08 March 2022

Australian Appeals Court Seems Sceptical of Push to Name DABUS ‘AI’ as Inventor

I am not a robotOn 9 February 2022, the appeal by the Commissioner of Patents against the decision of Justice Beach finding that the ‘AI’ machine known as DABUS (‘Device for the Autonomous Bootstrapping of Unified Sentience’) could be named as inventor on a patent application filed by Dr Stephen Thaler, was heard before an expanded Full Bench of the Federal Court of Australia.  The five-judge panel comprised Chief Justice Allsop, and Justices Nicholas, Yates, Moshinsky and Burley.  The appointment of an expanded panel (three judges is usual), including the Chief Justice, is notable, and suggests that the court considers the question of whether a machine can be an inventor for the purposes of the Patents Act 1990 to be one of particular legal importance.  The case was argued for Dr Thaler by David Shavin QC (who also appeared at first instance), and for the Commissioner of Patents by Sophie Goddard SC (appearing for the first time in this matter).  Hamish Bevan (who represented the Commissioner at first instance) argued the Commissioner’s case in reply.  The proceedings were conducted via video conference, and I observed online.

I think that it is fair to say that Dr Thaler did not experience such a smooth ride on appeal as he did at first instance.  Where Justice Beach seemed positively enthusiastic about the idea of allowing patent applications to be filed, and patents granted, for inventions autonomously devised by machines, the judges on the Full Bench appeared sceptical of a number of the arguments presented on behalf of Dr Thaler, and at times subjected Mr Shavin to fairly sharp questioning and criticism.  There are some aspects of the original decision that are unlikely to survive the appeal, although Dr Thaler only requires one line of reasoning to stand in order to prevail.

Based on the arguments presented, there are a number of sub-questions that the Full Court may need to decide.  First, there is a question of whether the case is simply about determining the meaning of the word ‘inventor’ as a matter of pure statutory construction, or about deciding whether or not to develop the concept of the inventor according to a common law methodology.  Either way, the court then needs to decide whether or not a non-human, machine inventor should be included.  In doing so, it may need to consider whether non-human inventors are compatible with other provisions of the Act, and in particular with the section that deals with the entitlement to be granted a patent.  A machine cannot possess or assign intellectual property rights, so provisions that most naturally apply to transfers of title between (natural or legal) persons may not necessarily be compatible with ‘AI’ inventors.  Justice Beach found that two different mechanisms specified in the legislation could apply to enable Dr Thaler to claim ownership of patent rights on inventions made by DABUS.  He will need at least one of these to survive the appeal.

There are a number of avenues by which the Commissioner could succeed in her appeal, but by the same token there are multiple opportunities for Dr Thaler’s defence of the appeal to be successful.  It is difficult to predict which lines of reasoning the court may adopt, and indeed it is not unlikely that the judges will hand down multiple concurring and/or dissenting opinions.  On balance, my sense is that the appeal is more likely than not to succeed – perhaps by a majority rather than unanimous judgment.  If so, then Australia will rejoin the many other jurisdictions in which corresponding applications naming DABUS as inventor have been rejected.

The remainder of this article provides a summary of the major arguments presented at the hearing.  It is somewhat lengthy, but considerably less so than the full day spent at the Federal Court!

The Role of ‘Agreed Facts’

As a precursor to my analysis of what took place at the hearing, it is worth noting the existence of certain ‘agreed facts’ in the case.  For readers who may be unfamiliar with the concept, an ‘agreed fact’ is something for which there is not necessarily any evidence, but which the parties nonetheless stipulate should be assumed to be true for the purposes of the court proceedings.  Under section 191 of the Evidence Act 1995, agreed facts are taken to be undisputed, need not be proven by evidence, and cannot be contradicted or qualified by evidence.  Ideally, this is because they are actually true.  But sometimes – as here – they are employed to test an application of the law under the assumption that they are true.

Here, it has been agreed that DABUS devised the inventions disclosed and claimed in Thaler’s patent application autonomously.  Further, it has been agreed that it did so without human control or direction from Dr Thaler (or anybody else).  This has the consequence that it is not open to the court to find that DABUS is not the inventor, or that there is any person who may be considered as inventor, either in place of, or jointly with, DABUS.

