Intellectual Property in the Age of Artificial Intelligence: Who Owns AI-Generated Creations?

“It’s not where you take things from – its where you take them to.” – Jean-Luc Godard

AI-generated artworks have been contentious in recent times, particularly as the internet witnessed a wave of Studio Ghibli-styled artwork and the likes, which now appear across sectors –from ephemeral uploads on personal social media platforms, to restaurant menus and for other large-scale advertising, marketing and/or commercial strategies– raising an underscoring question: “who holds the rights to AI-generated works?” This article explores the controversial question of ownership and inventorship in works generated by artificial intelligence (AI) and provides key insights on the contemporaneous approaches to AI innovation.

Prior to further investigation, it is essential to recognise that this abstraction falls within the ambit of intellectual property (IP) law wherein IP is defined to refer to “[…] creations of the mind, such as inventions; literary and artistic works; designs; and symbols, names and images used in commerce”.[1] Derived from this sphere of law, people have protected their imagination and gained social or financial recognition through their IP rights, in line with Article 27 of the Universal Declaration of Human Rights.[2]

As the international body for IP information and policy, the World Intellectual Property Organisation (WIPO) supports developing and implementing international law on IP, including through key treaties such as the Paris Convention, the Berne Convention, and the Patent Cooperation Treaty (PCT). International and national laws protect various types of IP –namely, patents, industrial design, trademarks, copyright and related rights. With the disruptive and transformative nature of technology and digital infrastructure, WIPO has had to confront challenging questions regarding the regulation of the innovation ecosystem. Particularly, innovations within AI algorithms have raised curiosities regarding AI’s role in the invention process and, thereby, the larger role of the IP legal ecosystem in addressing these uncertainties.

Since its launch in November 2022, ChatGPT has become a landmark development in generative AI (GenAI), drawing unprecedented global attention to the field. From concerns on AI outputs and copyright-related infringements, to current debates regarding ownership of AI-generated works, existing IP frameworks struggle to provide clear answers or vary on technical grounds, and these issues are being litigated across the globe.

In its factsheet, WIPO recognises

“Generative AI tools can create new content, such as text, computer code, images, audio, sound and video, in response to a user’s prompt, such as a short, written description of the desired output. Current examples of generative AI tools include ChatGPT, Midjourney, Copilot and Firefly”.[3]

The use case of AI in the invention process has been categorised into the following:[4]

i. As a new model of invention

ii. As a tool utilised by humans for or in invention

iii. As an incorporative feature of the invention

iv. As an autonomous generator of material without human input

In its perspective, WIPO highlights that AI is a subject matter which triggers concern for mainly two types of IP –copyright and patents.[5] For the purpose of this article, the primary focus lies on the intersection of AI and IP law to delve into the idea of invention and evaluate legal developments.

The question of ownership of AI-generated creations especially arises within patenting, wherein AI tools are commonly used for the generation or production of work. This may be in the form of assistance further to human intervention or autonomously generated without human intervention. 2019 witnessed a rapid incline of AI-related patenting, with over half of the inventions identified having been born in scientific publications as early as 2013.[6]

Given that IP laws encompass both intangible aspects (corpus mysticum) and tangible objects of creation, innovation or distinction (corpus mechanicum), the use of AI in invention falls under current IP frameworks. However, coupled with the evolving understanding of AI’s role in the invention process, national IP frameworks grapple with the consideration of AI as an inventor.

At the heart of IP is the principle of exclusivity – awarding authors, creators or inventors with the right to their creation. This principle ensures that the right-holder determines how their creation may be used and prevents others from commercially exploiting the same without permission. This trickles down into a time-bound right for patent owners, and the same for copyright-holders which later enters the public domain for free use.

In its wake, GenAI has unsettled traditional IP laws and raised questions about potential legal challenges. The following sections discuss the legal challenges, followed by the implications of AI intervention as an autonomous inventor, referring to a relevant case study.

