Introduction

In Kenya, the use of artificial intelligence (AI) is no longer a futuristic concept but one that is being discussed, utilised, and developed to significantly transform the healthcare sector. AI as a driver of change has sought to revolutionise the health sector¹ by filling the health gap and enhancing outcomes through small businesses and corporations that have come up with AI-based medical products.²

One of the key areas where AI is making a substantial impact is in telemedicine and mobile health (mHealth) applications. Mobile Health (mHealth) refers to the use of mobile devices to deliver health-related services and information. In Kenya and across other sub-Saharan African countries, mHealth has emerged as a critical tool in expanding access to healthcare, especially in cancer care. These applications have been increasingly utilised to provide cancer education, facilitate early screening, aid in diagnosis, and monitor patients during treatment. Despite its potential, less than half of the 48 countries in Sub-Saharan Africa have integrated mHealth into their overall patient care. Kenya, alongside Uganda, Tanzania, and Nigeria, stands out as one of the few nations leading this integration, demonstrating a growing commitment to leveraging digital solutions for improved health outcomes.³

Currently, Kenya is taking several steps to prioritise the health of its citizens by introducing new initiatives or enhancing existing ones to improve healthcare delivery. Of the 9,696 health facilities in Kenya under the Health Sector Report of 2021, the Ministry of Health directly manages 42.9% of them.⁴ There are a number of policies and guidelines that have been put in place to further ground the goal of achieving effective health care in Kenya. These include the Kenya Health Policy from 2014 to 2030,⁵ the Kenya Universal Health Coverage Policy 2020-2030,⁶ the Kenya Community Health Strategy 2020-2025⁷, and the Kenya National eHealth Policy, among others.⁸ To deal with these challenges, the government has implemented several key initiatives, which include the Linda Mama pro, launched in 2017, aimed at providing subsidised NHIF maternity services to all Kenyan women, specifically, it aimed at reducing infant and maternal mortality rates.⁹ The Health Insurance subsidy programme initiated in 2014, which subsidizes NHIF premiums for poor households,¹⁰ and the Afya Care program which was launched in four counties to provide insights for nationwide implementation, and Digital Health has been at the forefront of digital health innovation in Africa through initiatives like Mtiba, which is a mobile health wallet that saves for healthcare expenses.¹¹ AI has been incorporated in the health sector in Kenya as well to make a significant impact.¹² Currently, within the context of medical imaging, there is the use of AI algorithms in diagnostic and imaging analysis, which provides a more accurate diagnosis. This has been seen in the partnership between USAID, the Centre for Health Solutions, and the Tamatisha TB program, which launched a computer-aided programme for chest x-rays, a triage tool for pulmonary tuberculosis.¹³ Additionally, there is the use of AI and predictive analytics for patient risks and the use of AI in automating the triage and workflow of emergency care.¹⁴ However, WHO has observed with great caution, stating that healthcare providers need to be aware of the risks of using AI in low and middle-income countries, as overestimation of the large multi-modal models is capable of leading to dangerous misdiagnosis or inappropriate treatment decisions.¹⁵

This article seeks to assess how AI is currently being applied in Kenya’s health sector with a particular focus on its potential to address persistent challenges such as misdiagnosis. It further examines the associated risks, especially algorithmic bias and data representations that may inadvertently exacerbate existing health inequities if left unaddressed. The analysis also explores strategies to mitigate these risks. Lastly, the article compares Kenya’s broad, multi-sectoral National AI Strategy to Tanzania’s more targeted and health-specific AI Strategy in consideration of existing initiatives. This comparative lens is intended to highlight how Kenya might move from policy formulation to more tailored and impactful implementation within the healthcare space.

