Harmonizing AI and IPRs

Otim Enoch

Assistant Lecturer, Department of Public and Comparative Law, Faculty of Law, Victoria University, Kampala, Uganda. Mobile line +256703977097. Email address: [email protected]

Abstract

Harmonizing Artificial Intelligence (AI) and Intellectual Property Rights involves navigating the intersection of technology and legal framework to address the unique challenges posed by AI innovations within the realms of intellectual property. This research explores the considerations in harmonizing AI and IPRs, focusing on issues such as ownership of AI-generated works, patentability of AI inventions, protection of AI algorithms, and the balance between fostering innovation and safeguarding intellectual property rights. By delving into these complexities this paper sheds light on the evolving landscape where AI and IPRs converge highlighting the need for adaptive legal frameworks to ensure a harmonious coexistence between technological advancement and intellectual property.

1. Introduction

The effects of AI are felt in various industries as it has the potential to introduce significant transformation to the way tasks are accomplished. With regards to intellectual property law, AI can automate a significant portion of legal work, significantly reducing costs to clients and increasing the efficiency of law firms (Jarrett and Choo2021). AI also has implications in the arts by producing potentially copyrightable works. In cases involving patent disputes, AI has the potential to accurately conduct prior art searches with greater efficiency than humans. This increasing use of AI in these fields of intellectual property law gives rise to concerns over the protection of AI products and liability issues. This creates a requirement for harmony between AI and intellectual property laws.

AI has witnessed a mathematical rise in recent years with a noteworthy emphasis on machine learning (ML). The field itself is interdisciplinary, covering several areas of computer science, psychology, and neurophysiology. Despite this, it can be generally characterized as the branch of computer science involving the development of software and hardware that is capable of intelligent behavior.

1.1. Definition of AI and IPRs

AI has many definitions which can vary depending on the context in which it is used. The term AI was first coined by John McCarthy back in 1956 as the science and engineering of making intelligent machines (Anderson, 2024). This would involve the replication of human intelligence and processes into machines so that they can be capable of mimicking human thought processes to accomplish tasks. This task would involve simulation of human intelligence and learning, which is a bit abstract in its current sense still today. In more modern contexts, AI is often used as an umbrella term to collectively mean any act of getting a computer to perform tasks that would otherwise require human intelligence. What is core to all types of AI is that there is an intent to simulate human thought and intelligence, whether that be via machine learning, state-based systems, or logical-based processes. Any simulation of human thought and intelligence will create an overlap with current IP protection laws, and they will be hard to enforce as AI has blurred the lines between man and machine creations.

We are coming up with action plans. However, it is undeniable that AI is beginning to take a far larger role in our lives, and certain legislative measures are becoming more necessary. If AI is the act of implementing intelligence into machines, then we must select specific types of AI to focus on intellectual property rights. The higher the level of intelligence in a machine, then the higher the possibility of human-like thought processes being simulated. When we reach the stage of simulating human-like thought processes, then it is reasonable to apply the same legislative framework for IP protection of said thought processes, as we do for humans today (Vescovo, 2023). However, this is not the issue that is faced today, and to better understand the implications AI has on IP legislation, we must first define what is meant by AI and likewise IP rights.

