Authored By: NYAKARA JOY SHALOM
STRATHMORE UNIVERSITY
I. Introduction
Kenya’s ambition to lead Africa’s technology sector is no longer rhetorical. In 2025, the government launched the National AI Strategy 2025–2030, explicitly building on the country’s reputation as the “Silicon Savannah” and committing to a framework that prioritizes AI-driven innovation across agriculture, healthcare, finance, and public services.1 The AI market in Kenya is projected to grow at an annual rate of approximately 28% through 2030, with Google, Amazon Web Services, Microsoft, and IBM all having established development centers in Nairobi.2 Yet as Kenyan corporations increasingly deploy AI systems to generate commercial outputs — software, marketing content, data-driven analyses, and creative works — a critical legal question goes unanswered: who owns what the machine produces?
Under the Copyright Act (Cap. 130) of Kenya, copyright belongs to human authors. Under section 22(3), copyright is granted only where human effort can be shown to have contributed to a work’s originality, as reiterated by the Copyright Tribunal of Kenya in Aryeh Movement Limited v. Cynthia Beldina Akoth Okello.3 Although this case has been narrowly applied, its effects are commercially wide-reaching: businesses that invest in artificial intelligence technology receive no protection for the valuable outputs they generate using AI tools.4 This paper seeks to demonstrate that an untapped provision under Kenya’s Copyright Act can fill this gap, and that filling it is imperative to the successful implementation of the country’s Artificial Intelligence strategy.
II. The Legal Framework: A Provision in Search of Application
The Copyright Act (Cap. 130) defines “author,” in relation to a literary, dramatic, musical, or artistic work or computer program that is computer-generated, as ‘the person by whom the arrangements necessary for the creation of the work were undertaken.’5 This provision, found in Section 2, is almost identical to Section 9(3) of the Copyright, Designs and Patents Act 1988 in the United Kingdom, which provides at least some form of protection for computer-generated works under British law. The provision is not novel; it has existed in Kenyan law for over two decades, yet has largely gone unnoticed.
There is a reason for this apparent lapse. The provision reflects an earlier era in which computers were used only to carry out direct commands from a human programmer. The current situation, in which artificial intelligence generates text, images, video, and other material based on patterns learned from millions of data points, is fundamentally different and far more complex. It is correspondingly difficult to identify the human being who made a sufficient effort to render the work original, as required by Section 22(3).
This difficulty is illustrated by the Copyright Tribunal’s judgment in Aryeh Movement Limited v. Cynthia Beldina Akoth Okello [2025]. Under Sections 22(3) and 23(1), the Tribunal ruled that because the authorship requirement implies human involvement, a work created wholly by a generative artificial intelligence program does not qualify for copyright protection.6 This produces an all-or-nothing outcome: either sufficient human creativity can be shown, or the work receives no protection under the law at all.7
III. The Commercial Gap and Its Consequences
The practical implications of this deficiency are profound. A Nairobi-based fintech that uses artificial intelligence to develop credit-scoring models, a Kenyan law firm that uses AI to generate large volumes of contractual templates, and an African media company that uses generative models to produce localized content are all corporations that invest money, incur risk, and control the process by which these works are created. Yet under the present regime, none of them holds copyright in the resulting work, because the process lacks sufficient human creative input.
This produces what may be called the corporate beneficiary problem: the organization that bears the financial and technical risk of using an AI system is systematically disqualified from owning the product of that process. The problem is sharpened by comparison with the work-for-hire doctrine that governs traditional employment relationships. Section 33 of the Copyright Act vests copyright in an employer where the author of a literary work is its employee.8 But an AI system is not an employee, which eliminates the work-for-hire framework as a ready-made solution.
The effects do not end there. Kenya’s National Artificial Intelligence Strategy 2025–2030 explicitly identifies intellectual property protection as one of the conditions necessary to attract investment in AI. Where the legal system fails to protect the intellectual property generated through AI, it creates a deterrent in precisely the place an incentive is needed. Such a failure could result in the under-utilization of AI by companies, the relocation of IP-holding subsidiaries to jurisdictions with stronger legal protection, or — worse still — the misattribution of AI-generated work to a human author simply to secure copyright. Each of these outcomes raises further legal questions of its own.
