Kenya Wants to Govern AI—and Invest in It. Can the Two Ambitions Meet?

Last Updated 1 month ago by Kenya Engineer

Two developments have brought Kenya’s artificial-intelligence ambitions into sharper focus.

The Ministry of Information, Communications and the Digital Economy recently concluded public consultation on the Draft Kenya Artificial Intelligence and Other Emerging Technologies Policy. The document seeks to guide the governance, development, deployment and use of AI while strengthening digital infrastructure, human capital, innovation, public trust and national economic resilience.

Separately, Nairobi Securities Exchange chief executive Frank Mwiti told Reuters that the exchange is developing what could become East Africa’s first AI-focused exchange-traded fund. The proposed product would give Kenyan investors exposure to a basket of international companies associated with AI.

The ETF remains under discussion with the Capital Markets Authority and could be delayed if concerns about inflated global AI valuations persist. It should therefore be treated as a proposal, not an approved investment product.

The policy and ETF perform different functions. One seeks to shape Kenya’s technological environment; the other would provide an investment channel. Yet placed side by side, they reveal whether the country’s AI strategy is connected to an investable domestic productive base.

The draft policy gets the direction largely right

The ministry describes the policy as a framework for a responsible, inclusive, secure and trusted AI ecosystem. Its emphasis on infrastructure, talent, innovation, sustainability and strategic autonomy reflects the main components required for meaningful AI development.

This is an important progression from treating AI purely as a software issue. Modern AI depends on extensive physical infrastructure: data centres, high-capacity communications networks, reliable electricity, cooling systems, cloud platforms, storage, cybersecurity and specialised processors.

Kenya therefore cannot become a major AI economy simply by increasing the number of people using generative tools. It must develop the engineering capacity required to build, host, secure and operate AI systems.

The policy’s final version should make this physical layer much more measurable.

For example, what national computing capacity does Kenya intend to develop? Will researchers and start-ups receive subsidised access to high-performance computing? What proportion of public-sector AI workloads should be hosted locally? How will energy and water requirements for data centres be assessed? Who will finance shared infrastructure?

Without targets, responsible institutions and budgets, infrastructure ambition risks remaining rhetorical.

Data governance is the second engineering foundation

AI systems depend on data, but Kenya’s most useful datasets are fragmented across ministries, counties, universities and private institutions. Many are incomplete, stored in incompatible formats or unavailable through secure interfaces.

The policy should establish practical standards for collecting, documenting, sharing and maintaining public-interest datasets. It should also distinguish between making data available and surrendering control over it.

Local AI systems require representative Kenyan data—languages, crops, diseases, traffic patterns, soils, weather, financial behaviour and industrial operations. Models trained primarily on foreign datasets may perform poorly when deployed in local contexts.

At the same time, poorly governed data-sharing can expose personal information, proprietary industrial data and sensitive government systems. Privacy protection, access controls, anonymisation and traceable authorisation must therefore be treated as technical architecture, not merely legal compliance.

The emerging national AI framework should align clearly with the Data Protection Act, the National Data Governance Policy, cloud-policy guidelines, cybersecurity requirements and sector regulations. Developers need to know which rules apply when an AI system crosses several areas at once.

Risk should be assessed by use, not fashionable terminology

The consequences of an AI system depend greatly on where and how it is used.

A writing assistant does not present the same level of risk as a system recommending medical treatment, controlling energy infrastructure, assessing loan applicants or identifying suspects. Regulation should therefore concentrate on applications capable of affecting safety, rights or access to essential services.

High-impact systems should be subjected to stronger requirements, potentially including documented training data, performance testing, cybersecurity evaluation, human oversight, incident reporting and independent audits.

For engineering applications, validation must extend beyond checking whether an algorithm produces generally accurate results. A predictive-maintenance system, for example, should be tested for false negatives that could allow equipment to fail. A traffic-management algorithm should be assessed for pedestrian safety and unusual road conditions. A clinical model should be evaluated on Kenyan patients and across demographic groups.

The policy should also make procurement agencies responsible for interrogating vendor claims. Public bodies should not acquire “AI-powered” systems without access to performance evidence, technical documentation and clear responsibility when the technology fails.