Interpreting ‘Inventor’ in Australian Patent Law

The question of whether DABUS can be validly named as the inventor on Thaler’s Australian patent application arises as a result of the requirement, in Regulation 3.2C(2)(aa), that the applicant of an application filed as the national phase of an international application under the Patent Cooperation Treaty (PCT) must ‘provide the name of the inventor of the invention to which the application relates’.  Thaler’s PCT application entered the national phase in Australia, identifying the inventor specifically as ‘DABUS, The invention was autonomously generated by an artificial intelligence’.  The Commissioner of Patents took the view that inventors for the purposes of the Patents Act 1990 must be natural persons, and that as DABUS is not a natural person, no inventor had been supplied for the purposes of Regulation 3.2C(2)(aa).  When Thaler declined to amend the application to name a human inventor – instead arguing that DABUS can and should be recognised as the inventor – the Patent Office ultimately determined that the application did not comply with the required formalities, and that it had therefore lapsed under Regulation 3.2C(5).

On the face of it, then, the only issue before the court is: what is the meaning of the word ‘inventor’ in Regulation 3.2C(2)(aa) (and, by implication, elsewhere in the Act and Regulations), and does it encompass DABUS in circumstances where it is agreed that DABUS is an AI machine that autonomously devised the inventions claimed in Thaler’s application?

And yet, very little of the argument before the Full Court related to the meaning of the word ‘inventor’, per se.  As in the original hearing before Justice Beach, the overwhelming majority of time, on behalf of both the Commissioner and Dr Thaler, was spent on discussion of Thaler’s entitlement to be granted a patent under subsection 15(1) of the Patents Act.  Quite early on in the Commissioner’s submissions, the Bench questioned how section 15 became involved, and how Justice Beach came to be adjudicating on how title is derived.  The response given by Ms Goddard was twofold.

  1. Firstly, Regulation 3.2C(2)(aa) is a relatively recent provision, added only in 2015 by the Intellectual Property Legislation Amendment (TRIPS Protocol and Other Measures) Regulation 2015.  The Explanatory Statement indicated that the reason for the provision is that information regarding the identity of the inventor ‘is required to ensure that the entitlement of the applicant to be granted a patent is clear.’  Accordingly, section 15 of the Act is implicitly invoked by the provision, by way of the Explanatory Statement.  And while entitlement is not an issue that arises at the time of filing (since any person may file an application), it must be established prior to acceptance of an application, and therefore arises immediately upon examination.
  2. Secondly, Dr Thaler himself made submissions as to how he could satisfy the requirements of paragraphs 15(1)(b) and (c) so as to be entitled to the grant of any patent issued on the application.

More generally (although I am not sure that this was expressly argued), since ‘inventor’ must mean the same thing throughout the Act and Regulations, that meaning should be consistent with sensible operation of the entitlement provisions in section 15.  There is little point in permitting an application to proceed if it cannot ultimately be granted due to the non-existence of any legally entitled recipient!

The Meaning of ‘Inventor’

On behalf of Dr Thaler, Mr Shavin reiterated substantially the same submissions that had been successful before the primary judge, i.e. that ‘inventor’ is an ‘agent noun’ (like ‘computer’, which once referred to a person but is now almost exclusively a machine) that can mean either a person or a thing that invents.  The Bench seemed to be interested in exploring this further, asking whether this would include any ‘thing’ – including an ordinary (non-AI) computer program, or even an animal – and also whether Mr Shavin was suggesting that the current meaning of the word ‘inventor’ includes an AI machine.  Mr Shavin appeared to baulk at the idea that an animal might be an inventor.  However, if I understood him correctly, he argued that ‘inventor’ is an ‘agent noun’ as a matter of grammar, and that this is the correct approach to construction regardless of actual usage.  I must say that I have significant difficulty with the notion that a statute should be construed purely on the basis of grammatical ‘fit’, without regard to what the community would regard as the meaning of the language employed.  And I got the impression that some of the judges may have had similar difficulties.

