AI Drug Discovery East Africa: Ownership and Access

 AlphaFold 3 is a useful starting point, not a shortcut to a medicine

AlphaFold 3 has become a powerful symbol of AI assisted science. The 2024 Nature paper describing the system reports that it can predict joint structures involving proteins, nucleic acids, small molecules, ions and modified residues, with major improvements for several kinds of biomolecular interaction. That can help researchers understand how potential therapeutic molecules may interact with biological targets. It does not mean that an AI system can produce a market ready medicine simply by predicting a structure.

For East African business law, that distinction is essential. A useful AI output can sit at the beginning of a long chain that includes laboratory work, candidate selection, preclinical research, clinical evidence, manufacturing, patent strategy, licensing and regulatory approval. The legal question is therefore wider than “who owns what the AI discovered?” It is who owns each legally relevant layer, who is entitled to commercialise it, what evidence is required before patients can receive it, and whether access conditions were built into the research relationship early enough to matter.

The AI provider does not automatically own the medicine

A company that provides an AI model may own or license the software, model weights, interface, confidential know how and contractual rights governing use. None of that automatically makes it the owner of every molecule, formulation or therapeutic use identified by a research team. Patent ownership is determined through inventorship, employment rules, assignments, collaboration agreements and the law governing the particular invention. Data licences and material transfer agreements can add further rights and restrictions.

That means a research partnership should separate at least four questions that are often collapsed into the word “ownership.” Who may use the AI system? Who owns pre existing data, code and biological materials? Who owns new patentable inventions and know how produced during the project? Who receives rights to develop, manufacture, license or sell the resulting medicine? If those questions are left to a general clause saying that each party retains “its intellectual property,” a successful discovery can expose ambiguity at precisely the moment when investors and regulators require certainty.

East African patent law still starts with human inventors

Kenya, Rwanda, Tanzania and Uganda are contracting states under the ARIPO Harare Protocol, which provides a regional filing route for patents in designated states. The current ARIPO system requires patent applications to identify the invention through a description, claims and other formal materials and applies the familiar standards of novelty, inventive step and industrial applicability. It does not create a special owner for inventions generated with artificial intelligence.

National law reinforces the human centred starting point. Kenya’s Industrial Property Act states that the right to a patent belongs to the inventor, subject to assignment and other rules. Uganda’s Industrial Property Act defines an inventor as the person who actually devises the invention, gives the right to a patent to the inventor and contains rules for inventions made in employment or commission. WIPO’s current AI and IP materials similarly distinguish human created inventions, which can be protected through existing frameworks if they qualify, from machine created inventions, where international debate continues. For an East African research company, the safer assumption is therefore not that “the AI owns it” or “the user owns it,” but that the project must identify the human inventive contribution and then trace ownership from those inventors through employment, assignment and collaboration arrangements.

A patent and a medicine approval answer different questions

A patent can prevent unauthorised commercial exploitation of an invention for a limited period. It does not certify that the product is safe, effective or of appropriate quality for patients. A company may hold a strong patent and still be unable to market the medicine. It may also develop a medicine that can be lawfully supplied even where no patent is obtained, provided regulatory requirements are met.

The East African Community Medicines Regulatory Harmonization programme has developed common guidelines and joint assessment procedures intended to reduce duplication in medicine evaluation. The EAC publishes guidelines for registration of pharmaceutical products, vaccines, biotherapeutics, variations and abridged assessment. The regional process supports scientific review, but national regulatory authority remains central to market authorisation and local oversight. WHO’s Good Reliance Practices likewise encourage regulators to use trusted assessments efficiently while retaining national responsibilities such as vigilance, market surveillance and oversight of local manufacturing and distribution. An AI generated prediction therefore has evidential value only to the extent that it contributes to the scientifically required development record. It does not bypass the medicine regulator.

Public funding creates leverage only if the access terms are written

The most important access question often arises long before a patent is granted. An East African university, public research institute or ministry may contribute laboratories, clinical networks, public data, researchers or grant money to a discovery partnership. If the agreement says nothing about affordable access, the fact that public resources supported the work may not automatically produce a legal right to a particular price or licence later. The exact result depends on the funding law, contract and applicable patent rules.