IV. Toward a Purposive Interpretation and Legislative Reform
The definition of “author” in Section 2, as it applies to computer-generated work, offers a solution to this dilemma without requiring legislative amendment. Under that definition, authorship lies with ‘the person by whom the arrangements necessary for the creation of the work were undertaken.’ Where a corporation undertakes the procurement, training, deployment, and control of an artificial intelligence system, those actions can reasonably be described as the arrangements necessary for the creation of the resulting work. Read purposively, this aspect of the definition supports treating the deploying corporation as the author.9,10
The Copyright Tribunal and KECOBO should adopt a similar approach where applicable. Doing so would not necessarily require reversing the ruling in Aryeh Movement, since that decision concerned pure AI outputs produced without any corporate arrangement.11 Instead, the doctrine should develop a distinction between pure AI outputs produced without purposive arrangements and those produced through them.
In the absence of clear judicial development, legislative amendment would be the more reliable route. The Copyright Act could be amended to add a subsection recognizing that copyright subsists in computer-generated works and vests in the entity that made the necessary arrangements for their creation, for a shorter term — say, 25 years — than that granted to human-authored works. This would ensure that companies investing in computer systems and software to produce such works receive meaningful recognition, without extending to them the full, permanent-style monopoly protection accorded to human authors. A disclosure requirement for works registered as computer-generated would also help preserve the integrity of ownership records.
This approach would align with the UK position under CDPA s. 9(3), adapted to Kenya’s development context.12 It would also signal to the international technology investors drawn to Kenya by its AI Strategy that works created within Kenyan territory can be protected.
V. Conclusion
A momentous decision awaits Kenya. The Copyright Act already contains what the country needs to lead the continent in leveraging AI for economic development. But because its “arrangements” clause has been read narrowly, the statute has left Kenyan AI-generated creations in a no-man’s land that undermines the very objective the AI Strategy sets out to achieve. The Aryeh Movement decision, while doctrinally justifiable on its facts, illustrates a gap that the Copyright Act is only partly equipped to close.
The way out is two-fold. First, KECOBO or the Copyright Tribunal could adopt a purposive interpretation of Section 2’s definition of computer-generated works, recognizing corporate authorship for entities that have taken conscious steps to arrange for the deployment of AI technology. Second, where that interpretive approach proves insufficient, Parliament could amend Section 22 so that computer-generated works arising from such arrangements vest in the company and receive limited protection under copyright law. Kenya does not have to wait for global consensus on AI and intellectual property rights to act.
Note(S):
1. TechAfrica News, ‘Kenya Unveils AI Strategy 2025–2030 to Lead Africa’s Digital…’ [title incomplete in source — please verify full title]; National AI Strategy 2025–2030 (Government of Kenya, 2025).
2. US Commercial Service, ‘Kenya Digital Economy’ (Country Commercial Guide, 2024).
3. Copyright Act (Cap. 130) s 22(3).
4. Aryeh Movement Limited v. Cynthia Beldina Akoth Okello [2025] COPTA/E001 (Kenya Copyright Tribunal).
5. Copyright Act (Cap. 130, Laws of Kenya) s 2, para (g).
6. Copyright Act (Cap. 130) ss 22(3), 23(1).
7. Aryeh Movement Limited v. Cynthia Beldina Akoth Okello [2025] COPTA/E001 (Kenya Copyright Tribunal).
8. Copyright Act (Cap. 130) s 33.
9. Bowmans Law, ‘Kenya: Navigating Copyright and Generative AI in the Music Industry’ (2 September 2025).
10. Wamae & Allen Advocates, ‘Legal Status of Copyrights or Trademarks Generated by AI Agents’ (Legal Update No. 106, 2025).
11. Aryeh Movement Limited v. Cynthia Beldina Akoth Okello [2025] COPTA/E001 (Kenya Copyright Tribunal).
12. Copyright, Designs and Patents Act 1988 (UK) s 9(3); Andres Guadamuz, ‘Do Androids Dream of Electric Copyright?’ (2017) 2(1) Intellectual Property Quarterly 169.
