📌 1. Core Copyright Challenges for AI/Computer‑Generated Heritage Replicas

When AI or computational systems produce digital reconstructions or replicas of heritage sites (e.g., 3D models, VR environments, photorealistic images), several legal obstacles arise:

A. Authorship & Ownership

Under Kenyan law, copyright protection is tied to a human author or legal person. Current statutes presume human creative effort as a precondition to copyright. Purely machine‑generated works with minimal human input may not qualify.

B. Originality

Copyright in Kenya requires that a work be original — reflecting significant human skill, labour or judgment, not merely technical output created without human guidance.

C. Infringement Risks from Input Data

If an AI model was trained or fed copyrighted architectural plans, photos, or artistic depictions of heritage sites without permission, output visuals may inadvertently reproduce or derive from copyrighted material, risking infringement.

D. Incidental Inclusion & Secondary Use

Use of copyrighted images or elements within a larger digital environment may sometimes be treated as incidental and not infringing — but this depends on whether the inclusion is essential and recognisable.

E. Enforcement and Proof Issues

Establishing that output violates rights requires evidence of similarity, ownership, and unauthorised reproduction — these evidentiary standards have been tested in Kenyan copyright cases.

📚 2. Kenyan Case Laws Interpreting Copyright Principles

There are no reported Kenyan decisions exclusively about AI‑generated heritage replicas yet, but the following cases illustrate how Kenyan courts currently interpret copyright ownership, authorship, and infringement — laying the groundwork for future AI disputes.

✅ Case 1 — Mwangi Kirubi t/a Aa Click Picture Works Africa v Ink Productions Ltd (2022)

Facts

The plaintiff sued a film producer for reproducing his photograph in a drama production without permission.

Legal Findings

Ownership of copyright exists even without formal registration — the creator’s testimony and evidence can establish authorship under the Copyright Act.

However, the court held that incidental inclusion of a copyrighted visual in a film or broadcast does not, on its own, constitute infringement where the work is not a substantial part of the defamatory production.

Implication for AI Replicas

Digital reconstructions that merely include copyrighted images of heritage sites as background elements may not be infringing if the copyrighted content plays no central role — although this hinges on how “substantial” or “recognisable” such elements are.

✅ Case 2 — Nairobi Map Services Ltd v Airtel Kenya Ltd (2019) (Court of Appeal)

Facts

The appellant claimed that a television advertisement reproduced an administrative map (its copyright).

Key Principle

The Court of Appeal upheld the High Court’s decision that incidental inclusion of a copyrighted work (here the map) in another product was not infringement under section 26(1)(c) of the Copyright Act.

Relevance

If an AI replica contains copyrighted imagery of cultural artefacts as incidental components (e.g., as textures, backgrounds), similar defenses might apply — especially where the original is not the focus.

✅ Case 3 — Kimani v Safaricom Ltd & Others (2023)

Facts

A musician brought a claim against Safaricom and digital platforms for distributing his music (Skiza Tunes) without proper licensing.

Outcome

The court held that distributors must conduct due diligence to confirm that copyright licences are valid.

Damages were awarded for unauthorised reproduction, broadcasting, and distribution.

Inference for Digital Replica Platforms

Digital marketplaces or reconstruction platforms streaming or selling AI‑generated heritage visuals must ensure clear rights to architectural or artistic material used in training or production, or face infringement liability.

✅ Case 4 — Safaricom Ltd v Transcend Media Group (2020)

Facts

Safaricom alleged infringement of copyright works in the implementation of a digital marketing campaign.

Key Points

The court reaffirmed that copyright subsists without registration, and protection arises from the mere existence of original creative work.

Establishing ownership is critical before claiming infringement.

Application to AI Output

Reconstructed digital heritage visuals — if demonstrably original and creatively composed by humans using AI tools — may be copyrighted by the human authors or rights holders without registration.

✅ Case 5 — Kenya Copyright Board v Said (2024) (High Court)

Facts

This appeal concerned procedural and jurisdictional issues related to a copyright enforcement claim.

Outcome

The appeal was allowed, overturning the lower court’s judgment and dismissing the suit.

Significance

While not directly about originality or AI, this case highlights the technical nature of proving infringement in Kenyan courts and illustrates that procedural steps and clear ownership must be established for copyright enforcement.

📌 3. Emerging Kenyan Position on AI‑Generated Works

Although direct case law about AI‑generated digital replicas is sparse, recent administrative rulings and tribunal interpretations give insight:

AI and Human Authorship — Copyright Tribunal

In Aryeh Movement Ltd v Cynthia Beldina Akoth Okello (Copyright Tribunal, 2025), the Tribunal held that purely AI‑generated works are not eligible for copyright protection unless there is sufficient human input and originality.

Statutory Basis — Kenyan Copyright Act

Section 22(3) of the Act requires that a work must reflect sufficient effort giving it original character and be in a material form — this inherently presumes human creativity.

Interpretation by Practitioners

Legal commentators and IP specialists note that under current Kenyan law, machine‑only output is likely unprotected, and any copyright must be grounded in the human’s creative contribution or intervention in generating the output.

📌 4. Why These Challenges Matter for Heritage Site Replicas

Digital replicas of heritage sites often incorporate:

Photographs and scans of physical locations (possibly copyrighted),

Architectural blueprints or artistic renderings from third parties,

AI models trained on copyrighted images of the sites without consent.

Legal Risk Areas

Authorship disputes — who owns the output if multiple humans and AI contribute.

Copyright eligibility — whether purely AI‑generated parts can be copyrighted.

Infringement claims — whether output reproduces substantial parts of copyrighted source material.

Moral rights — original creators of architectural or artistic expression might claim distortion, misattribution, or integrity violations.

Licensing obligations — platform intermediaries must ensure they are licensed or risk liability — similar to music distribution cases like Kimani v Safaricom.

📌 5. Practical Guidance in Light of these Cases

To navigate these issues and mitigate legal risks:

✔️ Document Human Creative Input

Author(s) should demonstrate specific contributions, decisions, and editing that distinguish the final output from mere machine output.

✔️ Clear Rights to Input Data

Secure licences or permissions for any copyrighted images, plans, or drawings used in training AI models.

✔️ Artefact Segregation

If output includes copyrighted content owned by others, determine whether its use is incidental or core — incidental use may sometimes avoid infringement claims.

✔️ Contracts & Ownership Agreements

When multiple contributors (e.g., model trainers, curators, editors) participate, use written agreements to clarify ownership of resulting works.

📌 6. Summary — Copyright Frontier in Kenyan Digital Replicas