1.2. Importance of harmonization

The importance of harmonization of AI and IPRs can be established through Article 1(3) of the Agreement on Trade-Related Aspects of Intellectual Property Rights (TRIPs), which describes the necessity to "reduce distortions and impediments to international trade." This statement focuses on a crucial aim of international law: to create a uniform set of rules that multinational corporations can follow in the international business environment, the foremost area where AI and IPR harmonization can be found. J Sbardellati discusses AI as being "integrated into one or more modules of a product, or an aspect of technological innovation. AI is likely to be a crucial competitive tool and the object of intellectual property rights" (Sbardellati, 1999). This statement shows the potential for AI to be a trade-distorting or enhancing tool by suggesting that the implementation of AI into a product or process will provide a company with a form of technology that can vary from others in its productivity and output. Suppose the AI technology created by a firm is key to a product's success. In that case, it will want to ensure that it can protect that technology from being imitated or duplicated, and it will want to use that AI product to enhance international trade in that particular market or to venture out into new markets with that AI tool. The objective for these companies is to be able to have a predictable set of IPR laws they can follow to protect this type of technology and to ensure that the IPRs for that technology will enhance its trade within markets and globally. The foundation for this harmonization process can be found in TRIPs itself, as Article 1(2) states, "Members may, but shall not be required to, implement in their law more extensive protection than is required by this Agreement, provided that such protection does not contravene the provisions of this Agreement." The flexibility allowed in TRIPs for its member states gives the initiative for AI technology-based companies to coordinate and lobby at both a national and international level to influence the creation or amendment of IPR laws that will suit their particular type of technology. This coordination will ensure that a particular set of laws favorable to protecting AI technology will be created and will also reduce the chances of a discrepancy between different states' IPR laws regarding AI technology, which could cause trade distortions between similar products in different markets (Athreye et al.2020).

2. Challenges in Harmonizing AI and IPRs

2.1 Lack of clarity in AI-generated works Defining an AI-generated work can prove to be a difficult task. Currently, it may be a mixture of human and AI input. However, the goal for the future of AI technologies would be moving towards AI software that requires no human input to create output or unsupervised AI. At this point, the machine-generated work would have no author behind it. This would leave the issue of whether the work would be attributed as having been created by the person who activated the AI or the owner of the machine. If the machine is repaired, the work belongs to its owner. It is suggested that the work would have the same status as the work of a person in employment, in that it is work done in the course of employment and the employer will be considered the author. However, this concept and the actual status of employment can vary from country to country and would not apply to work done by a self-employed machine owner (Hoffmann et al., 2020).

AI technologies are increasing in their capabilities at an exponential rate. They are now able to produce outputs that closely resemble those of human origin, spanning from works of art to literature and even scientific discoveries. The traditional understandings of intellectual property rights (IPRs) may have an issue applying to such output. Generally speaking, for any work to be attributed as owned and created by an author, there must be a certain level of intellect and human intervention involved. AI-generated works, however, are often created without direct human involvement (Salami, 2021). Additionally, for works to be copyright protected, they must have been the product of a national or domiciled person or person of a country that is a signatory of an international copyright treaty. This adds another level of complexity, as borders do not bind AI technologies and can produce work at the behest of any programmer from any country.

2.1. Lack of clarity in AI-generated works

It is quite challenging because the existing IPR law in every country is still based on the assumption of copyright IPR for humans and legal entities. If it is classified that the AI system is the primary author, it can lead to the revision of IPR rules and contribute new provisions on copyright IPR for AI systems as legal entities. But it can also lead to technological unemployment issues because the result of the work is not owned by the human and it can reduce the value of the work itself (Hong et al., 2022). The other problem if the human is still categorized as the primary author is that the AI system only acts as a tool. There are no special provisions regarding AI as a tool to make the work because the existing law is more focused on the provisions of subordinates and employment.

AI is a sophisticated system that can learn and innovate by itself. The current AI system has been built to create its work, for example, music, artwork, and even literature. This raises a question of whether the AI system itself can be categorized as the primary author and claim the IPR, or if the human using the AI system is the one responsible. If the AI is categorized as the primary author, the IPR of AI-generated work will last until 50 years after the work is published. If the human is categorized as the primary author, then the IPR depends on the human itself (Smits and Borghuis2022).

2.2. Ownership and authorship attribution

This definition is not wholly satisfactory when applying to the ownership of AI-generated works, since current legislative provisions do not cover the situation where the creator is a machine. The Copyright Act dictates that the first owner of a copyrighted work is the author, defined as "the person who creates the work". This definition is not satisfactory since it is not uncommon for a person to create a work in the scope of their employment or against a commission. In such cases, the first owner of the copyright is the employer or the person who commissioned the work (Salvagno et al., 2023). This provision would need to be interpreted regarding the ownership of AI-generated works, and utilizing the idea of ascribing authorship to the person who can be deemed to have directed the creation of the work would prevent the anomalous situation where robots could be the legal authors of copyright works.

