Africa’s Copyright Push Turns to AI Training and Platform Royalties
In August 2026, the U.S. government’s “IP for Growth” initiative pushed the African creative-industry copyright debate beyond conventional anti-piracy enforcement and toward digital distribution, royalty collection and generative AI. After its Geneva launch, music-industry workshops were held in Lagos and Johannesburg, followed on August 19 by a public briefing from Katherine Hiner, the USPTO’s intellectual property attaché for sub-Saharan Africa. Regional and industry reporting on August 20–21 amplified two linked policy signals: digital copyright systems must do more than recognize rights on paper, and AI training disputes are increasingly being tested against existing copyright doctrines rather than treated as a separate technological exception.
One figure drew particular attention: the workshops cited an estimate that Nigeria and Kenya alone leave roughly US$286 million in recorded-music revenue uncollected each year. The policy response presented publicly is not a single “platform royalty law.” Instead, it links WIPO internet treaties, technological protection measures, rights-management information, collective management organisations (CMOs), anti-piracy enforcement and AI copyright analysis. For platforms, rightsholders and creators, the practical issue is whether ownership, licensing, data use and payment flows can be made traceable, explainable and enforceable at the same time.
Copyright policy is becoming an income infrastructure question
The clearest message in the public briefings is that digital copyright enforcement is being framed as infrastructure, not merely litigation. Hiner highlighted the WIPO Copyright Treaty and the WIPO Performances and Phonograms Treaty, alongside technological protection measures and rights-management information. The commercial logic is straightforward: content may move frictionlessly across borders, but licensing scope, ownership data and payment obligations cannot be allowed to disappear in that movement.
This helps explain why CMOs are central to the discussion. Music rights are fragmented across compositions, sound recordings, performers, publishers and territorial arrangements. At scale, platforms cannot negotiate individually with every rightsholder. A well-run CMO can reduce transaction costs, but only if repertoire data are accurate, tariffs are transparent, cross-border reciprocal agreements reconcile properly, distributions are timely and unmatched income is handled credibly.
It would therefore be misleading to describe the current “IP for Growth” agenda as a public demand for mandatory collective management across all streaming and short-video services. The public material supports a narrower, but still consequential, proposition: where direct licensing is commercially impractical, collective licensing bodies must function as auditable and trustworthy nodes in the payment chain. The regulatory pressure is likely to focus on concrete questions such as why revenue was not collected, why it was not matched, and why it did not reach the correct beneficiary.
Fair use is not a blanket licence for AI scraping
The AI discussion also requires careful reading. In the August 19 briefing, Hiner said that established doctrines such as fair use and fair dealing would play an important role in balancing creators’ interests with technological innovation, while stressing that disputes must be assessed on their facts and through existing legal processes. That is materially different from announcing a general exemption for scraping copyrighted works to train AI systems.
For generative-AI platforms, the legal question cannot be managed by simply asserting that training is “fair use.” In practice, disputes are likely to turn on the nature of the works, the purpose and scale of copying, how training data were obtained, whether protectable expression is retained or reproduced, the effect on existing or emerging licensing markets, and the specific doctrine adopted in the relevant jurisdiction. Some countries apply fair use, others fair dealing, and others rely on more specific text-and-data-mining exceptions. Cross-border training makes those differences harder to ignore.
If African jurisdictions borrow from U.S.-style fair-use reasoning in future AI policy, that would not automatically give platforms an entitlement to ingest content. A more realistic outcome is a fact-intensive dispute structure in which platforms must document provenance and purpose, rightsholders must identify the works and economic interests affected, and regulators or courts draw the line. For businesses, data inventories, licence-chain records and deletion mechanisms are more operationally valuable than a broad abstract statement that training is lawful.
Platform governance is moving beyond takedowns to royalty flows
Digital copyright compliance has long been associated with notice-and-takedown systems, pirate links and account enforcement. The US$286 million uncollected-revenue estimate points to a different problem: lawful consumption may already be taking place while revenue is still not being identified, measured or distributed correctly. The loss may result from piracy, but it can also arise from missing metadata, failed rightsholder matching, incomplete territorial licensing information, CMO-platform reconciliation problems or licensing gaps between short-form video, user-generated content and traditional music rights models.
This shifts the governance question from “Did the platform remove infringing content?” to “Does the platform have a reliable system for identifying rights and settling revenue?” Streaming services may possess granular usage data, yet accurate payment still depends on matching that usage to identifiers, publication data and territorial rights chains. Short-video and UGC platforms face an even more complex stack because one post can simultaneously implicate recording rights, composition rights, performance interests, adaptations and user-created material.
The next policy development to watch is not necessarily a single African royalty regime. More plausible is a gradual rise in minimum expectations for reporting, rights metadata, CMO transparency and dispute handling. That would increase compliance costs for platforms, but it would also impose stronger data-quality duties on rightsholders. Without reliable ownership metadata, platforms cannot pay accurately; without verifiable usage data, rightsholders cannot audit or claim effectively.
What creators, collecting bodies and AI platforms should fix now
For labels, publishers and creators, the first priority is not to wait for legislation. Rights chains should be converted into verifiable data assets. Recording ownership, composition rights, performer interests, commissioned works, territorial licences, sublicensing and CMO mandates should reconcile with one another. For international distribution, registration and collection arrangements should be checked market by market. Unallocated revenue often does not mean the right does not exist; it means the system cannot identify the correct payee.
Streaming and short-video platforms should broaden copyright governance from complaint handling to licensing and settlement controls. Repertoire sources, conflicting claims, duplicate claims, automated matching, suspense accounts, cross-border deductions and CMO reconciliation all deserve internal-audit treatment. Where AI is used for recommendations, generation, music tools or content transformation, platforms should also distinguish between processing required to deliver a service and ingesting works for model training. A single broad terms-of-service clause is a weak substitute for that distinction.
Generative-AI companies should focus on training-data governance: preserve records of source and acquisition method, separate licensed data from publicly accessible material and rights-restricted material, maintain rightsholder complaint and deletion channels, and assess direct or collective licensing for high-value catalogues. Fair use or fair dealing may form part of the legal analysis, but it should not substitute for provenance and permissions management.
“IP for Growth” remains a policy and capacity-building initiative rather than a binding cross-border code. Its direction is nevertheless clear. The next phase of African digital copyright competition will be shaped not only by who controls content, but by who can prove where that content came from, who authorised it, how it was used and where the money went. AI training and platform royalties sit on the same policy agenda because both ultimately test whether rights information can keep pace with digital scale.
