For years, Kenyan practitioners advising on intellectual property have had to borrow answers from other jurisdictions whenever a client asked an increasingly common question: who owns work created with the help of artificial intelligence? The Copyright Tribunal’s recent decision in Aryeh Movement Limited v Akoth does not settle that question definitively, but it is the first time a Kenyan adjudicative body has put its mind to it in a reasoned way.
The case itself was, on its face, a jurisdictional dispute. The applicant was challenging a decision of the Kenya Copyright Board (KECOBO) on a copyright registration, and the Tribunal’s primary task was to determine whether it had the power to hear the matter at all. Nothing in that framing suggested the case would become a reference point for AI and authorship.
The Tribunal stated that the Copyright Act does not contain a bespoke regime for AI-generated works. There is no provision that tells us what happens when a machine, rather than a person, produces an output that looks like a protectable work. Under section 22 of the Copyright Act, a work qualifies for protection only where it is original, and originality in our law has always meant that sufficient effort has been expended to give the work its own character, coupled with the requirement that it be reduced into material form. That test was written with a human creator in mind. The question the Tribunal had to consider was whether it could stretch to accommodate a non-human contributor.
Its answer was that it could not, at least not on its own as effort, skill and judgment are attributes of a mind, not a computational process. The Tribunal was not prepared to sever “originality” from the idea of human authorship simply because the tools available to creators have changed. In arriving at this position, it drew comfort from comparative authority with the same conclusion reached by the United States Copyright Office and the United Kingdom’s Copyright, Designs and Patents Act framework for computer-generated works, and by the European Court of Justice’s long-standing insistence, that copyright protects only the author’s own intellectual creation. None of those jurisdictions have identical statutory language to Kenya’s, but the underlying philosophy is the same: copyright is a reward for human creative choice, not for computational output, however sophisticated.
What is more useful for practitioners is the second half of the Tribunal’s reasoning that the recognition that AI involvement does not automatically disqualify a work from protection. Where a human author uses AI as an instrument, in the way a photographer uses a camera or a composer uses software, and retains creative control over the selection, arrangement and expression of the final product, copyright can still subsist in that human-authored layer. This is an important clarification because it avoids two extremes: treating every AI-assisted work as unprotectable, which would be commercially unworkable given how embedded these tools now are in ordinary creative and professional practice, and treating AI-generated output as automatically protectable, which would strip the concept of authorship of any real meaning.
The Tribunal, however, declined to rule on whether the specific works in dispute were copyrightable, because the works were never produced in evidence and there was no material on record showing how much of the final product was attributable to human input versus the AI system used. Going forward, parties who want to claim authorship over AI-assisted work will need to build an evidentiary record using drafts, prompts, revision history, and a clear account of the creative decisions made by the human party rather than assume that a finished product will speak for itself.
This decision, obiter as it is, has done useful work in flagging the direction Kenyan law is likely to take. Practitioners advising clients in publishing, advertising, media, software and increasingly in legal drafting itself should start treating the human contribution to AI-assisted work as something to be documented, not assumed. Contracts allocating IP ownership in commissioned work should now expressly address AI involvement, and businesses relying on AI tools to generate marketing copy, designs or code should keep records demonstrating the human judgment exercised over the final output.
KECOBO’s registration practices will also need to evolve to ask the right questions at the point of filing, rather than leaving these issues to be litigated after the fact. Until Parliament legislates specifically on AI and authorship, which is very likely on the horizon given global regulatory trends, this decision offers Kenyan practitioners their first, cautious signpost: while AI is increasingly being used, the author, for now, still has to be human.
