Can You Patent Software in Australia? A Poker-Machine Case Just Changed the Answer

For about a decade, if you were building software in Australia and asked a lawyer “can I patent this?”, the honest answer was a wince and a “…probably not.”

So most founders stopped asking. They filed in the US instead, or locked their best ideas in a drawer marked trade secret, or just shipped and hoped. A whole generation of Australian software companies grew up believing the actual clever bit — the thing that made them valuable — couldn’t be protected at home.

That belief is now out of date. Between September 2025 and February 2026, a fight over poker machines that had been grinding through the courts for years finally ended — and quietly rewrote the rules on whether you can patent software in Australia.

Here’s what changed, why it was broken in the first place, and what it means for what you’re building.

The decade Australia said “no” to software patents

Start with the absurdity that kicked it all off.

Picture a poker machine — a physical one, with spinning reels, gears and motors. The patent office says: lovely, here’s your patent. Now build the exact same game in software, on a screen, doing exactly the same thing for the player. For years, the answer to that one was: sorry, that’s just an “abstract idea” running on a “generic computer.” No patent for you.

Same invention. Same result for the user. One’s protected because it has gears; the other isn’t because it has code. If that sounds like nonsense, it’s because it was — and it’s the trap a lot of genuinely clever Australian software fell straight into.

A quick word on the legal plumbing, because it explains everything that follows. To be patentable in Australia, an invention has to be a “manner of manufacture” — the threshold for patentable subject matter under s 18(1)(a) (and, for the now-abolished innovation patents at the centre of this saga, s 18(1A)(a)) of the Patents Act 1990 (Cth). That quaint phrase is inherited straight from s 6 of the Statute of Monopolies 1623 — yes, 1623 — and it’s the gate every invention, software or otherwise, has to pass through. Notably, Australia has no express statutory exclusion for software (unlike Europe, which formally excludes “programs for computers as such” under Art 52 of the European Patent Convention). The whole question turned on judge-made interpretation.

The machinery behind the “no” was a test the courts had built on top of that gate — informally, the “two-step test.” An examiner would ask: is this a computer-implemented invention? If yes, does it show an “advance in computer technology”? In practice that meant the examiner mentally stripped out the “generic” computer, looked at whatever was left, and if that looked like a business method or a mathematical scheme, rejected it. Software that did something genuinely new got dissected until the new part could be waved away as a mere idea.

Case after case fell to this: schemes for protecting assets (Grant v Commissioner of Patents [2006] FCAFC 120), methods for building financial indices (Research Affiliates LLC v Commissioner of Patents [2014] FCAFC 150), systems for assessing people’s competencies (Commissioner of Patents v RPL Central Pty Ltd [2015] FCAFC 177), tools for displaying business intelligence (Encompass Corporation Pty Ltd v InfoTrack Pty Ltd [2019] FCAFC 161), and a digital advertising engine (Commissioner of Patents v Rokt Pte Ltd [2020] FCAFC 86) — knocked down one after another as “abstract ideas on a computer.” Australia earned a reputation, fairly, as a hard place to get a software patent.

The poker machine that wouldn’t quit: the Aristocrat case

Enter Aristocrat, one of the world’s biggest makers of electronic gaming machines (EGMs — “pokies,” if you’re being honest about it).

Back in 2017–18, the patent office rejected Aristocrat’s innovation patents on its digital “feature games” using exactly that reasoning: clever, sure, but it’s just software on a standard machine, so not a manner of manufacture. Aristocrat disagreed — and, unusually, had the resources to argue the point all the way. So began one of the longest-running patent sagas in Australian legal history.

It went up. It came down. It went up again. In 2022 it reached the High Court, which split three-all — an almost unheard-of deadlock (Aristocrat Technologies Australia Pty Ltd v Commissioner of Patents (2022) 274 CLR 115; [2022] HCA 29). Because the court was tied, the rejection technically survived: under s 23(2)(a) of the Judiciary Act 1903 (Cth), when the High Court is equally divided, the decision appealed from is simply affirmed — no binding precedent is created. But the telling part wasn’t the result; it was that all six judges criticised the old two-step test. Nobody was defending it anymore.

