As technology surrounding artificial intelligence ("AI") continues to improve, so does its capacity for increasingly complex tasks. In fact, some AIs have become so advanced that they now have the capacity to conceptualize and create inventions.
In Canada, inventions that are new, useful and non-obvious can be afforded patent protection. Under the patent system, in exchange for inventors disclosing their inventions to the public, they are afforded a 20-year monopoly from the time of filing, during which time they have the exclusive right to practice the invention. In the normal course, the inventor who conceives of the invention has the right to apply for and be a named inventor of the patent. However, when the invention is conceived of by AI, it challenges this system, and begs the interesting question: can an AI qualify as an inventor of a patent?
Various jurisdictions around the world have recently had to tackle this question in response to patent applications filed by Mr. Stephen Thaler. Thaler is the creator of a system called DABUS, short for "Device for Autonomous Bootstrapping of Unified Sentience". DABUS is described as a "creativity machine", an AI system that uses a series of neural networks to "think" and create novel inventions. It has so far conceived of two inventions: a "fractal container", designed to allow for coupling of multiple containers, improve grip, and improve heat transfer; and a "neural flame", a flashing light designed to emit a uniquely-identifiable light signal that attracts enhanced human attention. Thaler filed patent applications for the inventions in 17 different countries.1 On each application, DABUS, rather than Thaler, is listed as the inventor. Because patent law is statutory, and each country has distinct patent laws, various jurisdictions around the world have had to individually address the question of whether AI can legally be considered an inventor of a patent.
Australia, the US, the UK, and the EU have all addressed this issue and rejected the DABUS patent applications. Australia is the most recent country to comment on the matter. In April 2022, a 5-member panel of the Full Court of Australia reversed a decision from Australia's Federal Court, which had initially held that AI could legally be an inventor. On appeal, the Full Court found that an "inventor" within the meaning of the legislation must refer to a human or natural person.2
The Full Court's decision canvassed Australian jurisprudence and historical legislation linking patent protection to an actual act of invention by the "true and first inventor". The Full Court then the cited to section 15(1) of the Australian Patents Act, allowing patents to be granted to "a person who is an inventor", and held that "person" means a natural person. Further, the court referenced provisions of the Australian Patents Act that allow for inventors to assign of entitlement to the patent to others. These provisions require there to be a legal relationship between the actual inventor and the person first entitled to the grant. The Full Court found the actual inventor must be a natural person, since something without a legal identity, such as AI, could not legally give effect to such a grant of assignment. The Full Court also relied on precedent holding inventions to be the result of "human action" and Australia's legislative history which supported the finding that "the origin of entitlement to the grant of a patent lies in human endeavor".
The Australian Full Court's decision is consistent with other jurisdictions that have declined to allow AI to be an inventor under patent law, and that have declined the DABUS applications. In September 2021, the US District Court upheld the decision by the US Patent Office ("USPTO") in respect of the DABUS patents, finding only natural persons, and not AI, can be inventors.3 The Court construed the reference to "individuals" in the definition of "invention" in the US Patent Act as necessarily referring to a natural person. The District Court's decision is currently on appeal.
The EU and UK have reached similar decisions on the DABUS patents. In September 2021, the England and Wales Court of Appeal (EWCA) rejected the DABUS patent applications and found that AI cannot be an inventor.4 Examining various provisions of the UK Patents Act 1977, the Court found that the statute's reference to the "inventor", who is required to be the "actual deviser" of the invention, must consist of a "person or persons". The European Patent Office (EPO) Legal Board of Appeal similarly released a press communiqué in December 2021, announcing that it had upheld decisions refusing to grant the DABUS patent applications.5 Although the detailed reasons have not yet been released, the Board of Appeal refused to allow the applications because "[u]nder the EPC, the inventor had to be a person with legal capacity." A recent decision from the IP Office of New Zealand similarly denied the DABUS application, holding that New Zealand patent legislation's reference to "inventor" refers only to natural persons and does not contemplate non-human inventors.6
Thaler has seen success in one jurisdiction: South Africa, where the patent, listing DABUS as the inventor, was granted in July 2021. However, South Africa does not have formal examination requirements for patent applications filed in the jurisdiction, so neither their Patent Office nor their Courts have had to substantively grapple with the legality of AI being an inventor.
The Canadian Perspective
Canadian courts have not yet had to address the question of whether AI can be an inventor in the context of patent law. However, since DABUS-generated inventions have been filed in Canada, a decision on this issue is likely on the horizon.
The Patent Act does not define "inventor", and there has been limited judicial commentary on its meaning. In considering a pharmaceutical patent, the Supreme Court (SCC) inferred an inventor is the "person or persons who conceived of" the patented invention.7 This holding has been interpreted by the Federal Court as: "[i]n Canada, the language of the jurisprudence assumes that an "inventor" is a natural person".8 While these decisions did not consider the question of whether AI can be an inventor, their reasoning may be relied upon to resist AI-patent applications, as seen already in Australia, US, and Europe.
The Canadian Patent Office's (CIPO) recent commentary on Thaler's application supports this hypothesis. In December of 2021, CIPO issued a compliance notice9 finding Thaler's application contravenes section 27(2) of the Patent Act and section 54 of the Patent Rules,10 which require that an application for a patent be filed by "an inventor or the inventor's legal representative", that the inventor be identified, and that a statement of entitlement to the patent be filed by the applicant. CIPO found the application non-compliant because the inventor for this application is a machine, "and it does not appear possible for a machine to have rights under Canadian law or transfer those rights to a human." Interestingly, CIPO did not completely shut the door. CIPO stated that Thaler could remedy his application "by submitting a statement on behalf of the Artificial Intelligence (AI) machine and identify, in said statement, himself as the legal representative of the machine." CIPO's statement suggests Thaler may be eligible to receive patent protection for DABUS-generated inventions, provided DABUS is not listed as the sole inventor without more. Instead, Thaler could presumably rely on the language in section 27(1) of the Patent Act, which allows patents to be granted for an invention "to the inventor's legal representative".
