Can an AI system be an inventor? Full Court says no

This article was updated on 24 May, 2022 to reflect the special leave application Dr Thaler’s counsel filed in the High Court.

In a decision that brings Australia into line with recent decisions in the US, Europe and the UK, the Full Court of the Federal Court of Australia has unanimously found that an artificial intelligence (AI) system cannot be an ‘inventor’, while also identifying a number of policy issues to be addressed.  

The Full Court overturned Justice Beach’s decision, in which his Honour found that an AI system could be an inventor for the purposes of the Patents Act 1990 (Cth) (Act) and Patent Regulations 1991 (Cth) (Regulations).[1] Instead, the Full Court held an ‘inventor’ must be a ‘natural person’ – based on the historical centrality of the role of a human inventor in patent applications, the scheme of the Act and Regulations, and the natural reading of section 15(1) of the Act.[2]  

The Full Court’s decision follows similar outcomes in the UK,[3] US,[4] NZ[5] and EU[6] where Dr Stephen Thaler had filed similar patent applications.  

While the Full Court’s decision seemingly represents a setback for industries employing AI systems in research and development, it has helpfully set the stage for a policy debate around this issue in Australia. Importantly, the Full Court identified several policy questions raised by AI inventorship and clarified that there may be scope for patents to be granted where an invention developed using AI might be regarded as having a human inventor.[7] 

As was widely assumed to be the case prior to Justice Beach’s decision, innovators seeking patent protection for inventions created utilising AI systems must name human inventors.

Background – The Deputy Commissioner and Federal Court’s decisions

Dr Stephen Thaler filed Australian Patent Application No. 2019363177 for ‘Food container and devices and methods for attracting enhanced attention’ (Application),[8] in which his AI system “DABUS” was named as the ‘inventor’.[9]

As we reported early last year, the Deputy Commissioner of Patents found the Application had lapsed for invalidity as it was a legal impossibility for DABUS to be ‘the name of the inventor of the invention’ which was required to be provided in the application by Regulation 3.2C(2)(aa).[10] 

Dr Thaler sought judicial review in the Federal Court where Justice Beach held that an AI system could be the ‘inventor’ for the purposes of the Act and Regulations.[11] His Honour reasoned that ‘an inventor is an agent noun; an agent can be a person or a thing that invents’ and nothing in the Act or Regulations precluded an AI ‘inventor’.[12] His Honour recognised the benefits of AI based inventions in, for example, the pharmaceutical industry and considered that a finding of AI inventorship was consistent with the object of the Act to ‘promote economic wellbeing through technological innovation’.[13]  The Commissioner of Patents appealed to the Full Court.[14]

The Full Court’s decision         

The Full Court considered ‘the central question’ before it was ‘whether a device characterised as an artificial intelligence machine can be considered to be an “inventor” within the meaning ascribed to that term’ in the Act and Regulations.[15]  The Full Court adopted a conventional text based approach to statutory construction,[16] cautioning against an approach whereby the Court imputes what it regards as desirable policy to the legislation and then characterising that as the purpose of the legislation.[17]

The Full Court identified that ‘inventor’ was not defined in the Act and Regulations,[18] and then considered the Explanatory Statement which stated that the purpose of Regulation 3.2C(2)(aa) was ‘to ensure that the entitlement of the applicant to be granted a patent is clear’.[19] This statement provided the starting point for the Full Court’s subsequent analysis of an inventor’s role in the ‘path to entitlement to the grant of a patent’, particularly in relation to s 15(1) of the Act.[20]  

The Full Court’s analysis revealed that patent law had proceeded on the assumption that the inventor was a natural person whose ingenuity was to be rewarded by a grant of monopoly over their invention.[21] The Full Court found that historically the inventor had to have human attributes, including the ability to make representations on the nature of the invention and possess a legal personality, being the person from whom an entitlement to the grant of a patent had to be derived.[22]  

Further, the concept of an ‘inventor’ had to be workable within s 15(1) of the Act, which regulated the entitlement to a grant of a patent.[23]  

S 15(1) of the Act provides that a patent for an invention may only be granted to a person who:

1. is the inventor; or
   
   
2. would, on the grant of the patent for the invention, be entitled to have the patent assigned to the person; or
   
   
3. derives title to the invention from the inventor or a person mentioned in paragraph (b); or
   
   
4. is the legal representative of a deceased person mentioned in paragraph (a), (b) or (c).

