Kenya will charge foreign manufacturers about KSh 3.2 million (US$2,500) to approve the highest-risk artificial intelligence medical software, according to a new framework published by the Pharmacy and Poisons Board.
- •By the regulator's definition, the fee applies to Category IV medical device software, the most dangerous classification in a four-tier risk system, which could cause death or serious irreversible harm if they malfunction.
- •The framework covers standalone AI diagnostic systems, cloud-based clinical decision tools, and software embedded in medical hardware; products that until now had no dedicated regulatory pathway in Kenya.
- •The guidelines further require a Software Bill of Materials, a component-level inventory of every element inside a medical software product, including open-source and third-party code, down to version and build numbers.
“A key feature of the framework is the distinction between standalone software and software embedded in hardware devices, a separation [] necessary to address risks associated with independent digital tools, particularly those driven by artificial intelligence and machine learning,” said Dr Ahmed Mohamed, Acting CEO of the Pharmacy and Poisons Board.
Foreign manufacturers seeking approval for Category I and II software, lower-risk tools that inform rather than drive clinical decisions, will pay US$500, while the same approval costs local manufacturers US$250. Category III software, covering high-risk applications that directly impact critical patient management decisions, is priced at US$2,000 for foreign manufacturers and US$1,000 for local ones.
An administrative fee of US$300 applies to foreign applicants across all categories, compared to US$100 for local manufacturers. All approvals will be valid for five years under the Health Products and Technologies Act.
The mandated fees are a fraction of what comparable approvals cost elsewhere. In the United States, FDA clearance for a medical device routinely runs into the tens of thousands of dollars, with complex AI submissions demanding significantly more. Meanwhile, European conformity assessments under the bloc's Medical Device Regulation can reach hundreds of thousands of euros.
The low price point sends a clear commercial signal that Kenya is positioning itself as an accessible regulatory entry point for health-tech companies eyeing the African market at a moment when global manufacturers are increasingly looking beyond saturated Western markets for growth.
Traceability, and Risk
“The guidance also tightens requirements on labelling and traceability. Software distributed via physical media must include clear instructions for use, while web-based or downloadable applications must display key information, including intended use and safety warnings, within the user interface. Developers will be expected to provide installation guidance and maintain robust version control systems to support traceability and corrective actions,” Dr. Mohamed said.
Companies seeking approval for AI-enabled medical devices must submit detailed documentation on training datasets, model architecture, bias management, and clinical validation metrics including sensitivity, specificity, and area-under-curve scores.
For AI systems with continuous learning capabilities (software that modifies its own behavior after deployment in hospitals and clinics), manufacturers must demonstrate anomaly detection mechanisms, rollback capabilities to prior validated states, and full traceability between real-world training data and clinical outputs. Moreover, annual post-market performance reports must be filed with the Board for the entire lifecycle of the product.
The framework anchors medical software regulation within Kenya's broader digital infrastructure agenda, drawing on the Digital Health Act of 2023, the National Cybersecurity Strategy running through 2027, and the Kenya AI Strategy extending to 2030.
Its risk classification system aligns with the International Medical Device Regulators Forum's global framework, and applications will be processed through the Board's existing online portal with a unique reference number assigned to each submission for lifecycle tracking.