1. Current status of AI in the Kenyan Health Sector

In the recent MEDEXPO hosted by the Ministry of Health in collaboration with the county government of Nairobi, the Kenya Association of Pharmaceutical Industry and the Kenya Association of Private Hospitals, held between 14th and 16th May 2025, health experts called for the adoption of AI and advanced diagnostics, noting it as a game changer in the health care sector.¹⁶ In his speech, the Director of Health Products and Technologies at the Ministry of Health, Dr Tom Menge, emphasised the need to embrace AI diagnostics and cautioned practitioners not to lose sight of the ethical and equitable access to care. He compared Kenya to Iran, which has made remarkable strides in producing essential locally manufactured medicine for cancer treatment, biopharma, and nanotechnology.¹⁷ This was a comparison made out of concern given Kenya’s disjointed supply chain systems for drugs, which is critical for treatments such as cancer.¹⁸ AI as a transformative force in the healthcare system is being used to assist in diagnosing illnesses, assessing patients, anomaly detection, and even medical imaging, which provides faster and improved patient outcomes.¹⁹ Recently, in June 2025, Agha Khan launched an AI-powered clinic decision support platform that aims to tackle the existing health gaps with AI, such as delayed diagnosis and suboptimal treatments.²⁰ It provides real-time evidence that guides clinicians through a language interface available in Kiswahili and English, therefore making it easily accessible. It operated on grounded data sources such as the Electronic Health Record repository that had over 3.2 million patient records and health trends.²¹ Second, clinical practice guidelines that are tailored for low and middle-income settings bring it closer home, and lastly, it contains over 300 peer-reviewed publications on prevalent diseases in Kenya.²² Through harnessing generative AI, Afya Gemma aims to scale equitable, data-driven care across low-resource areas to ensure healthcare access to individuals in rural and underserved communities.²³ Additionally, several innovative startups are transforming healthcare delivery. The Pathological Network enhances diagnostic capacity in hospitals by ensuring the consistency and quality of laboratory results. Damu Sasa facilitates the rapid access to blood products for patients in need through a cloud-based, end-to-end blood services information management platform. Meanwhile, Jacaranda Health, in partnership with Google, is advancing ultrasound technology by developing a handheld device powered by artificial intelligence, eliminating the need for bulky traditional machines.²⁴ AI through a chatbot named Sophie Bot is used to ask questions about reproductive and sexual health services.²⁵

The current state of AI in Kenya’s healthcare sector is promising, marked by innovative applications in areas of diagnostics, reproductive health, medical imaging, and clinical decision-making. These advancements reflect a commitment to leverage AI for equitable, efficient, and accessible healthcare across the country. Nonetheless, there remain significant challenges.Addressing these issues will be crucial for ensuring that AI adoption in Kenya’s healthcare sector is both sustainable and inclusive.

2. Navigating biases

AI bias in healthcare springs up from two major areas: algorithmic bias and a bias in the data used to train healthcare tools. Algorithmic bias within this sector is the use of an algorithm in a way that increases disparities in health systems by compounding already-existing ones in socioeconomic position, ethnicity, ethnic background, religion, gender, or disability. The algorithm’s output becomes prejudiced due to false assumptions based on the data fed into it. According to Ferrara, bias is defined as a systematic error in decision-making processes leading to unfair outcomes, which is a critical concern²⁶. Ntoutsi defined algorithm bias as the inclination or prejudice of a decision made by an AI system which is for or against one person or group, especially in a way considered to be unfair.

Before one can navigate biases in artificial intelligence systems, it is important to acknowledge that such biases are already present and consequential in Kenya’s healthcare. AI tools are increasingly being used in the diagnostic support of diseases like malaria, tuberculosis, and cervical cancer. However, these tools often rely on trained data and are adopted from developed countries and therefore are created for their local demographics. Many low and middle-income countries face significant challenges in clinical diagnosis and guideline-based care. Limited resources, overstretched healthcare systems, and a shortage of healthcare professionals mean clinicians see a high volume of patients with a wide range of health complaints daily and make rapid diagnosis and treatment decisions, often with limited information.²⁷

To date, trials assessing LLM-based clinical decision support systems have largely been based on simulated cases and historical clinical interactions, not real patients in real time. Only three RCTs have rigorously evaluated AI-based clinical decision tools in Africa, and greater investment from donors and governments is required to build the evidence base at scale. Although medical AI is a relatively new field, it is important to acknowledge that the underlying disparities in healthcare that drive bias in medical AI are not recent developments. Rather, they are rooted in longstanding historical driving forces of inequality in health systems, which themselves reflect even longer-standing discrimination and other forms of structural oppression.²⁸

In order to reduce the amount of bias that exists within health systems, one of the measures that might be implemented is to guarantee that these systems have access to reliable and objective health data that is used for training purposes. In many cases, the data obtained, particularly inside Kenya, is just a reflection of a small majority, the majority of which resides in urban areas. Because of this, it is important to place an emphasis on the accuracy of the data. It will be possible for the algorithms to achieve their aim of improving health outcomes across the nation if they are able to give more accurate diagnoses through the use of systems that are based on correct and varied data.

