A deep dive into patentability of AI, ownership and entitlement to AI-created inventions

The surge in the use of AI can be attributed to recent developments achieved in relation to algorithms, increased computing power and the availability of vast amounts of data. This in turn raises the question of how the current patent system is responding to AI? 

In part A of this article, we examine the subject matter eligibility standard for AI-created inventions in Australia, commonly known as the "threshold issue".

As machine involvement in inventive processes increases, in part B we look at the current framework for who can be considered an inventor for AI-created inventions in Australia and consider whether a change to the existing framework is required.

A. Patentability of AI

The threshold issue: manner of manufacture

The discussion surrounding the patenting of AI in Australia focuses extensively on what is known as the "threshold issue", namely, whether AI is proper patentable subject matter. Section 18 of the Patents Act 1990 (Cth) (the Act) states that an invention is patentable if it is a "manner of manufacture", the meaning of which has been developed over time by Australian Courts.

In a leading High Court decision in 1959, National Research Development Corp v Commissioner of Patents [1959] HCA 67, the Court avoided imposing a literal interpretation on the phrase "manner of manufacture" and instead adopted the test that the invention must "create an artificial state of affairs of economic utility". The adoption of this approach has meant that patentable subject matter in Australia is flexible and able to develop as new technologies emerge.

However, patent applications covering computer-implemented or machine learning inventions (such as AI) frequently stumble at this "manner of manufacture" test. This is often because as the idea sought to be claimed is abstract and the computer is a tool in which the invention is performed, rather than the invention being based in that computerisation.

Since December 2018, there have been two key decisions of the Federal Court of Australia dealing with this threshold issue. First, in Rokt v the Commissioner of Patents [2018] FCA 1988 (see our article here), Justice Robertson overturned a decision by the Commissioner of Patents and concluded that computer-implemented business methods can meet the "manner of manufacture" test, where the following three conditions are satisfied:

1. the invention solves a technical problem, as opposed to just a business problem, with a technical solution;
2. the computer is integral to the invention, rather than merely an intermediary; and
3. the invention involves steps that are foreign to the normal use of computers.

In reaching his decision, Justice Robertson found that known components had been integrated into a single system in an innovative and previously unknown way. In addition, his Honour concluded that the invention drew together different streams of information and worked with them in a way that was new, making the combination of these techniques new.

In light of these findings, Justice Robertson concluded that the invention brought together some new elements and some known elements to form a working combination that had not previously been achieved and involved the use of computers in a way that was "foreign to their normal use".

Although the decision in Rokt was considered by many to provide a glimmer of hope for the patenting of computer-implemented inventions in Australia, unsurprisingly, the Commissioner of Patents filed an appeal of Justice Robertson's decision to the Full Court of the Federal Court. The hearing of that appeal was stayed pending the outcome of the much-anticipated decision of a 5 judge panel of the Full Federal Court in Encompass Corporation Pty Ltd v Infotrack Pty Ltd [2019] FCAFC 161 (Encompass) which was handed down in September 2019 (see our article here).

In the Encompass first instance decision, Justice Perram determined that the claimed inventions (two innovation patents) were merely a combination of three previously known methods. This led his Honour to conclude that the claimed invention did not result in an "improvement in the computer". As such, Justice Perram found the patents were invalid on the basis that they did not concern a manner of manufacture.

On appeal, in a unanimous decision, the Full Court upheld Justice Perram's findings that the patents were invalid on the basis that they did not concern patentable subject matter. In doing so, the Full Court emphasised that if claims, as a matter of substance, "merely require generic computer implementation" of an otherwise abstract idea, then the claimed invention cannot be a manner of manufacture.

Many in the profession and industry hoped that the Full Court in Encompass would clearly set out the boundary between an abstract idea and a computer-implemented invention. However, the Full Court expressly declined to provide this guidance stating that the appeal did not raise any significant question of principle nor did it provide the occasion for the Court to set out the metes and bounds of patentable computer implemented inventions.

The invention in Rokt does not appear to suffer the same deficiency in description as was relied upon by the Full Court in the Encompass decision. Therefore, perhaps the Rokt appeal will provide the much needed clarity that the profession and industry are seeking on this point.

The distinction between AI and "AI-created inventions"

It is important to draw a distinction between the patentability of AI (computer-implemented inventions which disclose a type of AI) and the patentability of "AI-created inventions", being inventions created by AI systems, which may or may not be computer-implemented inventions.

As mentioned above, questions of the patentability of computer-implemented inventions have been the subject of Australian Court judgments. However, the Courts have not yet considered questions of ownership of, and entitlement to, patent rights for AI-created inventions. Given the recent activities in this space outside of Australia, it will only be a matter of time before this issue comes before an Australian Court.

