INTRODUCTION
Artificial Intelligence (AI) has existed since the 20th century.1 However, generative AI, a relatively new advancement, has become the most prevalent concept in the field. Generative AI is a form of artificial intelligence algorithm that can be used to create new content, including audio, code, images, text, simulations, and videos.2 Generative AI models learn the patterns and structure of their input training data and then generate new data that has similar characteristics.3 The use of generative AI models such as Chat GPT, Bing, Bard, Copilot and MidJourney have raised several concerns from both legal and ethical points of view. More recently, concerns have been raised specifically around intellectual property (IP) rights and the potential infringements that may arise in the use generative AI.
In this article, we will consider the implications of generative AI on patents and examine:
- whether patent law protects AI-generated inventions; and
- whether AI can be an inventor for the purpose of a patent application.
PATENTS
A patent is a right granted for an invention that is new and useful.4 The owner of a patent has the right to prevent others from exploiting the invention within the period of the grant, which is usually twenty (20) years.5 The patent application requires a full disclosure of the invention, how it came about and how it works.6 This is to ensure that after the protection period, the public can gain full possession of the invention and use the technical details disclosed therein to improve upon the invention, to design around it, or to engage in other innovative activities.7
Does patent law protect AI-generated inventions?
Generally, countries may grant patents to inventions which meet the patentability requirements even where they are made with the assistance of AI tools.8
These requirements may differ marginally from country to country but to qualify for patent protection an invention must:
- be new (novelty);
- include an inventive step (non-obviousness); and
- be used in a known industry (utility).
Most countries are silent on whether inventions generated wholly by AI can be patented. This is mostly because there remains some uncertainty around, amongst other things, whether an AI-generated invention meets the requirement of non-obviousness. An invention is said to involve an inventive step under common law, if it is not obvious to a person skilled in the art, having regard to any matter which forms part of the state of the art.9 This requires an invention to be one that is not readily apparent to a person having ordinary skill (and imagination) in the relevant field.10 When assessing whether the inventive step requirement has been met, a patent office or court would look at the state of the art (generally anything in the public domain that relates to the invention) and determine whether the invention is one that can easily be deduced from it or involves some further inventive process to come to.11 Whereas the novelty requirement is a quantitative requirement- "is the invention new?", the inventive step requirement is a qualitative one- "is the invention obvious?".12 If it is deemed to be obvious, then no patent protection can be granted. The inventive step requirement is to ensure that patents are granted to meritorious inventions, as opposed to obvious extensions and modifications of prior art.13
AI-generated inventions create a unique complexity for the application of the test of non-obviousness. For instance, a determination of what forms part of the prior art and who (or what) should be considered as a reasonably skilled person in the field is not one that can be easily made for AI. Should a similar AI be considered as a "person with ordinary skill" for the purpose of deciding non-obviousness? Should the prior art include all output by AI models in the relevant field? Would any invention meet such a standard? Moreover, since AI models are trained on large data sets consisting of existing material, it may be difficult to conclude that the invention generated is not merely a recombination of existing ideas in different ways14 but one with an inventive step.
The drive for an extension of patent protection to AI-generated inventions is said to stem from the need to encourage innovation, which in of itself is the aim of patent law. It has been argued that there would be no point in patenting AI-generated inventions since this aim cannot be achieved.15 The counter argument has however been that although the prospect of holding a patent will not directly motivate an AI, it will encourage some of the people who develop, own, and use AI.16 Allowing patents on AI-generated works, therefore, will promote the development of inventive AI, which will ultimately result in more innovation for society.17 Thus, where an AI-generated invention somehow meets the patentability requirements, there is no reason it should not be protected under the patent law regime.18
Despite the policy considerations, patent law as it currently stands, would require several revisions and clarifications to be able to properly accommodate AI-generated inventions. In its current form it is unclear whether patent law can extend its protection to AI-generated works. An option would be to provide some judicial or legislative clarity on the non-obviousness requirement for inventions wholly generated by AI as a first step to accepting the patentability of such inventions. Beyond further guidance, it may be prudent to require full disclosure in patent applications, indicating the exact involvement of AI in the invention to prevent persons from taking advantage of the different standards for non-obviousness.19 This could serve to clearly delineate the lines for granting patent protection to AI inventions.