When dealing with AI and IPRs, a differentiation between ownership and authorship is essential. Ownership is a matter of possession of rights. When a computer is creating works without human intervention, the question of who owns the computer arises. In many cases, the owner of the AI will be the owner of the work as per the agreement between the parties (Brown, 2021). Where the AI has created works without direct instructions or oversight from the owner, the position is less clear. In this case, it would be prudent to revert to the default position that the work is owned by the person who can be deemed to have created it.

2.3. Copyright protection for AI-generated content

In providing copyright protection for AI-generated works, the requirement that the work is original and reflects the creator's skill, judgment, and labor presents difficulty. In the case of computer-generated works, the author of the work will be the person who causes the work to be created, that is, the person who inputs the data to produce the work. However, this will not apply if it is proven that the work's creation did not involve human intervention (human authorship). In LADBROKE, the Court of Appeal accepted that in some cases computer-generated work is capable of being original and held that in these circumstances authorship would reside with the person who is responsible for deciding the criteria by which the work is produced. But it is still unclear how this decision applies. Working with the current legislation, it may be difficult to prove that the work reflects the skill and judgment of the human author. This may lead to the work being deprived of copyright (Wan & Lu, 2021). To combat this problem, it has been suggested that the Copyright Designs and Patents Act 1988 be amended to state that the author of a computer-generated work should be taken to be the person who makes arrangements for the creation of the work. This amendment would require the UK to adhere to the Berne Convention, which sets out that for literary and artistic work, the author shall be the national of the country to which the Convention is in force. However, to prevent a plurality of authorship, the Act should continue to provide that the copyright will subsist until 70 years after the death of the person with deemed authorship (Paul, 2021).

3. Legal Framework for Harmonization

The international treaties concern potentially conflicting obligations under the WTO TRIPS Agreement and other agreements led to the establishment of two organizations that are active in the field of patenting and human health, and that have a bearing on biotechnology: the World Health Organization (WHO) and the WTO. The common objective of both organizations is to promote public health, but they have different roles and constituencies. In 2001, the Doha Declaration on the TRIPS Agreement and Public Health at the WTO Ministerial Conference in Doha established that the TRIPS Agreement does not and should not prevent members from taking measures to protect public health. It confirmed that the Agreement can and should be interpreted and implemented in a manner supportive of WTO members' right to protect public health, and in particular, to promote access to medicines for all (Haugen, 2021). The Declaration was designed to provide increased clarity on the extent of LDCs and countries with insufficient manufacturing capacity to utilize the flexibility already inherent in the TRIPS Agreement for securing access to medicines. The Declaration led to the amendment of the TRIPS Agreement to enable countries to export pharmaceuticals made under compulsory licenses to countries that lack sufficient manufacturing capacity to produce the product in question. A decision of the WTO General Council in 2003 implemented this amendment to the TRIPS Agreement (Urias & Ramani, 2020).

3.1. International treaties and agreements

At an international level, the Agreement on Trade-Related Aspects of Intellectual Property Rights (TRIPS) is in place to create a minimum standard of IP protection across member countries. Although this agreement is not AI-specific, the range of technology it covers is so wide that it shall inevitably apply to AI. TRIPS defines technology through different sectors such as telecommunications or computer software. Each of these areas has its specific patent protection (Reyad). Therefore, patents for AI software would always be covered by the general TRIPS definition of patents, and the guide to enforcement of such patents provided by TRIPS would also apply to AI. However, as AI is such a constantly progressing technology, patent laws must be continually updated to reflect the new capabilities of AI.

The legal framework for the harmonization of AI and intellectual property rights is well-recognized internationally. To a large extent, the rights in rem (patents) and the rights in personam (usually copyright) in AI technology are similar to the traditional rights for other forms of technology (Picht and Thouvenin2023). Patents can provide a strong means of protection for AI technology as they can cover anything from a product or process feature to a concept or method. Copyright may protect software - an AI tool is protected as a work that derives literary or artistic content, providing it is original, compared to other software. However, the use of AI may also affect patent and copyright laws and the way they operate.