An interrogation of the legal landscape

• • As mentioned above, the definitional scope of IP incorporates different rights for traditional IP, supported by legal frameworks, and was previously created with the intention to address and protect the creations of individuals without consideration of AI-generated work. Resultantly, it is an unexplored and grey area in the contemporary legal terrain.
  • As with most IP rights, the protection availed is territorial in nature, which unveils the uncertainties that are bound to arise from the lack of harmonious interpretation on AI-generated works. This discrepancy especially harms transnational businesses and activities. Advertently, as evidenced by the DABUS claims discussed in detail below, seeking clarity by resorting to forum shopping.
  • Implications on various IP rights arise as the training process for AI involves data mining/data scraping and other IP-protected data alongside potentially harmful and/or illegal content. AI models often rely on copyrighted material without clear accountability for authorship or protection against infringement. This reveals friction between holders of IP rights and AI investors compounded with the legal uncertainty, such as whether using copyright-protected material in training AI constitutes the infringement of copyright protection.
  • Data insecurity caused by inappropriate and illegal outputs produced by GenAI–such as biased, discriminatory and harmful content.

An invention of AI?

The burning question–can AI be considered as an inventor?–is sought out in a pioneering series of patenting claims surrounding DABUS, where one Stephen Thaler created an AI machine, Device for Autonomous Bootstrapping of Unified Sentience (DABUS), and filed for national patent applications, naming DABUS as the inventor. Thaler claims that the machine is to be credited for two inventions, i.e.,

1. 1. Fractal food and beverage container
   2. Fractal light-flashing device

With the initiation of litigation in over fifteen jurisdictions and pending appeals, presented below is an overview of initial multi-jurisdictional responses:

i. Australia

The case of Thaler v. Commissioner of Patents,[7] wherein the Deputy Commissioner of Patents determined that the patent applicant did not comply with Regulation 3.2C(2)(aa) of the Patent Regulations. At the first appellate instance, the Court allowed an AI machine to be named as an inventor, keeping it consistent with current technological reality.

The decision was overturned in 2022, with the Federal Court ruling in favour of the Commission of Patents, deeming the Deputy Commissioner’s conclusion as correct, i.e., that the application was inconsistent with the regulation by naming DABUS as an inventor.[8] In the same year, the appeal made to the High Court was rejected.

ii. New Zealand

Thaler applied for a patent in New Zealand in May 2021. Following the Assistant Commissioner of Patents’ refusal to accept the patent application based on permitting exclusively human inventorship. In appeal, the High Court of New Zealand takes account of similar overseas applications in parallel, namely Australia, United Kingdom, and the United States of America,[9] and in its judgment dated 17 March 2023,[10] held that it deemed it inappropriate for the Court to expand the definition of an inventor through statutory interpretation. Accordingly, the appeal was dismissed.

 iii. United Kingdom

In 2018, Thaler filed two patent applications for the inventions, stating that he was not the inventor. A year later, the Hearing Officer issued a decision on behalf of the Comptroller, holding that DABUS was not a person as envisaged in the statute, with no entitlement to grant a patent to Thaler on the grounds of owning DABUS, and called for withdrawal of the applications. Subsequently, in the appeal before the High Court, the application was deemed deficient. The consequent appeal was also dismissed.[11] Lastly, the appeal before the Supreme Court was dismissed,[12] noting that the Court did not err in its decision and affirmed the Comptroller’s decision.

 iv. United States of America

In Thaler v. Hirshfeld et. al., the plaintiff approached the Court with a one-count complaint against the defendants’ refusal to recognise an AI machine to be an inventor under the Patent Act.[13] In 2021, the District Court reviewed the legislative language of the statute and ruled in favour of the United States Patent and Trademark Office (USPTO), refusing the application. However, the Court recognised that

“As technology evolves, there may come a time when artificial intelligence reaches a level of sophistication such that it might satisfy accepted meanings of inventorship. But that time has not yet arrived, and, if it does, it will be up to Congress to decide how, if at all, it wants to expand the scope of patent law”.[14]

Despite upholding USPTO’s decision in the Federal Circuit Court, Thaler approached the Supreme Court with the core question of whether “the Patent Act categorically restrict(s) the statutory term ‘inventor’ to human beings alone”, resulting in the refusal of the petition. In lieu of these developments, USPTO also published new guidelines clarifying that for patentability, the inventor must be a natural person and filed guidance relating to the use of AI and outlining legal and ethical responsibilities of users.[15]