B. Ownership of and Entitlement to AI-Created Inventions

DABUS

On the surface, a patent application for a food container and a patent application for a flashing light for use in emergencies do not seem like inventions that could disrupt the foundations of patent law. However, the named inventor on each of these applications has raised complicated issues of ownership and entitlement for the United States Patent Office (USPTO), European Patent Office (EPO) and United Kingdom Intellectual Property Office (UKIPO).

Meet DABUS. DABUS is an AI system named as the inventor of the above-mentioned inventions. While DABUS was not created to solve any particular problem and is perhaps best known for creating surrealist art, its training in general information from various knowledge domains has allegedly led it to independently conceive of novel inventions.

DABUS' inventor is Stephen Thaler. Thaler believes that DABUS is the inventor of the inventions and should be named in the patent applications. Thaler, together with the team from The Artificial Inventor Project, argues that "if similar training was been given to a human student, the student rather than the trainer would meet the inventorship criteria as an inventor". However, the challenge with this reasoning is that a student is a person, whereas DABUS is not.

Could AI systems be granted patent rights?

While an AI system may create an invention that meets the requirements for patentable subject matter, the question remains as to whether an AI system has standing to be granted patent rights over that invention?

While the law in Australia does not stipulate who (or what) may apply for a patent, section 15 of the Act allows a patent for an invention to only be granted "to a person who:

(a) is the inventor; or

(b) would, on the grant of a patent for the invention, be entitled to have the patent assigned to the person; or

(c) derives title to the invention from the inventor or a person mentioned in paragraph (b); or

(d) is the legal representative of a deceased person mentioned in paragraph (a), (b) or (c)." (emphasis added)

For the purposes of section 15 of the Act, a "person" means any legal person and includes a body politic and a body corporate, as well as a natural person.

Based on the current state of the law in Australia, similar to trusts and partnerships, AI systems are not currently considered a "person" and therefore could not be granted a patent or associated patent rights. Similarly, an AI system cannot validly assign its rights, and therefore cannot presently be listed as an inventor.

This raises the question of whether current patent laws regarding inventorship need to be revised to take into account inventions where an entity other than a natural person, or company to which a natural person assigns an invention, should be able to be listed as an inventor or an owner of a patent?

Interestingly, the USPTO recently sought public input regarding patenting of AI in an effort to determine whether AI systems can or should be named as an inventor or applicant for inventions arising, at least in part, through AI (e.g., machine learning systems). Notably, the USPTO appears to be trying to understand whether current patent laws should be revised to take into account inventions where an entity or entities other than a natural person contributed to the conception of an invention.

On 28 October 2019, the UKIPO updated its formalities manual for patent examiners by adding the following sentence "An 'AI inventor' is not acceptable as this does not identify 'a person', which is required by law. The consequence of failing to supply this is that the application is taken to be withdrawn."¹   It seems likely that the update to the formalities manual was at least partly in response to the two patent applications filed at the UKIPO that named DABUS as the inventor.

Should AI systems be granted patent rights?

Consideration of whether AI systems could be named as inventors or granted AI rights is often coupled with the question of whether AI systems should be granted patent rights.

The Artificial Inventor Project, led by Professor Ryan Abbott at the University of Surrey, reason that if AI systems cannot be recognised as inventors, the owners of the AI systems will not have any protection for the ideas generated by their work, which may discourage them from pursuing further development. Not recognising the AI as an inventor threatens innovation by "failing to encourage the production of socially valuable inventions" argues Professor Abbott.²

On the other hand, Professor Abbott has previously acknowledged that as creative computers become more and more sophisticated, at some point in the future, it is possible that they could have a disruptive effect on human creativity by pre-empting much or all human invention^.3

In considering these views, it is important to remember the context in which patent rights arise. Patent rights are granted by statute as an exception to the general prohibition against monopolies. That is, patent rights exist to encourage innovation, incentivise commercialisation and enable disclosure of inventions to the public for use after the monopoly period. Therefore, an AI should only be permitted to own a patent, or be named as an inventor of a patented invention, if it can be shown that this step will promote the objectives of the patent system.

While it is highly unlikely that any patent application naming DABUS as the inventor would proceed to grant in Australia based on the current state of the law, we can expect to see further debate on whether AI should be recognised as an inventor and/or owner of patent rights in the coming years.

Authors: Nina Fitzgerald, Partner; Eoin Martyn, Senior Associate; and Caroline Christian, Lawyer. 

^{End Notes}

^{1. Section 3.05 of the Patents Formalities manual issued by the UKIPO.}

^{2. Financial Times A patent predicament: who owns an AI-generated invention? (7 October 2019).}

^{3. Ryan Abbot, 'I Think, Therefore I Invent: Creative Computers and the Future of Patent Law' (2016) 57(4) Boston College Law Review 1120}