The Ruling
Kenya's Copyright Tribunal has ruled that a work made with AI assistance cannot be copyrighted unless a human contributed genuine creative input — and that AI itself, lacking legal personality, can never be listed as an author. The decision came in Aryeh Movement Limited v. Cynthia Beldina Akoth Okello (COPTA/E001/2025), a dispute over a collection of Bible Scripture Stories that Akoth wrote with AI tools while contracted by Aryeh Movement, a Nairobi-based media outfit.
The facts are unremarkable until they aren't. After the working relationship broke down, Akoth discovered Aryeh had registered the stories with the Kenya Copyright Board (KECOBO) in its own name. She filed a complaint; KECOBO revoked Aryeh's certificate; Aryeh appealed to the Tribunal. On appeal, the Tribunal made two findings that matter well beyond this one contract dispute. First, on the merits: "There has to be an element of sufficient human intervention to determine the extent of copyrightability of works generated by AI," and neither Akoth nor Aryeh could show it — meaning the underlying stories may not be protected by anyone. Second, on process: KECOBO overstepped by trying to resolve an authorship dispute at all, since its statutory role is limited to maintaining the copyright register, not adjudicating who owns what (CM Advocates; Zawya).
The Legal Hook
The Tribunal grounded its reasoning in Section 22(3) of Kenya's Copyright Act, 2001, which withholds protection from a literary, musical, or artistic work unless "sufficient effort has been expended on making the work to give it an original character" and it has been "reduced to material form" (Copyright Act No. 12 of 2001). That threshold has always presupposed a human doing the effort. The Tribunal simply confirmed that a prompt is not effort in the sense the statute means, and that AI, having no legal personality, cannot be the "person" the Act's authorship definitions require.
That puts Kenya squarely alongside the United States, where the D.C. Circuit held in Thaler v. Perlmutter that "multiple provisions make clear that authors must be human beings" — a ruling the Supreme Court left standing in March 2026 by denying certiorari (D.C. Circuit opinion). The UK and EU have reached similar conclusions. Kenya's Tribunal did not invent a novel doctrine; it imported an increasingly global consensus into a statute written years before generative AI existed.
The Case for the Line
It's worth taking the strict view seriously before critiquing it. Copyright exists to reward the specific labor of human creative choice — not merely to reward whoever presses a button first. If AI-generated text qualified for protection on its own, registries would fill with machine output claimed by whoever registered first, exactly what Aryeh tried to do here. A low bar for AI authorship would also let intermediaries — publishers, contractors, platforms — capture rights over work that no one meaningfully created, squeezing out the human writers, illustrators, and translators the system is supposed to protect. In a market like Kenya's, where AI tool adoption is now among the highest in the world — 97.5% of online adults reported using at least one AI tool in the past month, per DataReportal's Digital 2026 report — that risk is not hypothetical (KICTANet). A tribunal that let "I typed a prompt" satisfy Section 22(3) would have effectively deleted the originality requirement for an enormous and growing share of new content.
Where Proportionality Should Bite
The floor is right; the test above it is not yet usable. "Sufficient human intervention" tells a court what to look for after a dispute has already destroyed a business relationship — it tells nobody, ex ante, how much editing, prompting, curating, or restructuring is enough. Contrast this with the more granular (if still contested) U.S. approach, where the Copyright Office and courts have pointed to specific human acts — selection, arrangement, and substantive modification of AI output — as the operative test, rather than a single undefined phrase. Kenya's creative and publishing sector, already leaning hard on AI tools, now has a legal standard it cannot apply prospectively; it will only find out where the line sits when it gets sued.
The Tribunal's second holding compounds the problem. By ruling that KECOBO cannot resolve ownership disputes and must leave them to litigation, it closed off the cheap administrative remedy that a solo author like Akoth actually used to get relief. Future disputes over AI-assisted works — which, given adoption rates, will not be rare — must now go straight to the Tribunal or the courts, raising costs for exactly the independent creators the human-authorship rule claims to protect.
A Codification Opportunity
KECOBO opened public comment on a new Copyright and Related Rights Bill in March 2026 — the first full statutory rewrite since 2001 (KECOBO, Invitation for Public Comments). That bill is the right vehicle to convert this Tribunal ruling from a one-off precedent into a workable rule: codify the human-authorship floor, but attach concrete factors — degree of prompt specificity, iterative editing, curation and arrangement of outputs — so creators and the publishers who contract them know where they stand before, not after, a dispute. Kenya got the destination right. It now needs the statute to supply the map.