Thaler also traversed into copyright law in Thaler v. Perlmutter wherein the plaintiff approached the Court following the refusal by the United States Copyright Office (USCO) to register an AI-generated artwork as a creation autonomously generated by an AI algorithm. The main issue was whether such autonomously generated work by an AI system, ‘Creativity Machine’, with a lack of human authorship, could be copyrighted. Rendering its judgment in 2023, the Court affirmed the denial by USCO and emphasised on the fundamental prerequisite of human creativity for copyright protection.[16]

Regarding patent filing, it should be noted that South Africa was the first jurisdiction to recognise DABUS as an inventor through its patent office’s acceptance of the patent application, though this was through an examination on formality rather than a substantive review. Since then, Saudi Arabia’s patent office has also accepted Thaler’s application listing DABUS as the inventor, though the implications of this decision within the broader GCC patent framework remain to be fully determined. At a regional level, while the European Patent Office (EPO) has addressed AI inventorship cases, it maintains that under the European Patent Convention (EPC), inventors must be natural persons, though it has also recognised that owners of AI systems may have certain rights in the patent application process.[17]

The DABUS litigation exemplifies the fundamental challenges in reconciling traditional IP frameworks with emerging AI technologies. While some jurisdictions like South Africa and Saudi Arabia have recognised AI inventorship, most major jurisdictions have resisted, citing the need for human involvement in inventorship and authorship. This reflects a broader challenge within IP law to reconcile traditional legal frameworks with the growing capabilities of AI. The situation highlights the necessity for legislative updates and international cooperation to address the evolving landscape of AI in intellectual property. The U.S. Court’s recognition of potential future scenarios where AI could meet inventorship criteria suggests a possible shift in the legislative and judicial approach in the future, depending on technological advancements and societal acceptance.

Important to note is Thaler’s petition before the U.S. Supreme Court, in which he calls attention to a looming issue: “AI-generated inventions are increasingly important in many sectors of the economy. But without a reliable ability to patent their breakthroughs, companies large and small will have little protection for the significant investments they must make in research and innovation […]”.[18]

To summarise, an assessment of these country profiles reflects the confines within which national courts have approached the case, mindful of the gravity and relevance of the matter from a jurisprudential point of view, while tackling the specific issues raised before the relevant court. The point of “who or what is the inventor?” is a matter influencing innovation and economic growth.[19]

This legal discourse has inspired and run parallel to examining its ethical implications. Below are salient questions and points stemming from existing legal action and research:

• • Are AI-generated works considered as derivative or original? How would an artist be credited and how would one’s contribution be quantified for just attribution? On a related note, is it ethical to use works that have undefined origins?
  • AI-generated works may negate the incentivisation of producing artwork or creations that an individual would profit from, and potentially their heirs by extension. Further, how would this impact the market value of artists whose works or contributions chip into AI training sets?
  • From a creative standpoint, what would be the cultural and social impact of creations produced by GenAI? Could this gradually result in a homogeneous style of art? How might the role of AI-generated inventions or creations devalue human creativity and innovation?
  • Given that AI models are trained on existing information and material, including personal information, IP-protected materials, or sources that are not credited, this allows for the preservation and proliferation of societal bias and illegal content. To illustrate, Prisma Labs was charged with a class action lawsuit for the breach of personal information through unlawful data collection and illicit storage of such data. Lensa AI, under the company, was marketed as an app for editing and retouching, requiring individuals to upload selfie images to be processed as an artistic image. In addition to biometric extraction, the app often generated highly sexualised images of women.[20]

It is integral to connect these curiosities with concrete solutions and as a starting point for exploration, engagement, and the development of this subject matter. Given the scholarship on these contemporaneous concerns from the WIPO and other key interlocutors, the upcoming segment is forward-looking.

