Introduction
Part 1 of this blog post series introduced DABUS, an artificial intelligence (AI) machine and its two AI-generated inventions. We looked at treatment of AI inventorship in the context of patent applications in various jurisdictions. We concluded the discussion by briefly introducing a recent Australian Federal Court decision.
In Part 2 of this blog post series, we will analyze the Australian Federal Court’s judgment in more detail, and compare it to decisions reached by the United Kingdom Intellectual Property Office, the European Patent Office, and the United States Patent and Trademark Office. Finally, we will consider how a similar analysis might unfold in Canada.
The legal personality of AI (or lack thereof) and inventors as natural persons
The United Kingdom Intellectual Property Office (UKIPO) was the first to reject two patent applications that listed DABUS as the inventor.1 Looking at the wording of sections 7 and 13 of the Patents Act 1977,2 the UKIPO found a “clear expectation” that the terms “inventor” and “person” referred to a natural person (i.e., a human) and not an AI machine.3
The English High Court (High Court) affirmed the UKIPO’s rejection.4 The High Court did not distinguish between ownership and inventorship. Rather, the High Court’s analysis focused on DABUS having no legal personality and therefore no legal capacity to transfer its rights to Dr. Thaler:
“There must either be an application by the inventor within section 7(2)(a) (which cannot be made because DABUS is not an inventor nor a person) or the inventor must have transferred the right to apply enabling Dr. Thaler to apply under one of section 7(2)(b) or (c) (which again cannot be in this case).”5
The English Court of Appeal (Court of Appeal) has since rejected Dr. Thaler’s appeal and upheld the UKIPO’s and the High Court’s decisions.6 Although it was a 2-1 majority decision, all three judges agreed that only a natural person can be classified as an inventor.7
The European Patent Office (EPO) and Boards of Appeal (BOA) adopted a similar approach to ownership and inventorship, also focusing on the requirement for inventors to be natural persons. Tying both ownership and inventorship to having “legal personality,” the EPO stated:8
“AI systems or machines have at present no rights because they have no legal personality comparable to natural or legal persons. Legal personality is assigned to a natural person as a consequence of their being human…”9
The oral proceedings for the appeal are scheduled to be heard on 21 December 2021. One of the issues the BOA will consider is Article 52 of the European Patent Convention and the degree to which human intervention is a required element of a patentable invention. This issue was brought forward by the president of the EPO and his request to provide commentary was granted.
An application listing DABUS as an inventor was similarly rejected in the United States. The United States Patent and Trademark Office’s (USPTO) response to the patent application focused on the statutory language of 35 U.S.C. § 100, 101, and 115 for the terms “inventor” and “joint inventor” as referring to natural persons.10 The US Court of Appeals for the Federal Circuit reaffirmed this requirement.11 Moreover, the Manual of Patent Examining Procedure (MPEP) indicates that the threshold question for inventorship is “conception.”12 Conception is defined as “the complete performance of the mental part of the inventive act.”13 The USPTO reasoned that the use of terminology such as “mental” and “mind” in the MPEP indicated that conception must be performed by a natural person.14
Australia’s take on AI inventorship
While the United Kingdom, Europe, and the United States have all rejected the notion of AI inventorship, Australia has opened up the possibility. On 30 July 2021, the Australian Federal Court in Thaler v Commissioner of Patents (Thaler),15 reversed the Australian Patent Office’s rejection of Dr. Thaler’s patent application and found that an AI machine could be an inventor. This decision marks the first instance where a jurisdiction with a formal patent prosecution process has accepted an AI machine as an inventor.
Below, we summarize the key aspects of the court’s reasons in Thaler. (A more in-depth discussion of Thaler can be found here.)
The court approached the topic of AI inventorship by separating the question of whether an AI machine can own or control a patent from whether an AI machine can be an inventor.
In terms of ownership and control over a patent, the court held that only a human or other legal person can be an owner, a controller, or a patentee and an AI machine cannot own a patent:16
- Inventions are a class of intangible assets capable of physical possession.
- Such possession would give rise to ownership.
- Dr. Thaler is the owner, programmer and operator of DABUS.
- Dr. Thaler is the owner of these inventions and any inventions made by DABUS when they came into his possession.17
This part of the court’s analysis was consistent with the decisions from the United Kingdom, Europe, and the United States, discussed above.
However, the court found that just because an AI machine could not own a patent did not preclude an AI machine from being an inventor. In other words, the question of inventorship is entirely separate from ownership and control. In terms of inventorship, the court found that an AI machine can be an “inventor”:18
- There is no provision in the Australian Patents Act19 that expressly rejects the idea of an AI system being an inventor.
- There is no specific aspect of patent law that would drive a construction of the Patents Act as excluding non-human inventors (unlike copyright law requiring a human author or existence of moral rights).
- The word “inventor” is not defined in the Australian Patents Act or Patents Regulations,20 and can be construed to be either a person or a thing.