For the Commissioner, Ms Goddard maintained that according to its ordinary meaning an ‘inventor’ is a human being.  Furthermore, the policy underlying the patent system has always been to reward and encourage human ingenuity.  The emphasis in earlier cases has been on the ingenuity of the human inventor, e.g. in the context of whether the person identified was the ‘true and first inventor’.  Clearly, a machine cannot be incentivised to invent by the promise of a patent reward.  In this context, it was submitted that construing the word ‘inventor’ to encompass an AI machine involves an extension of the concept that is beyond the limits of judicial lawmaking, and that should be left for the legislature following appropriate community consultation.

In this regard, the Commissioner picked up the reasoning that had been effective for Dr Thaler before the primary judge, based on the High Court’s decision in D'Arcy v Myriad Genetics Inc [2015] HCA 35, that the word ‘inventor’ – like the concept of ‘manner of manufacture – invokes 400 years of history and common law development.  Ms Goddard observed that the plurality in Myriad had referred to the requirement for ‘human action’ in making an invention (which was why, in that case, a naturally occurring genetic sequence was not patentable).  She also noted that the High Court had cautioned against judicial extension of the concept of ‘manner of manufacture’ to encompass new classes of subject matter without due consideration of the consequences in terms of public policy and balancing conflicting interests.  She submitted that a similarly cautious approach should apply in the case of the term ‘inventor’.

There was a challenge from the Bench (I think it was the Chief Justice) to the effect that the present case involves a question of statutory construction, whereas Myriad was about the possible extension of the concept of ‘manner of manufacture’ according to a common law methodology.  That is to say, here we are concerned only with the meaning of a word, whereas in Myriad the ‘meaning’ of the phrase ‘manner of manufacture’ was not the issue, because it acts as a placeholder for the evolving scope of patentable subject matter which has been expressly left to the courts for development.  There is some force to this contention although, as Ms Goddard pointed out, the issue needs to be addressed because it was invoked by the primary judge, and formed part of his reasoning at first instance.

Nonetheless, perhaps as a result of this scepticism from the Bench, it seemed to me that Mr Shavin retreated somewhat from the ‘common law development’ argument that had proven so effective at first instance, in favour of a more conventional ‘statutory construction’ approach.  Whether this will help is not clear to me.  It will require the court to accept that the word ‘inventor’ should be interpreted to have a meaning for which there is no evidence, either in the form of dictionary definitions, or any examples of actual usage encompassing an ‘AI machine’ inventor.

Speculative Entitlement – Paragraph 15(1)(b)

If the word ‘inventor’ is to encompass non-human actors, this must be compatible with the overall operation of the Patents Act.  There would be no point in recognising a machine inventor if there were no mechanism for a patent to be granted to anybody.  And, under section 15 of the Act, a patent may only be granted to ‘a person’ (natural or legal) falling into at least one of four exhaustive categories.  The first of these (paragraph 15(1)(a)) is the inventor, and all parties agree that DABUS is not a person, and cannot qualify.  The fourth (paragraph 15(1)(d)) is the legal representative of a deceased person, which clearly cannot apply here.  That leaves paragraphs 15(1)(b) and (c).

Paragraph 15(1)(b) represents a form of ‘speculative’ entitlement on the part of a person who ‘would, on the grant of a patent for the invention, be entitled to have the patent assigned to the person’.  The most common application of this provision is in the case of an applicant that is the employer of the original inventor.  Typically, the rights of the employer are secured by an employment contract – either in express terms, or at common law – and a specific assignment may or may not be executed by the inventor.

Arguing on behalf of Dr Thaler, Mr Shavin asked the court to accept that paragraph 15(1)(b) could apply in case where an inventor, such as DABUS, is not capable of acquiring, possessing, or assigning any rights.  Hypothetically, at least, Dr Thaler might be entitled to have a future patent assigned to him, even though DABUS is legally and practically incapable of providing such an assignment.  Listening to this argument, I was reminded of the scene in Monty Python’s Life of Brian in which Stan – sorry, Loretta – wants to be recognised as a woman (perfectly fine) because she wants to have babies (somewhat more problematic, from a biological perspective).  Judith has the solution: the collective should ‘agree that he can’t actually have babies, not having a womb, which is nobody's fault – not even the Romans’ – but that he can have the right to have babies’ (misgendering in original).  The logic here – of breathing life into a hypothetical right, despite the lack of any practical means of giving effect to it – seems comparably absurd.