3. Regulation and standard

Currently, the Guidance note on the processing of Health data provides robust protection for data processed in health institutions. It contains the principles that govern data protection, advocates for consent to ensure lawful basis processing, safeguards the rights of the data subject and lastly, data protection impact assessments.²⁹ Other applicable frameworks and policies include the Health Act, Health sector ICT standards and guidelines, Kenya Standards and Guidelines for m-health systems, Kenya National Patients’ Rights Charter, amongst others.³⁰ These frameworks and guidelines ensure the right to privacy, data protection and confidentiality of patients and in the process, address the issue of bias and discrimination in the processing of health data. The guidance note takes into consideration that with the presence of such risks and misuse of personal and health data, there should be amendments to the Data Protection Act to ensure full compliance and address privacy concerns as well.³¹
In the Kenya National AI strategy, the health sector has been accounted for in the majority of the sections. It starts with the aspiration to adopt AI technologies and lead in AI model innovation in the health sector amongst others in order to foster the growth of local AI ecosystems. It highlights how computer vision is being used for practical applications such as medical imaging analysis in the healthcare sector. In the current government’s bottom-up economy, key sectors like health have been prioritised.³² At the moment, the government has identified use cases within the AI spectrum of natural language processing that involve large language models which include a maternal chatbot in local dialect, so as to provide accurate pregnancy and childbirth information as well as an expanded disease advisory system which is building on already existing systems.³³ This inclusion shows that Kenya is taking great strides in ensuring the health sector evolves in the AI revolution as well.

In comparison with the state of AI development in Tanzania, a novel tool is being worked on for medical diagnosis and prognosis at the initial stages of the disease.³⁴ The main goal is to reduce health care costs and increase access to health care in a bid to reduce mortality. The tool is not aimed at replacing physicians but at helping them increase their efficiency. The brains behind it are using data from Bombo Hospital, based on which they created a data set with 9 selected attributes, which will be used to predict if an individual treated with HIV is classified to have a good or poor health status.³⁵ Other projects include applying convolutional neural networks for malaria diagnosis and the application of transfer learning on residual networks for diagnosing UTI, especially in rural areas, by leveraging convolutional neural networks, amongst others.³⁶

Tanzania is one step ahead by having a more targeted AI strategy for the health sector. This strategy focuses on establishing a Centre for Digital Health that facilitates digital health systems and ensures proper implementation in the health sector.³⁷ This will allow for different stakeholders in the health sector, together with innovators, clinicians, and digital health stakeholders, who will build these digital health innovations instead of designing stand-alone solutions. Additionally, the strategy points out challenges that AI in the health sector is currently facing, such as the absence of policies and regulations for AI development and implementation in Tanzania, poorly defined governance structures that bring together various stakeholders, and a lack of a specific approach towards AI in health, among others.³⁸ The strategy then shares 12 ways to ensure the successful implementation of AI in health in Tanzania.³⁹ It aims to have a shared vision led by the government and ensure the stakeholders are committed to this vision.⁴⁰ Second, it seeks to create a policy environment for AI by including AI in the national health policy to prioritise it within various health sector strategies.⁴¹ Third, in terms of leadership and coordination, AI should be an added agenda in the National Digital Steering Committee, the digital health secretariat, and other committees alongside a team that ensures AI implementation is mapped, tracked, and reported.⁴² Also, there is a need to include AI in the National Digital Health Strategy and the National Health Information Guidelines and design of the health sector to allow for resource mobilisation and coordination as well as to promote data privacy, protection, and secure data access.⁴³ An important strategy is to educate various stakeholders through training and capacity-building programmes.⁴⁴ This will ensure that they keep up with the emerging developments in AI, and these institutions should have curricula that advance healthcare providers with the understanding and use of AI health solutions that will aid in successfully deploying AI-enabled technologies that will meet the clients needs.⁴⁵ The implementation aims to align itself with recognised standards of safety, efficacy, and equity. In addition to that, it seeks to form well-developed professional ethics in the medical community for broader adherence by innovators, computer scientists, and other stakeholders, and come up with new ethical guidelines.⁴⁶ The government should also provide adequate incentives to encourage private sector and non-profit research to prioritise sufficient findings. Finally, the government should ensure sustainability, stakeholder engagement, and collaboration by ensuring an investment is made beyond the devices and data storage space, fostering a culture of collaboration, trust, and openness among AI stakeholders in the sector, and reviewing the already existing health policies so as to harmonise and support the implementation of AI.⁴⁷