Can AI be an inventor?
More recently, there has been some judicial activity on whether AI can be an inventor for the purpose of a patent application. In what is described as the Thaler cases20, Dr Stephen Thaler attempted in several jurisdictions21 to register a patent which had an AI system, "DABUS" listed as the inventor. DABUS was said to have invented an improved food container and a flashing beacon to be used in emergency situations.22 Thaler's attempts were refuted in all but South Africa and Saudi Arabia. The courts of the other jurisdictions refused the patent on the ground that their patent laws did not envisage AI as an inventor, but rather a natural person.23 Additionally, all the courts were emphatic that the question of whether AI could be an inventor under their respective domestic law was one of statutory interpretation and thus refused to intervene to make an inclusion that did not exist under statute.
United States
The issue that was before the Court of Appeals for the Federal Circuit was whether an AI model could be listed as an inventor for the purpose of a patent application. The court succinctly stated that it was purely an issue of statutory interpretation and that in defining an inventor or joint inventors, the Patent Act uses the term "individual". The definition of "individual" 24 had been settled to mean a human being thus leading to the inevitable conclusion that for the purposes of patent applications, an inventor can only be a person and that nothing in the Patent Act showed an intention of Congress to deviate from the default meaning of individual.
In February 2024, the U.S. Patent and Trademark Office (USPTO or Office) published new guidance on the patent inventorship analysis for AI-assisted inventions.25 The guidance provides that "while AI-assisted inventions are not categorically unpatentable, the inventorship analysis should focus on human contributions, as patents function to incentivize and reward human ingenuity."26 Although not legally binding, the guidance offers inventors, patent law practitioners and other stakeholders a framework for anticipating how the USPTO will consider patent applications where AI played a role and for understanding their continued obligations before the Office.
United Kingdom
The United Kingdom Supreme Court in dismissing the appeal of Dr. Thaler in Thaler (Appellant) v Comptroller-General of Patents, Designs and Trademarks (Respondent)27 stated that the scope and meaning of the term "inventor" under the Patents Act,1977 did not extend to a machine such as DABUS. The Court took a textualist approach and held that, as a matter of statutory interpretation (including pursuant to sections 7 and 13 of the Patents Act), an "inventor" must be a natural person.28
The Supreme Court noted that it was not opining on the broader questions as to whether technical advances generated by AI should be patentable or whether the meaning of the term "inventor" should be expanded to include AI systems. The court stated that these were policy questions beyond the scope of Dr. Thaler's appeal and that needed to consider the balance between the awarding of a monopoly (under a patent grant) and the benefits to the public from disclosure of the invention.
Germany
The German Federal Patent Court held that the AI, DABUS could not be named as an inventor as inventors were limited to natural persons. However, it allowed Dr Thaler, being the one who prompted or caused DABUS to generate the invention, to be named as the inventor.29
India
The Controller General of Patents refused the application by Dr Thaler since DABUS is not recognized as a person. However, the Parliamentary Standing Committee under the Department of Commerce of India has recommended to the Rajya Sabha (Council of States) some revisions to current legislation to provide for the recognition of AI as an inventor.30
South Africa
The patent office of South Africa granted the patent filed by Dr Thaler which listed DABUS as the inventor. Although granted, it is argued that this grant was merely concerned with the completeness of the application and did not go into the actual merits. 31 Thus, this could not be relied on as a precedent that South African law recognizes AI as inventors.
Ghana
Although the question is yet to be considered here in Ghana, both judicially and administratively, the emerging jurisprudence worldwide, along with the tone of the Patents Act, 2003 (Act 657) (the Patents Act), suggests that Ghanaian law would also consider an inventor to be a natural person and thus exclude AI. Although the Patents Act, does not define the term inventor, the Act has several provisions in relation to inventors which imply that an inventor must be a human being.