The Full Court found that the term ‘inventor’ within s 15(1)(a) denoted any person who materially contributes to or supplies the inventive concept.[24] The Court found that ‘the law relating to the entitlement of a person to the grant of a patent is premised upon an invention for the purposes of the Patents Act arising from the mind of a natural person or persons’.[25] Moreover, as s 15(1) specifically referred to a ‘person’ it was clear the ‘inventor’ in s 15(1)(a) had to be a ‘natural person’.[26]

On a ‘natural reading’ of s 15(1), ‘each of ss 15(1)(b), (c) and (d) provide for circumstances where a person becomes entitled to the grant of a patent by ultimately receiving that entitlement from the inventor in s 15(1)(a)’.[27] For the purposes of s 15(1)(b), considered in context, the assignment had to come from the inventor in s 15(1)(a) who also had to be a natural person to give effect to an assignment.[28]  Similarly, s 15(1)(c) required derivation of title from the same natural person who was the ‘inventor’ in s 15(1)(a), as the Full Court doubted the Parliament would have intended to ascribe a different meaning to ‘inventor’ in s 15(1)(a) and (c).[29]  

Having noted their agreement with important aspects of the UK Court of Appeal’s reasoning in Dr Thaler’s UK application, where the outcome was the same, the Full Court clarified that their task had been focussed on the language of the Act which materially differs from equivalent UK patents legislation.[30]  

Key policy implications

While the decision brings the Australian position on inventorship into line with other major jurisdictions, the impact on investment and innovation requires consideration from a policy perspective. There remains uncertainty about the extent to which intellectual property (IP) rights can ever subsist in AI devised inventions.

The concern, expressed in a recent UK IP Office public consultation paper, is that ‘if AI devised inventions are unable to be patented, there may be less investment in this technology’ or this ‘may encourage the use of trade secrets, which could harm follow-on innovation’.[31]  On the other hand, it is recognised that if an AI system could be classified as an inventor, there is the potential for large volumes of patents to be held by a small number of companies with access to the best AI technologies and training data.[32]  

The UK Government has sought responses on four policy options related to AI generated inventions:[33]

• Option 0: Make no legal change
  
  Only humans can be named as inventors in patent applications. Some stakeholders noted this approach was sufficient as most inventions involving AI would have a human inventor in the short term.  
  
  
• Option 1: Make the term ‘Inventor’ expanded to include humans responsible for an AI system which devises inventions
  
  Where no humans qualify as inventor, the inventor could be the human ‘by whom the arrangements necessary for devising of the invention are undertaken’. The people involved in the following activities could potentially be considered human inventors: ‘programming the AI, configuring the AI, operation the AI, selecting input data such as training data for the AI or recognising applications of the output of AI.
  
  
• Option 2: Allow patent applications to identify an AI system as an inventor
  
  It would be transparent that a non-human has devised the invention, either by amending legislation to allow an AI system to be named as inventor, or to remove the requirement to name an inventor if the invention is devised by an AI system. The human who made the arrangements necessary for an AI system to devise the invention would own the patent rights, i.e. those undertaking the activities listed in Option 1.
  
  
• Option 3: Protect AI-devised inventions through a new type of protection
  
  A sui generis right, parallel to the patent system, for inventions that fail to qualify for patent protection as no human can be identified. The right could involve alterations to the traditional system including a modified test for inventive step and shorter term of protection.  

Where to next for AI inventors in Australia?

The Full Court observed that debate as to ‘the role that artificial intelligence may take within the scheme of the Patents Act and Regulations’ was ‘important and worthwhile’.[34]  

The policy questions identified by the Full Court closely resemble those posed by the UK IP Office suggesting the findings from the UK consultation may be influential in the Australian context. 

The Full Court identified several ‘urgent’ policy questions including whether an ‘inventor’ should be redefined to include AI, and if so, to whom such an AI invented patent could be granted in respect of its output, what the standard of inventive step would be or what would be the continuing role of the ground of revocation for false suggestion or representation.[35]  

Similar to the suggestions in Option 1 and 2 above, the Full Court considered possible candidates for the grant of a patent based on an AI’s invention would include ‘one or more of: the owner of the machine upon which the artificial intelligence software runs, the developer of the artificial intelligence software, the owner of the copyright in its source code, the person who inputs the data used by the artificial intelligence to develop its output, and no doubt others’.[36] 

The Full Court’s observations were made on the basis of the uncontested fact that ‘Dr Thaler is not the inventor’,[37] implying a factual possibility that DABUS was capable of being the inventor. In contrast, according to IBM, for example, ‘intelligent machines will remain tools that assist humans, rather than invent independently, for a considerable time.’[38]  

It will be necessary to consider whether empirical evidence can establish if in fact: 

• current AI systems are sufficiently advanced to independently invent; and
  
  
• legal recognition of AI as an inventor in patent applications will incentivise innovation.

The issue of inventorship will likely need to form part of a broader law reform conversation focussing on patents and copyright in particular. For instance, unlike the unique position in the UK,[39] Australia does not yet have copyright laws which deem the person by whom the arrangements necessary for the creation of the work are undertaken to be the author of computer generated literary, dramatic, musical or artistic work.  

Dr Thaler’s counsel has filed a special leave application in the High Court. Subject to the outcome in the High Court, the Full Court’s decision clarifies that an Australian patent application naming an AI system will be unable to proceed.

In the meantime, we will watch this space for the outcome of Dr Thaler’s special leave application and any responses by the Australian Patent Office to the Full Court’s decision.