That is why access obligations are stronger when negotiated at the start. WHO’s 2024 briefing note on access oriented technology licensing explains that licences can deal not only with patents but also trade secrets, technical assistance and regulatory obligations, and can be structured to promote wider access to resulting health products. In an East African project, an access clause could address territorial licences, non exclusive rights for specified markets, supply commitments, reasonable pricing principles, technology transfer, local manufacturing, development milestones or rights that activate if the commercial partner stops developing the product. The correct package will depend on the project, but the principle is simple: public value should not be left to an informal expectation when it can be expressed as a contractual obligation.

Research agreements should anticipate the moment the AI result becomes valuable

A well drafted AI drug discovery agreement should decide how results move from research to development. If one party identifies a candidate through a model and another validates it in the laboratory, the agreement should state how inventorship will be assessed and who controls patent filing. If a university researcher and company scientist jointly contribute to the inventive concept, joint inventorship may follow under applicable law even if one institution funded most of the work. Ownership can then be allocated by employment rules or assignment, but inventorship itself should not be rewritten merely for commercial convenience.

The agreement should also address confidentiality and publication. Academic institutions may need to publish; a company may need time to file a patent before disclosure. Data used to train or query the system may carry contractual, privacy, database, genetic resource or research ethics restrictions. If biological samples or genetic resources are involved, national access and benefit sharing requirements may also matter. AI does not remove these older legal layers. It makes it easier for several of them to converge in one project.

Access is easier to design before monopoly and sunk costs harden

Once a candidate has passed expensive development stages, the negotiating positions change. A company that has financed trials, manufacturing scale up and regulatory submissions will understandably seek to recover investment. A public institution that contributed early science may then find that broad access requests conflict with an exclusive licence signed years earlier. Emergency tools such as compulsory licensing under national law and the flexibilities recognised by the WTO TRIPS framework remain important, but they are not a substitute for thoughtful deal design at the research stage.

East Africa can therefore treat AI assisted medicine discovery as a contracting and governance opportunity rather than waiting for a future patent dispute. Public funders can require an access plan. Universities can create standard licensing principles for health technologies. Companies can define foreground IP, inventorship procedures, development milestones and commercial rights before major value is created. Regulators can make clear that AI assisted evidence is welcome where scientifically valid but does not lower standards of quality, safety and efficacy.

The practical legal question is who controls the path to the patient

The deepest mistake would be to treat “the discovery” as one thing. An AI model may contribute to a prediction; researchers may design and test a candidate; inventors may obtain patent rights; a company may acquire those rights; a manufacturer may need know how; regulators must assess the medicine; and health systems must still be able to procure it. Ownership at one stage does not automatically control every stage.

For East African businesses and research institutions, the practical question is therefore not simply who owns an AI discovered drug. It is who can lawfully move the candidate from model output to patient access, under what rights, and with what obligations. AlphaFold 3 shows why that question is no longer theoretical. The legal work should begin when the research partnership is formed, because by the time a medicine succeeds, the most important ownership and access choices may already have been made.

Source note. This article is based on Josh Abramson and others, “Accurate structure prediction of biomolecular interactions with AlphaFold 3,” Nature (8 May 2024); the East African Community Medicines Regulatory Harmonization programme and its current medicines evaluation and registration guidelines; the World Health Organization, Good Reliance Practices in the Regulation of Medical Products, WHO Technical Report Series 1033, Annex 10 (2021), and Briefing Note on Access-Oriented Technology Licensing (2024/2025); the ARIPO Harare Protocol on Patents, Utility Models and Industrial Designs and ARIPO patent filing guidance; Kenya’s Industrial Property Act; Uganda’s Industrial Property Act; WIPO materials on AI, inventorship and intellectual property; and the WTO framework on public health and intellectual property, including the TRIPS Agreement and the Doha Declaration on TRIPS and Public Health.

Suggested citation: 

Ronald Serwanga, “AI Drug Discovery East Africa: Ownership and Access” East Africa Legal Insight (5 September 2026).