- The concept of “inventor” should be treated in a flexible and evolving way to align with the widening conception of “manner of manufacture” and the overall scheme of the Patents Act.
- The approach to construction should be consistent with the object clause of the Patents Act,21 which states that:
“The object of this Act is to provide a patent system in Australia that promotes economic wellbeing through technological innovation and the transfer and dissemination of technology. In doing so, the patent system balances over time the interests of producers, owners and users of technology and the public.” 22
The court reasoned that permitting AI systems to be inventors is consistent with the Patents Act by incentivizing and promoting technological innovation, leading to new scientific advantages.
In summary, the Australian Federal Court distinguished between patent ownership and inventorship and found that nothing in the Australian legislation restricted inventorship only to legal (or natural) persons.23 Further, permitting AI inventorship would promote the policy objectives of the Patents Act.
Canada’s likely treatment of AI
While Dr. Thaler did not apply for patent protection over DABUS’ inventions in Canada, it raises questions about how such an application would be received in Canada. As previously mentioned in Part 1 of this blog post series, the Canadian Patent Act does not define the term “inventor” or specify whether such an inventor must be human.24 Similar to Australia, there is no provision that expressly rejects the idea of an AI machine being an inventor.
The Supreme Court of Canada (SCC) shed some light on the term “inventor” in Apotex Inc. v Wellcome Foundation Ltd. (Wellcome): “the person or persons who conceived of” the invention, thus requiring that an inventor must at least be a “person.”25 However, this was in the context of co-inventorship between humans and the SCC did not address the possibility of AI-conceived inventions. This distinction is important because the SCC expressed the view that inventorship is a fact-driven inquiry.26 Thus, one reading of the SCC’s decision in Wellcome may be that inventors must be “persons” and therefore natural persons in line with the EPO, UKIPO, and USPTO decisions. However, another interpretation is that “inventor” is not defined in the Canadian Patent Act and that the Australian Federal Court’s interpretation may be open in Canada as well.
Another point to consider is whether Canada would separate the concepts of ownership and inventorship, as this was the notable distinction in the Australian decision. Currently, there are differences in treatment of ownership and inventorship in the Canadian patent system. For instance, inventorship cannot be modified by an agreement among the parties: a party either is or is not an inventor, and this is determined on a factual basis. However, ownership may be modified by agreements. The most common example is an employment agreement that may stipulate any inventions created during an employee’s course of employment belong to the employer.27
Moreover, patent owners can transfer or license their rights to others under section 49 of the Patent Act, but it does not confer the ability to transfer inventorship. In Thaler, the Australian Federal Court reasoned that Dr. Thaler prima facie obtained title from DABUS by way of Dr. Thaler’s possession of DABUS, his ownership of copyright in DABUS’ source code, and his ownership and possession of the computer on which DABUS resides, circumventing the need for an assignment.28 The reasoning in Thaler may translate over to Canada but would need to be tested in Canadian law. If Canada does agree that an AI machine can be an inventor, further questions must be addressed, such as whether an AI machine can be an applicant for a patent. The door for AI inventorship may be open in Canada, but whether that could support a patent application being accepted is unclear.
It is also worth noting that the purpose of the Canadian patent system parallels the object clause of Australia’s Patents Act; namely, to promote innovation and economic growth and well-being.29 In Canada, the courts take a purposive approach to statutory interpretation.30 However, unlike the Australian Patents Act, this purpose has been described in case law and is not explicitly described in the Canadian Patent Act.
To summarize, the jurisdictions we have discussed all agree that an AI machine cannot own or control a patent due to a lack of legal personality. Canada is unlikely to stray from this view. However, Canada’s position on AI inventorship is less clear. The absence of a definition in the Patent Act for “inventor” eliminates the express prohibition of AI inventorship that we have seen in jurisdictions such as the United States, subject perhaps to the SCC’s comments in Wellcome.
Moreover, the distinction between inventorship and ownership in Canada’s patent system and the similar purpose of the Patent Act to Australia’s Patents Act may favour adopting an approach more in line with the Australian Federal Court. If Canada accepts that AI can be an inventor and welcomes such patents into our patent system, this could open the door to more innovation in Canada and create an opportunity for AI-related inventions to flourish. Whether such patent applications also adequately address requirements of novelty, inventiveness, utility, and sufficiency, and whether they are ultimately granted will be the next legal frontier.
As we have discussed in Part 1 of this blog post series, the Government of Canada is looking to adapt and modernize the Canadian copyright regime to match the ongoing developments in AI and the Internet of Things, with the hopes of supporting innovation and investment in AI (which we described here). It is possible similar initiatives may be launched for patents in the future.
Regardless of how the law might currently treat AI inventorship, the Australian Federal Court’s decision raises a key question for all other jurisdictions – should AI machines be permitted to be listed as inventors from a policy perspective? We will explore this in Part 3 of this blog post series.
The authors would like to thank articling student Marisa Kwan for her assistance in preparing this update.