Section 4 generally addresses the right to the patent and the right to be named inventor which in several ways imply that the Patents Act considers inventors as human. The word "persons" is used in relation to an inventor in sections 4(2) and 4(3). Section 4(2) states that where two or more persons make an invention jointly the right belongs to them. Section 4(3) also provides for ownership of the right where two or more persons make the invention independently of one another. It is worthy of note that although there is no definition of inventor provided, the Patent Act uses the word "persons" in these sections showing a clear intention and understanding that inventors for the purpose of patents in Ghana are natural persons. Additionally, section 4(7) allows an inventor to make a special written declaration signed by the inventor stating a desire not to be named as inventor. Typically, the making and signing of a declaration is an act which is the preserve of human beings. This further strengthens the presumption that the Patents Act intended for inventors to be natural persons.
Given the position in the various jurisdictions described above, it is quite clear then, and for good reason, that courts are reluctant to extend the interpretation of patent statutes to allow for AI inventors. The right to be named as inventor is a moral right32 and as such it should only be enjoyed by natural persons. In addition, patent law presumes inventors are natural persons and thus have the capacity to enjoy or assign their rights to the invention. Likewise, legislatures are taking a cautious approach to amending laws especially when there is no international consensus on the matter.
Who should be named as inventor of AI-generated inventions?
From the above discussion, the question that arises is: who should be named as the inventor on record of AI-generated works? Should the definition be expanded to include AI or kept limited to human beings? Some have argued that just as corporate bodies and ships are granted personhood under the law, same should be extended to inventing AI.33
Recognizing AI inventorship would incentivize the use and development of AI especially in the invention process as well as prevent people from misstating inventors on patent applications and therefore prevent non inventors from enjoying rights they are strictly not entitled to.34 It would also ensure inventions find their way into the public domain as companies will be more willing to seek patent protection for their AI-generated inventions rather than protect them as trade secrets.35 The recognition of AI inventors would achieve more economic efficiency as there would be limited/no need for human capital (with their attendant limitations).36 Finally, it would do away completely with the thorny issue of who should be named inventor.37
On the contrary, the fear that exists is that recognizing AI as patent inventor could lead to the creation of severe monopolies (in relation to IP rights) in favour of companies or individuals with the huge capital needed to undertake AI invention38 and the proliferation of patents such that the incentive to invent is lost.39 Since the law starts from a presumption that the inventor is the patent owner, recognizing AI inventors could cause problems with ownership transfers.40 Although this may be remedied by granting legal personhood (or a limited form of it) to AI, several attempts in Europe and other places have shown that the world is not yet ready for such drastic changes.41 Lastly, it is considered that recognizing AI inventors would discourage humans who do not have access to AI from inventing and thus defeat the purpose of patent law which is to promote innovation.42
Despite the myriad advantages, the disadvantages seem to weigh more heavily against AI. If AI cannot be an inventor, who should be named? In answering this question, the Australian Full Court stated that:
- the owner of the machine upon which the AI software run;
- the developer of the AI software;
- the owner of the copyright in its source code; or
- the person who inputs the data used by the AI to develop its output,
could all be recognized as inventors of AI-generated inventions:43
If or when AI-generated inventions are considered patentable, it would be prudent, for the purpose of clarity and certainty, to name the one who uses the AI tool as an extension of their mind to create the inventions as inventor.44 Such a person would have conceived of an invention in their mind and thus warrant protection45 as opposed to a person who owns or creates (but does not use) the AI. This would serve to incentivize human inventors and at the same time encourage the use of AI in creating inventions46, answering at least two of the major issues on attributing inventorship.
Conclusion
The discussion so far points out clearly that patent law at present may recognize AI-generated inventions as patentable and makes a case for why it should. Yet, most jurisdictions do not accommodate AI as an inventor and this author has argued that this position is the soundest to protect the integrity of the patent system. Until and unless patent law is amended to explicitly include inventions wholly generated by AI, a case-by-case approach concerning the patentability of AI-generated inventions is best. In addition, it has been suggested that the patentability requirements be modified to properly cater for such inventions especially where patent law is amended to expressly include AI-generated inventions. With respect to inventorship, patent law could be widened to recognize persons who use AI models to create inventions as inventors with a reasonable standard for demonstrating intellectual or creative input. This would serve the aim of patent law which is to promote innovation (by humans) as well as encourage the development of AI whose utility has been widely demonstrated across several fields of industry.
