5% for those filed from 2019 up to the close, 39.0% for 2013–2018, and 45.4% across all pre-close divisionals. The post-close cohort’s 39.7% (for filings to the end of 2024, which have had time to mature) is unremarkable against that baseline. The figure that looks dramatic beside it — the 7,595 applications of the January–August 2021 deadline rush certified at just 3.1% — mostly measures a different thing: that rush was 96% original filings, and original filings were rarely certified. Like compared with like, the afterlife channel is simply divisionals behaving the way divisionals always have. These rates are floors as at the extract: 178 post-close applications alone were still alive and uncertified, so the rates can only rise.

What has changed is what certification buys. The back-dating that lets a post-close divisional through the s 52(3) gate also starts its clock in the past: s 68 gives an innovation patent “8 years from the date of the patent”, and that date belongs to the parent. A divisional filed today inherits a term that has already been running for years.

The median post-close divisional filed in the last months of 2021 had 6.5 years of term left at filing. By 2024 the median was 4.1 years; for 2025 and 2026 filings alike it is 3.4. Certification consumes more of it — the median right had 4.4 years remaining by the time it became enforceable — and the 137 post-close rights enforceable at the extract held a median of just 2.0 years. The route Parliament’s drafting left open leads into a room that gets smaller every year.

Nine patents, one drug, two court cases#

Whether any of this matters outside a register is a fair question, and one family answers it. The following account is drawn from the Federal Court’s own file records and sealed orders, and from IP Australia’s register; the patents’ filing, certification and end dates reproduce from the raw register extract via this piece’s analysis script.

Janssen Biotech, Inc., a Johnson & Johnson company, is the patentee behind Stelara, the ustekinumab immunology drug listed on the PBS; it acted in the Australian litigation alongside its local affiliate Janssen-Cilag Pty Ltd (together, Janssen, except where an entity is named). Between May 2023 and November 2024, Janssen Biotech filed nine innovation patent divisionals off a single parent application — AU2023201217, itself a divisional of the 2019 PCT-derived application AU2019346134, which was refused in January 2026; the parent itself remains under examination. All nine are titled for a method of treating ulcerative colitis with an anti-IL12/IL23 antibody, and each inherited a date of patent that had been running since September 2019 — the register projects every one of them to the same expiry day, 24 September 2027. For the three the Federal Court later revoked, the register also records a priority date of 24 September 2018, before the cutoff, which is what kept the divisional route open to the family. Eight of the nine were certified, most within about six months of filing. Each certification created a right that could be sued on.

The litigation ran in both directions. In January 2024, Samsung Bioepis AU Pty Ltd — whose ustekinumab biosimilar Epyztek was moving toward the Australian market — commenced Federal Court proceeding NSD65/2024 against Janssen Biotech concerning the first two divisionals. On 28 August 2024, Justice Downes ordered Janssen to offer to surrender both patents under s 137 of the Patents Act and to pay Samsung’s costs, with the proceeding to be discontinued once the Commissioner accepted the surrender; a remaining question about the basis of those costs was determined separately, and the file was finalised on 19 September 2024 with a judgment, [2024] FCA 1099, that could not be retrieved for this piece — so its reasons are not reported here. The register records the offer to surrender that September and the revocation that November.

By then Janssen had filed and certified three more divisionals — certified on 7 August 2024 — and nine days later, on 16 August 2024, Janssen Biotech and Janssen-Cilag commenced their own proceeding, NSD1134/2024, suing Samsung Bioepis for infringement. The court’s orders define the products in issue as Samsung’s ustekinumab formulations; the court record shows no interlocutory injunction application and no order granting one — the interlocutory applications the court determined concerned discovery. The final hearing concluded on 6 June 2025. On 9 June 2025 the court dismissed the applicants’ amended originating application, revoked every claim of all three patents, and ordered the applicants to pay Samsung’s costs. The order is expressed as an order of the court, not a consent order; no reasons for it have been located, so this piece does not report why the patents were revoked.

The rest of the family was already gone or soon went. The three further divisionals Janssen Biotech had filed in November 2024 and certified in February 2025 ended by the patentee’s own hand: the register records an offer to surrender published in March 2025 and revocation on surrender with effect from 5 May 2025, with no court order concerning them located. The ninth family member had already ceased, uncertified, in November 2024. The tally: nine patents from an abolished system’s residual channel; eight made enforceable; two surrendered under court order, three revoked by order of the Federal Court, three surrendered voluntarily, one ceased; costs against Janssen in both proceedings. The Productivity Commission’s 2016 findings quoted above described the system in general, years before any of this; nothing in the register or the court file records why any patent in this family was filed, and no such claim is made here.