3.2. National legislation and case law

National laws shaping legal protection of AI This is perhaps the area where the synergy of AI and IPRs has primarily come into focus as it is here where the various international norms on protection are concretized and applied. However, to date, the specific application of national legislation has been minimal. Given the global nature of AI research and development and the international nature of AI products, this is unlikely to change unless specific areas of AI are targeted for protection. National legal systems are largely based on the international norms on IPRs as set out in the TRIPS and other international IP agreements. However, IPR law has developed in different ways in different countries and the level of protection and specific protection methods can vary widely between nations. The main areas of IPR law relevant to AI are copyright, patents, and trade secrets. Copyright has long been regarded as an area of IPR law problematic for AI and, in general, does not give the extent of protection hoped for by an AI developer (Rodrigues, 2020). This is particularly a problem for machine learning systems which may produce work akin to that of a human. Patents have been the preferred method for protecting AI; however, there are many areas where AI simply would not be able to be patented mainly due to constraints on what can be patented included in Article 52 of the European patent convention. Trade secrets are often seen as one of the best forms of protection for AI as it does not have any formal requirements and may be obtained through practices such as keeping the algorithm secret or simply not releasing the AI into the public domain.

3.3. Best practices and guidelines

There are several best practices and guidelines available for producers and consumers of AI software, some of which come from sectors that have long experience in developing data processing and analytics technologies. It is a truism that the optimal time to build in IPR concerns is at the very commencement of product development. This is when the 'negotiation space' between the different stakeholders (e.g. coders, methods research experts, business strategists, and legal/compliance specialists) is at its most flexible, making it much more likely that an acceptable solution can be found. These solutions can cover a multitude of issues, from the specification of coding standards aimed at facilitating the later development of 'clean room implementations' of AI algorithms to the specification of employment agreements and job descriptions for data analytic professionals that allocate IPR ownership clearly and consistently (Lepori & Montauti). Since it allows a cheaper and less risky settlement of disputes than formal litigation, ADR should be considered as an alternative to court action in both contractual and non-contractual IPR matters relating to AI software. Depending on the nature and severity of the dispute, a wide range of ADR options may be open to the parties, from the use of a neutral expert to provide a binding opinion on a particular technical point, to full-blown arbitration.

4. Future Perspectives and Recommendations

We need to keep in mind that the laws and policies for intellectual property were built around human beings, and the sole reason these laws exist is to give people an incentive to create. With an AI creating a work, there will be issues of determining ownership of the work. If a computer program is designed or an AI is taught to create a work to which an IPR would normally be recognized, the ownership of this work is not so clear. This could be resolved by declaring that the person or legal entity at fault for the AI is the owner of the work, but fraught litigation may follow in determining what exactly the faulty action was. Furthermore, works created by an AI may not serve its intended purpose of increasing the incentive to create a work (Brown, 2021). An AI that generates a musical tune, for example, will be much more efficient in generating new melodies if it is to escape any potential liability of infringing on someone else's copyright of an existing melody. This is an issue because people have a copyright to a melody for their lifetime + 50 years, and often a melody is forgotten or the copyright holder is unknown. Any work involving translating, editing, or improving an existing work runs the risk of potential liability, rendering it not worth the investment. Lastly, a single work by an AI may infringe multiple copyrights from different people or entities. The more complex the work involved, the more likely it will parallel certain features of more than one pre-existing work similar in nature (Dornis, 2020).

Artificial intelligence is a technology that learns and adapts to its environment, performing tasks and improving its performance on them over time. In the future, AI will likely play a major role in the creation, modification, and use of intellectual property, affecting all IPR categories. The potential efficiency increases from the implementation of AI in the process of creating intellectual property are massive, however, many potential issues could arise. If we do not act now to anticipate and mitigate these issues, we may regret it in hindsight (Damioli et al., 2021).

