Artificial Intelligence as an Inventor: A brief Exploration of South African Intellectual Property Law

1. Introduction

The emergence of Artificial Intelligence has sparked significant changes across various sectors, including in the field of patent law. The ability of AI systems to develop new creations poses unique and challenging questions to the conventional structures of IP law. This article explores the implications of AI inventorship within the specific context of South African IP law, shedding light on the legal ambiguities and offering recommendations for future policy adaptations.

2. Understanding Artificial Intelligence

We can define AI as a multitude of computational systems that can execute tasks traditionally requiring human intelligence. This includes learning from experiences, interpreting language, recognizing patterns, and making decisions based on the scenarios presented to them.

Through what is called creativity machines, an AI system can foster innovation and creation, effectively shattering the paradigm of inventions as exclusively human endeavours. Generative systems and deep learning technologies can produce unique results that might not otherwise be possible or even conceivable through human intellect alone.

For instance, in various sectors such as music, engineering, pharmaceuticals, and art, AI has shown its prowess to innovate in ways that were previously unimaginable. AI's ability to create complex compositions in music, solve intricate engineering problems, generate potential pharmaceutical compounds, and even produce original art pieces exhibits its expansive capabilities.

This emergence of AI has the potential to completely reconfigure our understanding of what invention and creativity mean. As we move forward in this new frontier of AI-driven innovation, it becomes increasingly crucial to navigate the legal, ethical, and social implications of this technological advancement with caution and thoughtful deliberation.

3. Intellectual Property Law in South Africa

South African IP law is governed by several legal frameworks, such as the Copyright Act and the Patents Act. Within the realm of patent law in general, an inventor is typically a person who contributes to the invention's conception. However, the arrival of AI as potential 'inventors' suggests this definition may be inadequate.

South African patent law refers to the inventor in the masculine form as a person[1]. Unfortunately, in terms of definitions, the law did not advance with the definition of the term Inventor, despite having a definition of Patentee. Thus, it becomes necessary to verify what the law stipulates implicitly. See article 27 of the patent law, which stipulates the following:[2]

“Article 27. Who may apply for a patent.

(1) An application for a patent in respect of an invention may be made by the inventor or by any other person acquiring from him the right to apply or by both such inventor and such other person.”

The regulation that supports the patent law also refers several times to the human nature of the inventor[3]:

Article 22 of the regulation. Application for the grant of patents

…

d) where the applicant has acquired a right to apply from the inventor, an assignment or other proof, to the satisfaction of the registrar, of the right of the applicant to apply.

This means that it would be necessary to present an assignment document between the inventor and the applicant. However, the structure of such a document is predicated on the basis of a human inventor, considering that a legal transaction is born out of the will of the parties involved.

Furthermore, the address of the inventor is often mentioned. An AI system does not have an address[4].

Numerous other legal forms, such as the power of attorney and assignment documents, among others, reflect the human nature of the invention. These forms, inherently designed for human involvement, underscore the tacit assumption within the law that the process of invention is a human activity. They inherently imply the ability to express and act upon the intent, something that is, currently, distinctly human and beyond the capabilities of an AI system.

By implicitly structuring legal documents and processes around human agents, the law seems to maintain that inventions, at their core, are human endeavours. Therefore, extending these legal structures to accommodate AI-generated inventions would not just be a matter of updating the paperwork. It would represent a fundamental shift in how we conceptualize the nature of the invention and the legal structures that protect it.

South African IP law currently does not explicitly account for AI inventorship, leading to a legal grey area that could potentially stifle innovation or enable exploitation. Hence, it becomes crucial to explore how these legal structures can be adapted to accommodate AI’s role in inventing, which may involve the creation of a sui generis right.

4. Case Studies: The DABUS Patent in South Africa

A milestone was reached in the field of patent law when an AI system, known as DABUS (Device for the Autonomous Bootstrapping of Unified Sentience) created by Dr. Stephen Thaler, was acknowledged as the inventor for a patent in South Africa. This decision represented a significant shift, as it marked the first instance of an AI system being granted patent rights.

However, it's essential to understand the context of patent processing in South Africa to grasp the full implications of this decision. Unlike some other patent offices, South Africa's patent office does not undertake substantive examination of patent applications. Instead, it focuses on ensuring that applications comply with the necessary formalities.

This essentially means that no substantive decision regarding the merit of the invention, or the applicability of an AI system as an inventor, was technically made within the scope of this process. The patent office didn't formally examine the legitimacy of the AI as an inventor but primarily concerned itself with the completeness and correctness of the application forms.

This raises intriguing questions about the legal formalities involved in patent applications, particularly those related to inventorship and ownership. For instance, the standard requirement of an assignment document - which transfers rights from the inventor to the applicant - presents a conundrum in this case. As AI lacks legal personality, it cannot enter into a legal transaction, raising questions about the formal validity of the application.

Moreover, there seems to be no evidence or explanation as to how Dr. Stephen Thaler, the creator of DABUS, acquired the rights to the AI-generated invention. Without an assignment or other proof of transfer of rights from DABUS to Dr. Thaler, this omission presents potential formal inconsistencies in the application process[5].

A court might examine the nature of the inventor, the absence of a clear human inventor, and the inability of the AI to legally transfer rights, all of which could lead to a ruling that the patent is invalid due to non-compliance with the formal requirements laid out by law.

DABUS was named, also, as the inventor of two patent applications in the UK, EU, and the US. These jurisdictions rejected the applications, asserting that an inventor must be a natural person.

Contrasting South Africa's decision with the UK, EU, and US responses to the DABUS applications highlights differing attitudes towards AI and IP law. While South Africa seems to embrace AI inventorship, other jurisdictions adhered to the conventional human-centric view of inventorship. This divergence underscores the lack of international consensus on AI's role in IP, adding to the complexity of multinational patent applications and enforcement.