That set up the finale. On 16 September 2025, the Full Federal Court (Beach, Rofe and Jackman JJ) took the whole framework apart in Aristocrat Technologies Australia Pty Ltd v Commissioner of Patents [2025] FCAFC 131. It threw out the two-step test as too narrow and detached from first principle, and went back to the question the law had used since 1959: does the invention produce an “artificially created state of affairs” of economic utility — something real, with a useful result? That test comes from the foundational High Court decision National Research Development Corporation v Commissioner of Patents (1959) 102 CLR 252 — “NRDC,” the case every Australian patent lawyer can recite in their sleep.

The poker machines passed. And on 5 February 2026, when the Commissioner asked the High Court to step in one final time, the High Court refused special leave, finding insufficient reason to doubt the correctness of the Full Court’s decision. With no further avenue of appeal, the September 2025 reasoning is now the leading authority on software patent eligibility in Australia.

A decade of “probably not” was over.

The new software patent test, in plain English

Strip away the legal language and the question is now refreshingly human:

Is your software just an abstract idea sitting on a computer? Or does it make the computer — or the world — actually do something?

The first is still a no. The second is now a yes. The courts call the dividing line an “artificial state of affairs,” and IP Australia has since baked it into its examiner’s manual (the March 2026 rewrite of s 5.6.8.6). In practice, software lands on the patentable side when it does one of three things — and there are now recent, real cases for each:

• It makes the computer itself work better. Faster, leaner, more capable — a real improvement to how the machine does its job, not just a new task you’ve handed it. The textbook example is International Business Machines Corp v Commissioner of Patents (1991) 33 FCR 218, where switching to integer arithmetic to render a curve more efficiently — better computer graphics — was held patentable. The software changed how the computer natively computed, not just what it computed.
• It makes something happen in the real world. Your code controls a physical thing. In UbiPark Pty Ltd v TMA Capital Australia Pty Ltd (No 2) [2023] FCA 885, software that used smartphone signal strength to decide when to raise and lower the boom gates at a car park produced a concrete, tangible result off the screen — patentable.
• It solves a genuinely technical problem. Not “how do we sell more widgets,” but a real engineering problem. In Hytera Communications Corporation Ltd v Motorola Solutions Inc [2024] FCAFC 168, a method that improved scanning efficiency across timeslots in a TDMA digital-radio network — making the hardware materially better at sending messages — qualified.

Notice what all three share: the software does work. It produces a result you can point to. And the old trick of dissecting your invention until the clever part disappeared is gone — following NRDC and the High Court’s reasoning, the law now characterises your invention as a whole, in substance. Examiners are no longer allowed to give “undue or preponderant weight” to the inventive bits while dismissing the standard processor as a mere generic environment; off-the-shelf hardware still counts if it’s integral to how the thing works.

Two quick scenarios (in your language)

Abstract tests are slippery, so here’s how the line actually falls in practice.

The difference between A and B isn’t how impressive the product is. It’s whether the cleverness lives in the engineering or in the idea.

The catch: it’s not automatic

Before you rush off to patent your weekend side-project: this is not a free-for-all.

“An abstract idea on a generic computer” is still unpatentable — and plenty of software is exactly that. IP Australia’s own worked example of a no is a location-based payment method on a standard phone and processor, each just doing its normal job. Real cases draw the same line: in Dei Gratia v Commissioner of Patents [2024] FCA 1145, a clever last-mile logistics system was knocked back because the physical outcome (goods getting delivered) was too indirect — the substance was still a scheme. And in F45 Training Pty Ltd v Body Fit Training Company Pty Ltd (No 2) [2022], bolting physical exercise stations onto a computer-implemented scheme wasn’t enough to rescue it. The dividing line is real; it has simply moved to a far more sensible place.

