APAA e-Newsletter (Issue No. 49, October 2025)

Imminent implications of Gen-AI on Australian patent law & practice

Dr Simone Mitchell, Dr Nigel Lokan, Jaimie Wolbers and Dr Anna Goldys, MinterEllison (Australia)

As we experience the rapid rate of adoption of Generative-AI (“Gen-AI”), it is useful to consider some of the likely impacts of Gen-AI on patent law and practice. In this article, we discuss three areas of patent law and practice that Gen-AI is likely to impact in the near future, and its likely implications.

  1. Inventorship disputes

As confirmed in the Full Court’s decision of Thaler v Commissioner of Patents [2021] FCA 879 (“Thaler”), only a natural person can be an inventor for the purpose of the Patents Act 1990 (Cth) and Patents Regulations 1991 (Cth).

Despite this legal question being well-settled, unless there is significant legislative reform, the continued use of Gen-AI tools in research and development will no doubt raise a number of interesting questions in connection with the concept of inventorship.

Whether a person is characterised as an inventor is a question of law. The approach to ascertaining whether a person is an inventor involves a two-part enquiry, being:

  • identifying the invention; and
  • assessing whether an individual has made a material contribution to the inventive concept.

It has been observed that the issue of inventorship is “one of the muddiest concepts in the muddy metaphysics of patent law“.[1] Gen-AI raises the question as to whether an individual who:

  • develops prompts for a Gen-AI tool;
  • trains the Gen-AI tool;
  • develops the source code for the Gen-AI tool; or
  • owns the Gen-AI tool (or pays for the maintenance and running costs associated with the Gen-AI tool)

could be regarded as having made a material contribution to the inventive concept.

These were matters that were not considered by the Full Federal Court in Thaler, as the case proceeded on the premise that the Gen-AI tool “DABUS” was the inventor of the invention that was the subject of the patent application, and Dr Thaler (the developer and owner of DABUS) was not.

The consequence of misidentifying an inventor varies from jurisdiction to jurisdiction, however they raise entitlement issues which can give rise to grounds for revocation.

In the absence of any legislative change that might clarify these matters, we expect the emergence of inventorship disputes in the context of inventions that have involved the use of Gen-AI in their development.

  1. Are Gen-AI tools patent eligible subject matter?

In Australia, the courts and patent office continue to grapple with whether software and/or computer implemented inventions meet the requirement of a manner of manufacture, and are not a mere scheme, information or abstract idea.

The Full Federal Court recently delivered its judgment in Aristocrat Technologies Australia Pty Limited v Commissioner of Patents [2025] FCAFC 131 (“Artistocrat”). The case has a tortured history, having been subject to a High Court appeal (where six Justices heard the case and delivered a divided 3:3 opinion), remittal to the Federal Court, and now subject of a further appeal to the Full Federal Court. The case concerned whether the invention, the subject of certain patents relating to feature games and jackpot mechanisms on electronic gaming machines (poker machines), was a manner of new manufacture.

The Full Court (at [131]), adopting the approach set out by Justices Gordon, Edelman and Stewart of the High Court, has indicated that the approach to identifying whether software or a computer implemented invention meets the test of manner of manufacture involves asking whether, properly characterised, the subject matter that is alleged to be patentable is:

  • An abstract idea which is manipulated on a computer; or
  • An abstract idea which is implemented on a computer to produce an artificial state of affairs and a useful result.

The second aspect of this approach adopts the terminology established by the landmark High Court decision of National Research Development Corporation v Commissioner of Patents [1959] HCA 67; 102 CLR 252.

The Full Court considered that this approach is borne out in the prior cases, and includes a useful summary of those cases at paragraphs [132] – [133] of the judgment. At [132] the Full Court notes that:

…  the result in each of the Full Court decisions rejecting the patentability of certain computer-implemented claims was correct, in that each of them merely involved the use of a machine to manipulate an abstract idea rather than involving the implementation of the idea on a machine to produce an artificial state of affairs and a useful result. Those Full Court decisions concern: (a) the scheme for protecting assets from unsecured judgment creditors considered in Grant; (b) the scheme for constructing data concerning a non-capitalisation weighted portfolio of assets considered in Research Affiliates; (c) the scheme for assessing the competency or qualification of people in accordance with recognised standards considered in RPL Central; (d) the method and apparatus for displaying information to provide business intelligence considered in Encompass; and (e) the marketing scheme considered in Rokt. The same observation applies to the use of the GPS-enabled mobile devices for dealing with risk management information considered in Repipe Pty Ltd v Commissioner of Patents [2021] FCAFC 223; (2021) 164 IPR 1 (Perram, Nicholas and Burley JJ), and the improved logistics method considered in Dei Gratia Pty Ltd v Commissioner of Patents [2024] FCA 1145 (Rofe J).

The Full Court further noted at [134] that:

… in the 21st century, a law such has s18(1A) of the Patents Act, that is designed to encourage invention and innovation, should not lead to a different conclusion where physical cogs, reels and motors are replaced by complex software and hardware that generate digital images.

In applying this approach, the Court found that the claims of the Aristocrat patents were directed to a manner of manufacture.  The Australian patent office is in the process of reviewing its guidance in respect of software and computer-implemented inventions in light of this recent decision.[2]

This will be particularly relevant in the Australian patent office, as it has reportedly seen an influx of applications relating to AI technology (including AI hardware, expert systems and machine learning).  Many Gen-AI tool applications may, like earlier cases, simply use machines to manipulate abstract ideas. It will be interesting to observe these applications as they progress through the patent office, particularly in the wake of the recent decision of Aristocrat.

  1. Search and examination

Gen-AI tools have the potential to increase capacity to conduct searches for prior art amongst the patent attorney profession and in the Australian patent office. This is because large language models extend search capabilities well beyond key word searching, as some can comprehend the context, nuance and intent of the text they are reviewing.

The Australian patent office is already deploying tools to increase the efficiency of examiners, including to use machine learning to:

  • sort patent applications into the right technology groups to ensure it’s examined by the relevant teams;
  • determine which patent applications are likely to be ready for examination;
  • search global patent databases for relevant prior art based on applicant and inventor details; and
  • identify and prioritise foreign examination reports from related international applications.

The patent office has also disclosed that it is developing a proof-of-concept AI tool for its patent examiners. The tool will reportedly assist in ranking, comparing, and identifying documents most relevant to an application under examination, and offers quick summaries of key information. It is intended to assist examiners in making faster informed decisions.

It is evident that professionals must be in a position to confidently adopt and use Gen-AI in their day-to-day practice. Those who adapt to these tools will gain a competitive advantage, whilst those that don’t will likely be at a competitive disadvantage.

[1] Harris v CSIRO (1993) 26 IPR 469 at 476, citing Mueller Brass Co v Reading Industries 352 F Supp 1357.

[2] See: Aristocrat Technologies Australia Pty Ltd v Commissioner of Patents – Update | IP Australia