4.1. Ethical Considerations in AI and IPRs

In the context of AI and IPRs, it is of crucial importance to develop a set of ethical principles to ensure that global society can reap the benefits that AI promises while protecting individuals from its potentially harmful effects. It is beyond the scope of the present study to develop a comprehensive set of ethical guidelines, however, the following is a preliminary set of considerations that should underpin the development of AI-related to IPRs. Data generated by AI applications should be the copyright of the originating individual or organization. AI applications should respect the Databases Directive, meaning that there should be no extraction or re-utilization of data where this would be an unfair exploitation of the contents of a database, concerning the origin of data, and that there should be no impairment to the owner's right in the case where they have made the database contents public. AI applications should not infringe any copyright, patent, or design that is in place (Stahl, 2021). AI should not be designed to bring about new ways of circumventing IPRs, nor to produce adversarial attacks on IPR enforcement systems. If AI technology is being used to assist the creation or manufacture of a particular product, it should be designed to ensure that this product does not infringe the IPRs of another party. The highest risk area of AI and IPRs is likely to be concerning machine learning, where the learning algorithms are so complex that their results are not predictable by humans and it is therefore difficult to control whether the output is an infringement of someone else's IPRs. In such cases, it should be made clear who is the responsible party if there is an infringement, and if it is the machine learning agent itself, there should be some form of strict liability on the developer (Cao, 2021).

4.2. Collaboration between stakeholders

In the age of artificial intelligence (AI), companies can make strategic decisions and create innovative products at a faster rate than what can be regulated by the law. This is also true for companies in the field of AI, creating software and tools that implement AI in functions in which the companies may not fully understand or appreciate the implications of the technology. This results in a misalignment between law and technology, where the law is backward-looking and struggles to keep up with the pace of technological innovation. It is very difficult to predict how technology will develop and where it will be applied, making it difficult to create policy regarding something that is not well understood. When regulation is implemented, it may hinder future technological development because it is designed around the known uses of the technology (Indrawati & Kuncoro, 2021). A common strategy for innovative companies is to under-regulate their industry to prevent market failure and move to self-regulation once the industry has matured. This is typically not in the interest of the general public and leads to damaging consequences. AI technology is also commonly implemented into products and services of other industries. In the current environment, likely, the developers of the AI technology will not fully understand the IP rights of other products, possibly resulting in IP infringement. This will cause unwarranted hostility between the AI industry and other industries in which there will be calls for more stringent regulation on AI technology and increased protection of the IP rights of other industries.

4.3. Policy recommendations for effective harmonization

To achieve a globally harmonized system, two key approaches can be suggested. The first is to work towards achieving a higher level of international convergence. The second is to establish an international framework that takes into account local variations and implementation mechanisms. The former approach should be pursued for the direct protection of AI innovation. National disparities can hinder the free flow of AI technologies and know-how (Li et al., 2023). This approach would involve continuing efforts of the WTO TRIPS Agreement to establish a minimum harmonization agreement. Consensus on the required form of AI protection and permissible exceptions should be reached. It may be more flexible and practical in the field of AI innovation to create an agreement of contextual principles rather than achieve uniformity through binding rules. A project and approach of the World Intellectual Property Organization, the Singapore Treaty of 2004, aimed at coordinating and simplifying various patent procedures and achieving a higher level of general harmonization agreement between member nations. This agreement was stalled and eventually suspended in 2006 due to a lack of world consensus. The new leadership at the WIPO has shifted the focus away from AI and IP towards development matters. However, the recent increased attention AI has received across technology and scientific sectors, as well as its potential to significantly affect economic development in both developed and developing countries, means that now may be an opportune time to revive and reconsider such a treaty (Picht and Thouvenin2023). Those involved in AI innovation will also need to be hopeful of avoiding a repeat of the fate of the ill-fated European Community Patent, which became a costly and bureaucratic mess that only benefited lawyers. The latter approach, an international framework that takes into account local implementation, would be preferable in the field of AI IPRs. AI is still an emerging technology that spans diverse scientific markets and legal systems.