Here’s the bit that catches people out: how your invention is described can decide its fate. Post-Aristocrat, the battleground has shifted from is software patentable? (it is) to how an examiner characterises the substance of your invention. Examiners still tend to default to “this is just a business scheme on a generic computer” — and because the Full Court was careful to say the outcomes in cases like RPL Central and Rokt were correct (even if the old reasoning was flawed), those analogies remain available to them. In practice, it’s common to sail through the novelty and inventive-step objections in the first office action and then spend several rounds arguing subject-matter eligibility.

What wins those arguments is technical specificity. Specifications that spell out how the processor, memory or network is actually used — the causal link between the algorithm and the physical or technical result — read as engineering. Specifications that stay vague read as “a business idea with a computer bolted on.” That’s craft, and it’s where getting the right help early genuinely changes the outcome. (One trade-off to weigh: the more tightly you tie a claim to a specific implementation, the easier it can be for a competitor to design around — a strategy call, not a form-filling one. You’ll also need that detail to satisfy the separate support and sufficiency requirements in ss 40(2)–(3) of the Patents Act.)

Filing globally: Australia is now a strong launchpad

You don’t need the full international tour, but the headline matters: Australia has gone from one of the harder places to patent software to one of the more receptive ones — and the rest of the world has been drifting the same way. In the US, a 2025 USPTO reset on patent eligibility raised the bar examiners must clear before rejecting a claim as an “abstract idea” (good news for AI and machine-learning inventions). Europe has long let almost anything with a technical element clear the eligibility gate, fighting it out at the inventive-step stage instead. Canada looks for a “physicality requirement.” For software at the margins, Australia is now a genuinely competitive jurisdiction to file in.

That changes strategy. A granted or allowed Australian claim — one that has already demonstrated an “artificial state of affairs” — is a useful lever offshore. It lines up neatly with the US’s renewed focus on technological improvement and with Canada’s physicality test, and you can use it to request accelerated examination in those countries through the Patent Prosecution Highway. And from 1 March 2026, a new two-year EPO/IP Australia PCT pilot lets Australian applicants nominate the European Patent Office as their International Searching Authority — giving you an early, rigorous read on patentability, letting you skip the supplementary European search, and (if you also request preliminary examination with the EPO) cutting European examination fees by 75%.

The practical point: the order and timing of your filings across countries can save real money and preserve options. That’s a conversation worth having before you start, not after.

So what should software founders actually do?

If software is the heart of what your company is worth, three moves are worth making now.

1. Reconsider patents. If you were told years ago that your core tech “can’t be patented in Australia,” that advice was almost certainly given under the two-step test that no longer exists. It’s worth a fresh look — especially for anything that improves performance, controls hardware, or solves a real technical problem. Mind the clock, though: public disclosure (a demo, a launch, a paper, a pitch that isn’t under NDA) can destroy novelty under s 18(1)(b). Assess this before you ship.
2. Remember patents aren’t your only tool. Your source code is protected by copyright as a literary work the moment it’s written — no registration needed, under the Copyright Act 1968 (Cth) — and your real secrets can be protected as confidential information. The strongest position is usually all three at once: patents where they fit, copyright underneath, and secrecy around the crown jewels.
3. Get ownership sorted. A patent (or any IP) only helps if you actually own the invention — which gets messy fast once contractors, departed founders, an offshore team, and open-source components with copyleft licences are in the mix. Contractor agreements don’t assign IP automatically the way employment often does. A quick IP audit catches all of that before an investor or acquirer does it for you in due diligence.

The real shift here isn’t about poker machines. It’s that Australia has stopped treating software as second-class and started treating it like any other kind of engineering. If you build software for a living, your own country just became a much better place to protect what you build. (Want the official primer? IP Australia’s guide to what computer-related inventions can be patented is a useful starting point.)

This article is general information only and is not legal advice. The law described is current as at June 2026 and may change. For advice on your specific circumstances, speak with a qualified Australian patent lawyer or patent attorney.