From a Bahraini law perspective

In Bahrain, as in many jurisdictions, the question of who owns AI-generated creations is nuanced and involves interpretations of existing intellectual property laws. Bahrain’s IP framework is governed by Law No. 1 of 2004 on Patents and Utility Models, Law No. 6 of 2006 on Industrial Designs, and Law No. 22 of 2006 on Copyright and Related Rights, and is further influenced by international treaties and agreements, including the TRIPS Agreement, the Paris Convention, and the GCC Patent System, demonstrating its integration with both global and regional IP protection mechanisms. However, specific provisions addressing AI-generated works are not yet clearly defined in Bahrain’s legal system.

1. 1. Copyright Law: Under Law No. 22 of 2006, Bahrain’s copyright law traditionally requires a human author for protection. Works must result from human creativity and originality, as defined in Article 1 of the law. Therefore, AI-generated works without human intervention may not qualify for copyright protection under current laws. If an AI tool is used as an instrument by a human creator, the human’s contribution could be recognised, allowing for copyright protection of the resultant work.
   2. Patent Law: Under Law No. 1 of 2004, Bahrain’s patent law, like many other jurisdictions, defines an inventor as a natural person, and requires that inventions meet the criteria of novelty, inventive step, and industrial applicability. Therefore, AI cannot be named as the inventor under the current patent framework. Innovations created using AI tools would need to have a human inventor who contributed to the inventive process, even if AI played a significant role in the creation.
   3. Policy and Legal Development: As AI technology evolves, Bahrain, in line with its Economic Vision 2030 and commitment to digital transformation, may need to consider legislative updates to address the unique challenges AI-generated creations pose, particularly within the framework of existing IP laws. This could involve redefining authorship and inventorship criteria to accommodate AI’s role in creative and inventive processes.

In summary, under current Bahraini law and in accordance with Law No. 22 of 2006 and Law No. 1 of 2004, AI-generated works face significant challenges in obtaining IP protection unless there is substantial human contribution to the creative or inventive process. As the technology and its applications continue to develop, Bahrain may look to international practices and consider reforms to address these emerging issues effectively.

Considerations and Conclusion

In the pursuit of creating a comprehensive and competent legal landscape that is conducive for innovation, the following strategic considerations are a step towards carefully interlacing the worlds of AI and IP:

• • Uphold the fundamental principles of IP law to protect human creativity and innovation, while maintaining appropriate incentives for continued technological advancement.
  • Reassess and redefine the criteria for IP protection under Bahraini law to ensure effective and equitable measures regarding authorship, creation, and inventorship in the contemporary AI landscape with AI.
  • Foster international cooperation, particularly within the GCC framework, to maximise benefits and minimise risks while developing trustworthy AI systems that comply with regional and international IP standards.[21]
  • Anticipatory governance steered by diverse and forward-looking AI governance practices. While still in early stages of development, several jurisdictions have established leading practices and valuable precedents on AI governance.[22]
  • Policy development, broadly based on an interrelated tripartite classification:

1. 1. Advocate for multi-stakeholder engagement, including government entities, private sector, and academic institutions, in developing and updating IP policies and laws to establish robust governance mechanisms aligned with Bahrain’s legal framework. Simultaneously, promote investment in AI research and development in accordance with the Economic Vision 2030.
   2. Monitor and evaluate global approaches and policies, shaping legal jurisprudence and affecting well-rounded international standards and practices.
   3. Develop comprehensive legal policies under Bahraini IP law regarding ownership and licensing of AI-generated works, while establishing clear protective frameworks that define and safeguard individual contributions to AI-assisted creative processes.

• • Prioritise ethical considerations when targeting the intersection of AI and IP to establish sustainable and responsible AI systems. To achieve this, governments are and should implement national strategies and ethics frameworks.[23] For instance, the implementation of human-centred values and fairness,[24] and initiatives to reduce AI bias.[25]

To conclude, this niche area demands extensive consideration to harness the ethical and responsible use of GenAI. As of 2020, more than 340,000 AI-related patent applications had been filed globally since the emergence of AI technology, accompanied by over 1.6 million scientific publications in the field.[26] There is no question about the immense potential presented by AI technology. In light of the rollout of legal battles into other areas of IP, it is imperative to expedite the development of comprehensive legal frameworks that address the evolving IP ecosystem.