Two facts sit side by side at the end of this story, and no causal link between them is asserted. The patents asserted against Samsung’s Epyztek are gone. And as at 1 August 2026, Epyztek is still not listed on the PBS — the PBS status page records a positive PBAC recommendation from March 2025, with the sponsor’s notice of intent lodged only in July 2026 and listing “not yet occurred” — while a rival ustekinumab biosimilar, Celltrion’s Steqeyma, has been PBS-listed since 1 August 2025. Nothing in the court file or the PBS record explains the gap, and this piece does not.

The end is already scheduled#

The same one-sentence mechanism that created the afterlife also ends it. No innovation patent can have a date of patent on or after 26 August 2021 — that is the s 52(3) test itself — and the term is eight years from that date, with no extension available: the Patents Act’s extension-of-term provisions apply to standard patents only. The arithmetic is a ceiling, not a forecast: the last possible day any Australian innovation patent can be in force is 25 August 2029.

The register agrees. Of the 545 innovation patents enforceable at the extract, every single one has a projected expiry — its recorded date of patent plus eight years — on or before 25 August 2029, and the latest lands exactly on that day.

Fifty-five of the rights counted at the extract run out during the rest of 2026, 177 in 2027, 130 in 2028 and 183 in 2029. The post-close cohort is present in every instalment, including 49 rights that survive into the final year. A quarter of everything still enforceable — 137 of 545 — came through the residual channel. That share partly reflects the wind-down itself, since older cohorts have already expired, and the comparison windows are unequal — the post-close cohort spans almost five years of filings, the 2021 rush eight months — so no clean ranking exists; on single filing years, the post-close cohort’s 137 sits below 2021’s 203 and above 2020’s 118. On the day the last right lapses, the innovation patent will have taken exactly eight years to finish dying from the moment it became impossible to file one.

What could explain this instead#

A selection effect, not a behaviour change. Post-close filings certify at nearly twelve times the rate of the 2021 deadline rush, but that comparison mostly measures composition: the rush was 96% original filings (which certified at 2.2%), while the afterlife channel is almost entirely divisionals — and divisionals in every comparison window certify at 25–45%. A transitional divisional is a deliberate act by a holder that already has a live parent application, so the cohort selects for owners with a continuing interest in the family; divisional practice itself is ordinary and universal. This piece therefore treats the certification pattern as a description of who uses the channel, not as evidence that the closure changed anyone’s conduct.

Survivorship arithmetic. “A quarter of all enforceable innovation patents came through the post-abolition channel” is partly a mechanical consequence of an eight-year term: every cohort older than eight years holds none of today’s stock by construction, so recent cohorts dominate whatever remains. The single-year comparison is given above, and the post-close share is a description of the stock, not a measure of the channel’s scale.

The residue is tiny. The 508 post-close filings are 1.21% of the 41,887 innovation patent applications ever filed in Australia. Against the system’s own history, the afterlife is a rounding error; what makes it worth counting is that it is still producing enforceable rights years after the system closed, not that it is large.

Court events are too few, and too clustered, to be a trend. The register records ten Federal Court–ordered revocations of innovation patents in its entire history, and multiple patents can fall to a single order — the three 2025 revocations are one order, in one proceeding, on one family. On the dates the orders took effect, 2025 (three) ties 2023 (three) rather than leading. Counts this small support no claim about litigation rising or falling, and none is made. A related caution: the register’s “revoked” status mostly does not mean a court acted — of 229 innovation patents recorded as revoked, 173 carry neither a Federal Court order event nor a Commissioner’s hearing-decision event, having ended by other routes, chiefly surrender.

Recording and vintage limits. These figures describe what IP Australia’s register records, as at the 27 July 2026 extract — not necessarily every court outcome, since a revocation reaches this dataset only when the register records it, and the lag between a court order and its Official Journal publication has ranged from 24 days to almost five years. Certification rates are floors, counts of live rights will have moved since the extract, and the dataset itself was refreshed on 3 August 2026, after this analysis was frozen; re-running against the live file will produce slightly different counts.

The innovation patent is ending the way it began: as a drafting exercise. A system the Productivity Commission judged a failed experiment was closed by one sentence, survived through the interaction of that sentence with a 1991 back-dating rule, supplied the patents that a pharmaceutical incumbent and a biosimilar maker fought over in the Federal Court, and will finish — completely, arithmetically, and without any further decision by anyone — on 25 August 2029.