In his decision, Justice Beach provided some hypothetical examples in which paragraph 15(1)(b) might apply without any participation from the inventor, such as when an invention is stolen (and the thief might be obliged to assign a resulting patent), or when an invention is made under contract by a company that, in turn, employs the inventor.  But, as Ms Goddard pointed out in her submissions of behalf of the Commissioner, these examples suppose the interposition of some intermediary.  This mechanism does not solve the problem of establishing a chain of title back to an inventor that could never hold title.  Nor, importantly, do any of the proposed hypotheticals actually correspond to the agreed facts in this case.  There is not said to be any intermediary between DABUS and Dr Thaler.

The judges on appeal appeared to agree with the implausibility of paragraph 15(1)(b) applying, with considerable criticism being directed to Mr Shavin from the Bench.  The judges who commented seemed to be of the view that since a machine cannot obtain a patent, cannot be possessed of the right to obtain a patent, and cannot assign any right (that it does not, in any event, possess), further reference to thieves or employees could be of no assistance to Dr Thaler.  My impression was that the arguments based on paragraph 15(1)(b) are likely dead in the water.

Derived Entitlement – Paragraph 15(1)(c)

Paragraph 15(1)(c) is broader in its application, bestowing entitlement upon a person who ‘derives title to the invention from the inventor or a person mentioned in paragraph (b)’.  For Dr Thaler, Mr Shavin argued (as he had, successfully, before the primary judge) that the concept of ‘derivation’ is broad and not limited to assignment or transfer of title, and that paragraph (c) does not require that the source of title be a person, nor that this be the same person (i.e. a human inventor) referred to in paragraph (a).  In particular, he relied upon a principle of possessory title, i.e. that Dr Thaler can ‘derive’ title as the first person to come into possession of an invention generated by DABUS, particularly in circumstances where he also owns and controls the machine and associated software. 

Mr Shavin further submitted that title to inventions generated by DABUS must flow somewhere, for otherwise there may be an invention that cannot be patented simply because there is no person who can receive the grant, even if all the other requirements for patentability are satisfied.

Ms Goddard, arguing for the Commissioner, submitted that the ‘inventor’ must be the same in paragraphs (a) and (c), and therefore must be a person.  Accepting that ‘derives’ means to receive or obtain from a source, the question primarily concerns the source of origin, and not so much the obtaining.  The ‘property’ at issue, it was argued, is not the invention itself, but the right to apply for a patent – which is something that DABUS never has, and which therefore cannot be ‘derived’ by Dr Thaler.  Furthermore, to the extent that the term ‘possession’ appears in prior patent decisions, it is referring to something different from the possessory title relied upon by Dr Thaler.  Historically, possession in patent law has related to conception of an invention, i.e. a requirement that the inventor must have a sufficiently formed mental idea of the invention before it is set down and claimed in a patent specification. 

Additionally, to the extent that Dr Thaler relies on ‘progeny law’ (e.g. that the progeny of livestock, or the fruits of a tree, are the property of the owner of the progenitor), Ms Goddard submitted that the analogy is not applicable to intangible property, and that this does not, in any event, resolve the problem that if DABUS never had any right to apply for a patent in the first place, then there was nothing of which Dr Thaler could relevantly come into possession.

While the judges seemed more open to the possibility that paragraph 15(1)(c) might apply, as compared with 15(1)(b), I nonetheless sensed considerably more scepticism than had been expressed by Justice Beach at first instance.  Questions from the bench, particularly those directed to Mr Shavin, suggested that some of the judges, at least, might not be convinced that it is apt to speak of a person deriving title from a machine.  As one observed, in the case of progeny, for example, nobody would say that the farmer ‘derives title [to the offspring] from the livestock’.  Certainly, possession is a much less slippery concept in the case of tangible property!

Other Matters – ‘Evidence’, Objects and a Lacuna

Many things were discussed over the course of the day’s submissions.  Three are worthy of brief mention in this review.