References:

Jarrett, A., & Choo, K. K. R. (2021). The impact of automation and artificial intelligence on digital forensics. Wiley Interdisciplinary Reviews: Forensic Science, 3(6), e1418.

Kuziemski, M. & Misuraca, G. (2020). AI governance in the public sector: Three tales from the frontiers of automated decision-making in democratic settings. Telecommunications policy.

Anderson, M. M. (2024). AI as Philosophical Ideology: A Critical Look Back at John McCarthy's Program. Philosophy & Technology.

Vescovo, S. (2023). Rise of the Machines: The Future of Intellectual Property Rights in the Age of Artificial Intelligence. Brook. L. Rev.

Athreye, S., Piscitello, L., & Shadlen, K. C. (2020). Twenty-five years since TRIPS: Patent policy and international business. Journal of International Business Policy, 3, 315-328.

Hoffmann, N., Stahlbock, R., & Vo?, S. (2020). A decision model on the repair and maintenance of shipping containers. Journal of Shipping and Trade.

Salami, E. (2021). AI-generated works and copyright law: towards a union of strange bedfellows. Journal of Intellectual Property Law & Practice.

Hong, J., Edler, J., & Massini, S. (2022). Evolution of the Chinese intellectual property rights system: IPR law revisions and enforcement. Management and Organization Review.

Smits, J., & Borghuis, T. (2022). Generative AI and intellectual property rights. In Law and artificial intelligence: regulating AI and applying AI in legal practice (pp. 323-344). The Hague: TMC Asser Press.

Salvagno, M., Taccone, F. S., & Gerli, A. G. (2023). Can artificial intelligence help with scientific writing? Critical care.

Brown, R. D. (2021). Property ownership and the legal personhood of artificial intelligence. Information & Communications Technology Law.

Wan, Y. & Lu, H. (2021). Copyright protection for AI-generated outputs: The experience from China. Computer Law & Security Review.

Paul, B. (2021). Artificial Intelligence and Copyright: An Analysis of Authorship and Works Created by AI. Issue 5 Int'l JL Mgmt. & Human.

Haugen, H. M. (2021). Does TRIPS (Agreement on Trade‐Related Aspects of Intellectual Property Rights) prevent COVID‐19 vaccines as a global public good? The Journal of World Intellectual Property.

Urias, E. & Ramani, S. V. (2020). Access to medicines after TRIPS: Is compulsory licensing an effective mechanism to lower drug prices? A review of the existing evidence. Journal of International Business Policy.

Reyad, Z. H. (). Intellectual Property Are Not Well Protected by the Current National (UK) and International Intellectual Property Rights Ecosystems. aipublisher.org.

Picht, P. G., & Thouvenin, F. (2023). AI and IP: Theory to policy and back again–policy and research recommendations at the intersection of artificial intelligence and Intellectual Property. IIC-International Review of Intellectual Property and Competition Law, 54(6), 916-940.

Rodrigues, R. (2020). Legal and human rights issues of AI: Gaps, challenges and vulnerabilities. Journal of Responsible Technology.

Lepori, B. & Montauti, M. (). Bringing the organization back to Flexing structural responses to competing logic in budgeting. Accounting.

Dornis, T. W. (2020). Artificial creativity: emergent works and the void in current copyright doctrine. Yale JL & Tech.

Damioli, G., Van Roy, V., & Vertesy, D. (2021). The impact of artificial intelligence on labor productivity. Eurasian Business Review.

Stahl, B. C. (2021). Artificial intelligence for a better future: an ecosystem perspective on the ethics of AI and emerging digital technologies.

Cao, J. (2021). Mode optimization and rule management of intellectual property rights protection of educational resource data based on a machine learning algorithm. Complexity.

Indrawati, S. M. & Kuncoro, A. (2021). Improving competitiveness through vocational and higher education: Indonesia's vision for human capital development in 2019–2024. Bulletin of Indonesian Economic Studies.

Li, H., Yang, Z., Jin, C., & Wang, J. (2023). How an industrial internet platform empowers the digital transformation of SMEs: theoretical mechanism and business model. Journal of Knowledge Management.