Integrating artificial intelligence into the realm of intellectual property presents a complex array of challenges and opportunities. As AI continues to advance and generate creations that blur the lines between human and machine input, existing IP frameworks are tested and often found wanting. Recent landmark cases such as DABUS (Device for Autonomous Bootstrapping of Unified Sentience) in multiple jurisdictions, including the 2021 South African patent grant and subsequent global rejections, illustrate the ongoing challenge of defining inventorship and authorship in the age of AI. Courts and legislators face the daunting task of balancing the protection of human creativity with the recognition of AI’s contributions. Under Bahraini Law No. 1 of 2004 on Patents and Utility Models, as well as Law No. 22 of 2006 on Copyright and Related Rights, and similar to other jurisdictions such as the United States, current IP frameworks explicitly require human inventorship or authorship for IP protection, underscoring the need for legislative evolution to accommodate technological advancements. Moving forward, a harmonised approach between regional bodies (particularly the GCC Patent Office under the GCC Patent Regulation, the EU, and ASEAN) and international organisations (such as WIPO), coupled with ethical considerations, Bahrain’s Economic Vision 2030 objectives, and innovative policy development, is essential to create a cohesive legal landscape that fosters innovation while safeguarding the rights of creators across jurisdictions. Only through such measures can we adequately address the transformative impact of AI on intellectual property and ensure a sustainable future for both human and AI-generated innovations.

Authors:

Saad Al Doseri

Srivani P. Nair

[1] World Intellectual Property Organization (WIPO), What is intellectual property? (WIPO publication no. 450(E), 2020), p.1.

[2] Universal Declaration of Human Rights (adopted 10 December 1948 UNGA Res 217 A(III) (UDHR) art 27.

[3] WIPO, Generative AI: Navigating Intellectual Property (WIPO, 2024), p.2.

[4] WIPO, Getting the Innovation Ecosystem Ready for AI: An IP policy toolkit (Geneva, 2024), p.12.

[5] WIPO, ‘Conversation on Intellectual Property and Artificial Intelligence’ (Second Session, May 2020).

[6] WIPO, WIPO Technology Trends 2019: Artificial Intelligence (Geneva, 2019), p.14.

[7] Thaler v. Commissioner of Patents [2021] FCA 879.

[8] Commissioner of Patents v. Thaler [2022] FCAFC 62.

[9] Thaler v. Commissioner of Patents [2023] NZHC 554, para. 15-21.

[10] Ibid., para 33.

[11] Thaler v. Comptroller-General of Patents, Designs and Trade Marks, [2021] EWCA Civ 1374.

[12] Thaler v. Comptroller-General of Patents, Designs and Trade Marks, [2023] UKSC 49.

[13] Thaler v. Hirshfeld et. al., No. 1:20-cv-903, 2021 WL 3934803.

[14] Ibid., para 130.

[15] See Department of Commerce, Guidance on Use of Artificial Intelligence-Based Tools in Practice Before the United States Patent and Trademark Office (Federal Register Vol. 89, No. 71, April 2024).

[16] Thaler v. Perlmutter, No. 1:22-cv-01564 (D.D.C. Aug. 18, 2023) (2023).

[17] See Thaler v. Vidal; Petition for Writ of Certiorari, p.31, available at: <https://www.supremecourt.gov/search.aspx?filename=/docket/docketfiles/html/public/22-919.html>.

[18] Ibid., p.28.

[19] Supra note 16, p.2.

[20] Flora et. al. v. Prisma Labs Inc., 3:23-cv-00680, (N.D. Cal.).

[21] OECD AI Principles (May 2019).

[22] OECD, ‘Steering AI’s Future: Strategies for Anticipatory Governance’ (OECD Artificial Intelligence Papers No. 32, OECD Publishing, February 2025).

[23] OECD, ‘The State of Implementation of the OECD AI Principles Four Years On’ (OECD Artificial Intelligence Papers No. 3, OECD Publishing, October 2023).

[24] Ibid., p.15.

[25] Ibid., p.34.

[26] WIPO, WIPO Technology Trends 2019: Artificial Intelligence (Geneva, 2019), p.13.