The NSE proposal broadens access—but mainly to foreign value

The proposed ETF would most likely be denominated in Kenyan shillings, reducing the inconvenience and part of the currency exposure associated with individually accessing overseas markets. Mwiti said the basket could reference companies such as Microsoft and firms including OpenAI and Anthropic.

Strictly speaking, not every frequently cited AI company is publicly listed. An ETF cannot ordinarily purchase shares in private companies such as OpenAI or Anthropic in the same straightforward way it buys listed equities. Its final structure would therefore need to be explained carefully.

The fund could hold listed semiconductor manufacturers, cloud providers, data-centre operators and software companies. Alternatively, it could track an external AI index or use other permissible financial instruments to obtain exposure.

Before approving the product, the CMA and NSE should require clarity on index methodology, rebalancing, management fees, custody, liquidity and the meaning of “direct exposure to AI.” Without a disciplined methodology, an AI label can become a marketing device attached to almost any large technology company.

The exchange has also acknowledged the possibility of an AI investment bubble. Concentrated thematic funds can rise rapidly when a sector is popular but fall sharply when expectations change. The proposed product will therefore require prominent investor education, particularly if it targets younger first-time investors.

The more difficult question is where Kenya participates

An AI ETF may help Kenyans share in the growth of global technology companies. It does not, by itself, finance Kenyan AI innovation.

If the underlying holdings are overwhelmingly foreign, local savings will gain exposure to overseas AI businesses while Kenyan start-ups continue struggling to obtain patient capital, computing resources and markets.

That is not an argument against the ETF. It is an argument for completing the other side of the equation.

Kenya needs investable local technology companies, venture and growth-capital mechanisms, procurement opportunities and a credible route through which successful firms can eventually raise capital publicly. The NSE could examine complementary products that channel funds into Kenyan and African digital infrastructure, data centres, clean power for computing, semiconductor-support services and listed technology enterprises.

The national policy should consequently address not only innovation grants and incubation but also commercial scale. Many African start-ups survive pilot programmes but fail when they need long-term customers, infrastructure and later-stage investment.

Government procurement can help create that market, provided contracts are competitive, technically sound and structured to avoid permanent dependence on one vendor.

Energy and sustainability cannot remain peripheral

AI’s expansion will increase demand for electricity, cooling and data-centre capacity. Kenya’s renewable-heavy power system presents a competitive advantage, particularly for companies seeking lower-carbon computing.

But the advantage must be quantified. Data centres require exceptional reliability, often supported by redundant grid connections, batteries and standby generation. Cooling can also consume significant electricity and water, depending on the technology and local climate.

The final policy should promote energy-efficiency indicators such as power usage effectiveness, encourage heat- and water-conscious cooling designs and require transparent reporting from very large computing facilities. Locating new centres should also take account of grid capacity, fibre routes, water stress and opportunities for renewable generation.

This creates a natural bridge between AI policy and engineering policy. Kenya’s AI aspirations will ultimately depend on electrical engineers, network engineers, data scientists, cybersecurity professionals, cooling specialists and infrastructure financiers working together.

From a policy of intention to a programme of execution

The draft policy provides Kenya with an opportunity to establish direction before AI becomes embedded invisibly across the economy. Its final test will be whether it allocates responsibility and creates measurable actions.

A practical implementation framework should identify lead agencies, timelines, budgets and indicators covering computing capacity, datasets, workforce development, research, safety testing, start-up finance and public procurement. It should also define how progress will be independently assessed.

The proposed NSE fund, meanwhile, is evidence that AI has moved beyond laboratories and technology departments into mainstream investment discussion. That is significant—but it must not obscure the imbalance between consuming AI, investing in foreign AI and building AI within Kenya.

Kenya should aim to do all three intelligently. It can regulate high-impact systems without paralysing innovation, provide investors with diversified global exposure and deliberately create conditions in which Kenyan AI and infrastructure companies become investable themselves.

The stronger national ambition is not merely to own units in an AI fund. It is to ensure that, in time, some of the value represented in such funds is being engineered, hosted and created in Kenya.