Footnotes
1 The first AI programs were written by Christopher Strachey and Dietrich Prinz in 1951.
2 https://www.mckinsey.com/featured-insights/mckinsey-explainers/what-is-generative-ai
3 Pasick, Adam (March 27, 2023). "Artificial Intelligence Glossary: Neural Networks and Other Terms Explained". The New York Times. ISSN 0362-4331. Retrieved March 12, 2024
4 https://www.business.qld.gov.au/running-business/risk/ip/types/patents/definition
6 Ibid.,
7 Seymore, Sean B.: Symposium: The Disclosure Function of the Patent System; Vanderbilt Law Review Volume 69, Issue 6 at pgs. 1455&1456
8 The United States Patent Office released guidelines on 13 February 2024 on patenting AI assisted inventions: https://www.govinfo.gov/content/pkg/FR-2024-02-13/pdf/2024-02623.pdf
9 Molnlycke AB v Proctor v Gamble Ltd (1994) RPC 49,112.
10 Pfizer Ltd.'s Patent (2001) FSR 16
11 Windsurfing International Inc v Tabur Marine (Great Britain) Ltd [1985] RPC 59
12 Molnlycke AB v Proctor v Gamble Ltd (1994) RPC 49,112.
13 A. Griffiths, "windsurfing and the inventive step' 1999. IPQ 160, 163-4
15 https://www.wipo.int/wipo_magazine/en/2019/06/article_0002.html
16 Ibid
17 Ibid
18 https://intapi.sciendo.com/pdf/10.2478/vjls-2023-0002
19 See Heon, Lexi: Artificially Obvious but Genuinely New: How Artificial Intelligence Alters the Patent Obviousness Analysis, Seton Hall law Review, 2022, Vol 53:359
20 Thaler v Commissioner of Patents [2022] FCAFC 62 {Australia}; Thaler v Vidal & Ors 43 F.4th 1207 (Fed. Cir. 2022) {United States}; Thaler v Commissioner of Patents [2023] NZHC 554 {New Zealand}; Thaler v Comptroller-General of Patents [2023] UKSC 49 {United Kingdom}; Thaler v Taiwan IP Office (TIPO), 110 Xing Zhuan Su 3 (August 2021). The decisions of the other jurisdictions were by their patent offices.
21 Australia, the United States, New Zealand, the United Kingdom, South Africa, Taiwan, Israel, the Republic of Korea, Canada, Brazil and India
22 Trevor F. Ward, DABUS, An Artificial Intelligence Machine, Invented Something New and Useful, but the USPTO is not Buying It, 75 Me. L. Rev. 71 (2023) at pg. 81
23 Person here varies between natural and legal persons. For instance, the United Kingdom, United States, Australian and New Zealand courts concluded on a natural person whereas the EU Patent Office concluded on a legal person.
24 Mohamad v. Palestinian Auth.566 U.S. 449, 454 (2012)
25 See "Inventorship Guidance for AI-Assisted Inventions," February 13, 2024; see also the USPTO's supporting PowerPoint presentation, March 5, 2024
26 Ibid.,
27 Judgment dated December 20, 2023
28 Ibid.,
29 https://academic.oup.com/grurint/article/71/12/1162/6821266
30 https://www.lexology.com/library/detail.aspx?g=44cb22b6-5842-496a-8555-5efc81995bdc
32 https://www.wipo.int/edocs/mdocs/scp/en/scp_35/scp_35_7_summ.pdf
34 See Cole G. Merritt, A Compulsory Solution to the Machine Problem: Recognizing Artificial Intelligence as Inventors in Patent Law, 25 Vanderbilt Journal of Entertainment and Technology Law 211 (2023)
35 Ibid.,
36 See Fok, Ernest, Challenging the International Trend: The Case for Artificial Intelligence Inventorship in The United States, 19 Santa Clara J. Int'l L. 51 (2021)
37 Ibid.,
38 Ibid, note xxvii
39 Ibid, note xxix
40 Ibid.,
41 Ibid.,
44 https://www.brookings.edu/articles/ai-inventions-policy-options-and-a-path-forward/
45 See Villasenor, John, Reconceptualizing Conception: Making Room for Artificial Intelligence Inventions, 39 Santa Clara High Tech. L.J. 197 (2023).
46 Ibid, note xxxix
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.