The decision to grant the DABUS patent could be a significant development in South Africa's IP law. It could indicate a willingness to adapt to technological advancements and could potentially stimulate AI-related innovations in the country.

Nevertheless, we might argue that there has been a misinterpretation or misapplication of the formal requirements established for patent applications. If challenged in court, it's plausible that the patent could be invalidated due to failure to meet these legal formalities, despite the absence of an explicit human inventor.

Elaborating on this point, the uniqueness of this case could potentially lead to judicial scrutiny. A court might examine the nature of the inventor, the absence of a clear human inventor, and the inability of the AI to legally transfer rights, all of which could lead to a ruling that the patent is invalid due to non-compliance with the formal requirements laid out by law.[6]

5. Ethical and Legal Implications

Acknowledging AI as an inventor poses substantial ethical and legal dilemmas. A critical question is the assignment of moral and legal responsibility if an AI develops a harmful invention.

From the IP law perspective, the ownership of a patent becomes problematic. If AI is recognized as an inventor, who then owns the patent? Current law assigns patent rights to the human assignee, but this framework does not consider AI systems.

Moreover, allowing AI to patent its inventions could disrupt the patent system.

In considering innovation and the future, we should be cautious about accepting the protection of AI-generated inventions, as doing so could potentially undermine the very purpose of an IP system. The bedrock of IP systems worldwide is to foster and protect human creativity and invention, reflecting the toil and effort that a person puts into their creative or inventive process.

There is certainly room for argument that considerable intellectual effort goes into creating an AI system in the first place. Or, in the act of conceiving this AI system, the user may generate an innovative idea for a product or process and request the AI to bring this invention to life. However, it is crucial to make a clear distinction between human-aided inventions and those entirely generated by AI.

We're not discussing scenarios where AI merely supports or enhances the inventive process - rather, we're considering cases where the entire substance of a patent application, the complete technical solution, originates from an AI system, independent of direct human intervention.

Under these circumstances, the human user may not have been involved in the actual development of the AI system. Is it then justifiable to grant them commercial exclusivity based on an idea? It's vital to remember that ideas in themselves are not the subject of protection under IP systems. Rather, what is safeguarded is the tangible application of these ideas - the fruits of human intellectual labor. A blanket acceptance of AI-generated inventions, without critical scrutiny, could potentially subvert the foundational principles of the IP system.

Patents serve a specific purpose - they encourage human innovation by awarding temporary monopoly rights to inventors. If we allow AI systems to saturate the patent system with AI-generated inventions, this could inadvertently stifle human innovation. It's a delicate balancing act, one that could shape the future trajectory of innovation and the patent system itself. We must tread carefully, ensuring that our legal frameworks adapt and evolve to account for the rising tide of AI, while staying true to the original purpose of fostering human creativity.

6. Potential Solutions and Policy Recommendations

This case prompts a re-examination of South Africa's IP policy and legislation. Will existing laws be amended to explicitly accommodate AI inventors, or will a separate legal framework be established? Ensuring the ethical use of AI and navigating the potential monopolization of patents by AI will be central to this discourse.

From a comparative perspective, South Africa could potentially look to other jurisdictions that are struggling with similar issues. For example, in the US, there have been calls for legislative change to clarify the status of AI inventors, while in the EU, there is ongoing debate about whether the existing legal framework is sufficient to address the challenges posed by AI.

Furthermore, the global divergence on AI inventorship calls for international dialogue to harmonize laws and manage cross-border IP disputes. South Africa's pioneering decision could shape these discussions, serving as a potential model for other countries navigating the AI inventorship issue.

In light of these challenges, South African IP law could consider adaptations. One solution could be to define an inventor explicitly as a "natural person", thereby maintaining the status quo.

Alternatively, the law could allow the owner or the user of an AI system to have a sui generis right. This could promote innovation and the use of AI, but also would not potentially exacerbate patent trolling and market monopolization issues, which I believe we should be careful about and try other methods to promote innovation.

To tackle the ethical dilemmas, laws and regulations around AI ethics and accountability should be strengthened, ensuring a clear line of responsibility even when an AI system is involved in the invention process.

7. Conclusion

The role of AI as an inventor presents complex challenges to South African IP law, demanding substantial consideration. Legal ambiguities need to be clarified to unlock the benefits of AI-driven innovation while ensuring ethical use.

Given the rapid development and widespread application of AI technologies, this is not a hypothetical or futuristic issue but an immediate concern. South Africa, like other nations globally, faces the urgent task of revising their IP frameworks to accommodate this new era of AI systems.

This article was originally published in IPR Daily. See here the Chinese version.

[1] Thaldar D, Naidoo M. AI inventorship: The right decision? S Afr J Sci. 2021;117(11/12), Art. #12509.

[2] Patents Act 1978 (Act No. 57 of 1978, as amended up to Patents Amendment Act 2002)

[3] Patent Regulations, No. R6247 of December 15, 1978, as amended by Government Notice No. R1181 in Government Gazette No. 29413 of 1 December 2006

[4] Lavrichenko, Michelle, Thaler v. Vidal: Artificial Intelligence—Can the Invented Become the Inventor? (December 2022), p. 709

[5] South Africa was wrong to patent an AI’s ‘invention’ - David Cochrane and Christopher Mhangwane, 2022

[6] Oriakhogba, Desmond, Dabus Gains Territory in South Africa and Australia: Revisiting the AI-Inventorship Question (October 1, 2021). (2021) 9 South African Journal of Intellectual Property Law p.107.