Firstly, Ms Goddard noted that the extensive discussion of policy issues and (alleged) factual matters – such as the role of AI in the development of new pharmaceutical products – in the appealed decision was based on matter that was not in evidence.  I think that is right, and Mr Shavin did not seem inclined to argue the point too vehemently, choosing instead to focus on the broader policy objectives of the patent system – particularly the disclosure function of patent publication – rather than the specifics of any particular industry.  I suspect that, in the absence of any actual evidence before the court, the judges on appeal will have little regard to the existing role of AI in the process of invention.  There is plenty of other material before them upon which to base their decision(s).

Secondly, the application of the recently-added objects clause in section 2A of the Act was, of course, discussed.  Ms Goddard submitted that the primary judge failed to give sufficient consideration to the balance of interests of ‘producers, owners and users of technology and the public’, and focussed overly on the interests of producers and owners of technology, i.e. those stakeholders who most directly benefit from the grant of patent rights.  As I have already noted, Mr Shavin directed the court in particular to the disclosure function of the patent system, arguing that the benefits of publication would not accrue for an entire category of inventions (i.e. those devised by machines) that fall outside the patent system, and thus might instead be retained as trade secrets or confidential information.  I am not sure that this is an especially persuasive argument.  Nobody is compelled to participate in the patent system.  Many inventions that can feasibly be kept secret already are kept secret, while others are inevitably disclosed if and when they are commercialised or put into public use.  Furthermore, there are already categories of invention, delineated by subject matter, that are excluded from the patent system.  In the absence of any relevant evidence, it is difficult to draw any conclusion as to how the overall balance is affected by treating inventions devised by AI machines differently from inventions devised by humans.

Thirdly – and related to the above point – a number of the judges showed considerable interest in the question of whether or not there is a ‘lacuna’ (i.e. ‘an unfilled space; a gap’) in the Act, as argued by Mr Shavin.  Some on the Bench seemed to harbour suspicions that AI machines might be tools, like any other, and that as a practical matter their human operators could (and, perhaps, should) be named as inventors.  This line of questioning was frustrated by the agreed facts, which force the court to assume that DABUS is, indeed, the sole inventor, precluding any further consideration of Dr Thaler’s role.  However, even if the lacuna is real, the question remains of whether the court should fill the gap by construing ‘inventor’ to include an AI machine, or whether this is a matter for the legislature.

Conclusion – Likelihood of Success in the Appeal

There are a number of ways that the Commissioner can succeed in the appeal.  Firstly, the court may simply decide that the word ‘inventor’ in Regulation 3.2C(2)(aa) cannot, on any view, encompass a machine.  If the court adopts a ‘simple’ statutory construction approach, it would need to accept that ‘inventor’ has a meaning – encompassing AI machines – that is supported neither by dictionary definitions nor common usage, merely because it is grammatically viable.  If, on the other hand, it adopts a common law methodology (analogous to the High Court’s approach to ‘manner of manufacture’ in Myriad), it would need to consider an extension of the concept of an ‘inventor’ to be within the limits of judicial lawmaking, and not something best left for legislative determination.

Even if the court accepts that the term ‘inventor’ could encompass a machine, it might find that this is incompatible with the scheme of the Act in general, and with subsection 15(1) in particular.  From my observation, I believe that there is little prospect that the primary judge’s finding that paragraph 15(1)(b) could apply will survive the appeal.  The Bench seemed fairly scathing in its criticisms of Mr Shavin’s submissions on this point.

That leaves only paragraph 15(1)(c) to provide a mechanism enabling a person to gain entitlement to the grant of a patent from a non-human inventor.  This requires the court to accept that title can be derived via possession of the machine and/or its output.  There are clearly arguments for and against this proposition. However, the fact that Lord Justice Arnold (with whom Lady Justice Laing agreed), in the English Court of Appeal, found that there is ‘no rule of law that a new intangible produced by existing tangible property is the property of the owner of the tangible property’ certainly does not assist Dr Thaler.  The scheme of the Australian legislation may differ from that of the UK, but the principles of possession relied upon are derived from the common law, and there is no suggestion of any divergence up until now.

On balance, my sense is that the Commissioner will have at least a majority of the five judges on her side, and that the appeal is more likely than not to be successful.  But perhaps that is merely wishful thinking on my part – my views on the matter are no secret!


Post a Comment

Copyright © 2014
Creative Commons License
The Patentology Blog by Dr Mark A Summerfield is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Australia License.