This framework differs from that of Kenya in that, first, Kenya lacks a health-specific AI strategy for health but instead includes its inclusion features amongst other sectors in the National AI Strategy of Kenya. Additionally, there is no specific steering committee that oversees and ensures the successful implementation of AI in the health sector, leaving it to the various managements of the different health institutions. Some of these key features should be urgently borrowed if Kenya wants to ensure that, in the next five years, as per the timelines set in the Kenya National AI Strategy.

However, in our opinion, Kenya and Tanzania do not seem too far apart in ensuring the health sector is not left behind in the AI revolution. Given that health is one of the key sectors that ensures preservation of life, key attention must be given to remedies through AI that can address the new variants of diseases and viruses, and ensure better healthcare output

4. Policy and legal recommendations.

In order to minimise bias, Kenya should invest in robust regulatory policies with clear guidelines on how AI should be developed and tested. This includes addressing the lack of secure data sharing protocols in health institutions, which should be addressed in our data protection law, which will therefore allow for patient privacy even in an AI-driven sector. ⁴⁸ In our opinion, this could be seen in ensuring AI-specific data protection impact assessment.

Another crucial aspect of responsible and fair AI use is mitigating algorithm bias. This requires the utilisation of diverse and representative datasets. Establishing centralized health data repositories that gather data from various regions, socioeconomic backgrounds, and cultures within each country ensures that datasets are comprehensive and representative of the population. Collaboration with global health organizations and research institutions can offer access to diverse datasets, which can be integrated into AI training models for broader representation. Regular algorithm audits and bias testing are essential, including mandatory, periodic audits of AI algorithms to detect and address biases. These audits should involve testing algorithms on diverse datasets to ensure fairness. AI developers should publish detailed methodologies, including data sources and training processes, allowing independent verification of bias mitigation efforts.

Finally, creating stronger regulatory frameworks entails establishing clear guidelines and approval processes for AI technologies used in healthcare. It is important to develop and publish comprehensive guidelines that outline the requirements for approving AI technologies for healthcare use, focusing on safety, efficacy, ethical considerations, and patient privacy. The implementation of regulatory sandboxes, where AI technologies can be tested in a controlled environment, enables the evaluation of new technologies without putting patients at risk from unproven systems. Collaboration with international regulatory bodies is crucial, with a focus on adopting best practices from established regulatory authorities like the FDA (USA) and EMA (Europe), and customising them to fit the African context. The formation of regional consortia of African countries to standardise AI regulations can promote cross-border cooperation and consistency in AI healthcare standards.

It is crucial to develop centralised repositories for health data that compile diverse and representative datasets from across the continent. These repositories are essential for supporting accurate and unbiased AI models, which can lead to improved diagnostic and treatment outcomes.

Conclusion

AI is rapidly transforming Kenya’s healthcare sector, from diagnostic imaging and triage tools to mobile health platforms and maternal chatbots. These innovations promise to close gaps in care and improve outcomes, especially in low-resource settings. However, challenges such as algorithmic bias, unequal data representation, and privacy concerns persist, particularly when much of the training data originates from the Global North. Debiasing medical AI models will prove crucial in preventing the perpetuation and exacerbation of health disparities and ensuring all patients benefit equally from the future of medical AI.

To navigate this evolving landscape, Kenya must strengthen its regulatory frameworks, promote inclusive data practices, and ensure regular algorithm audits. Encouragingly, both Kenya and Tanzania are making strides in aligning AI with healthcare priorities, though stronger regional collaboration could accelerate progress. With deliberate, ethical action, Kenya can lead in creating an AI-powered health system that is not only innovative but also fair, accountable, and truly patient-centered.

https://itbrief.asia/uploads/story/2024/12/11/techday_26169f03bf0a929134f3.webp