Other Outstanding Questions on AI and Patent Law
The DABUS applications raise other important legal questions on AI and patent law:
- Inventorship: Should DABUS be considered the inventor, or should Thaler's work in creating DABUS entitle him to be considered the inventor for patent law purposes? This is the question that the Australian Full Court left open in its recent decision, noting that "the characterization of a person as an inventor is a question of law. The question of whether the application the subject of this appeal has a human inventor has not been explored in this litigation and remains undecided."11
- Ownership: Even if DABUS is considered the inventor, who should be the individual entitled to apply for or own a patent for an invention generated by an AI? This is yet another question that the Australian Full Court left open. Both the EWCA and EPO held that Thaler's ownership of DABUS, on its own, did not provide him with the right to apply for patents for inventions generated by DABUS for which DABUS is listed as inventor. In Canada, similar issues are currently being examined by the Canadian government regarding copyright ownership for AI-generated works.12 Although a final decision as to how ownership and authorship of the copyright has yet to be made, the outcome may influence how the issue is treated in the patent law context.
- Patentability Standards: There is uncertainty surrounding requirements for patentability for AI-generated inventions. For instance, in order for an invention to receive patent protection, it must be non-obvious. Should the obviousness standard be the same for AI-generated inventions, in comparison to those conceived of by humans? Further, patents must disclose the invention in a manner sufficient to enable a person skilled in the art of the patent to practice the invention. How might this standard need to change for an invention generated by AI using specific algorithms or training data? These and further questions are currently being discussed by stakeholders at the World IP Organization (WIPO),13 but have not yet been considered by the courts.
With the growing contribution of AI to innovation, it is inevitable that the Canadian government or judiciary will tackle these important questions. A 2019 study from Innovation, Science and Economic Development Canada (ISED) tracked an increase in the patenting of inventions involving AI technologies.14 As the capabilities of AI technologies improve, we will likely begin to see more AI-generated inventions being developed in Canada and elsewhere.
Whether these inventions are entitled to patent protection, and who might enjoy the associated rights that come with such protection, may profoundly impact various innovators' IP strategies going forward. The above introduction refers to the 20 year monopoly that strikes at the heart of patent bargain, serving to coax new inventions into the public domain. In the absence of this term of protection, innovators relying on AI technology may lack incentive to develop novel inventions. Alternatively, without recourse to protect their inventions under the patent system, developers may be reluctant to share their AI-generated inventions with the public, choosing instead to protect their works as trade secrets. It is evident that we are at a crucial stage in tackling these important issues, with the evolving public policy and legal commentary shaping the future of AI inventions.
1. For a full list of Thaler's patent applications listing DABUS as inventor, see The Artificial Inventor Project, "Patents and Patent Applications", online: https://artificialinventor.com/patent-applications/
2. Commissioner of Patents v Thaler,  FCAFC 62.
3. Thaler v. Hirshfeld, No. 1:20-cv-903 (LMB/TCB) (E.D. Va. Sep. 2, 2021).
4. Thaler v Comptroller General of Patents Trade Marks And Designs,  EWCA Civ 1374.
5. European Patent Office Press Communiqué on decisions J 8/20 and J 9/20 of the Legal Board of Appeal dated December 21, 2021, online: https://www.epo.org/law-practice/case-law-appeals/communications/2021/20211221.html
6. Stephen L. Thaler,  NZIPOPAT 2 at para 32.
7. Apotex Inc. v. Wellcome Foundation Ltd., 2002 SCC 77 at para 96.
8. Sarnoff Corp v Canada (Attorney General), 2008 FC 712 at para 9.
9. See Canadian Patent Application 3,137161 Office Letter dated November 8, 2021, online: https://www.ic.gc.ca/opic-cipo/cpd/eng/patent/3137161/summary.html?type=number_search&tabs1Index=tabs1_1.
10. Patent Act, R.S.C., 1985, c. P-4, s. 27(2); Patent Rules, SOR/2019-251, s. 54.
11.  FCAFC 62, supra note 2 at para 121.
12. Innovation, Science and Economic Development Canada, "A Consultation on a Modern Copyright Framework for Artificial Intelligence and the Internet of Things" (2021) at 11-16, online (pdf): https://www.ic.gc.ca/eic/site/693.nsf/vwapj/ConsultationPaperAIEN.pdf/$file/ConsultationPaperAIEN.pdf. Canada is considering three options in respect of copyright: (1) authorship is attributed to the human who made arrangements for the creation of the work; (2) AI-generated works are ineligible for copyright, since they do not involve human participation in creation of the work; and (3) a new standard would be created, where the work itself is "authorless", but ownership of the work transfers to the person who exercised skill and judgment in creating the work using the AI.
13. World IP Organization, "WIPO Conversation on Intellectual Property (IP) and Artificial Intelligence (AI), Summary of Second and Third Sessions" (4 November 2020) at 4-6, online (pdf): https://www.wipo.int/edocs/mdocs/mdocs/en/wipo_ip_ai_3_ge_20/wipo_ip_ai_3_ge_20_inf_5.pdf.
14. Innovation, Science and Economic Development Canada, "Processing Artificial Intelligence: Highlighting the Canadian Patent Landscape" (2019), online (pdf): https://www.ic.gc.ca/eic/site/cipointernet-internetopic.nsf/vwapj/AI_Report_ENG.pdf/$FILE/AI_Report_ENG.pdf.
To view the original article click here
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.