Kenya’s new draft Artificial Intelligence and Emerging Technologies Policy is open for public consultation, and it is not the kind of light-touch, wait-and-see document most African governments have published on AI so far. It transcends Kenya’s borders, builds two new regulatory institutions from scratch, and tries to answer a question most AI policy on the continent still dodges: who actually pays for all this, and who gets protected from it?
Major Highlights
- Extraterritorial reach: the policy applies to any company outside Kenya, OpenAI and Meta included, whose AI systems produce outputs used in Kenya or have a direct, foreseeable effect on people there. Being headquartered abroad no longer puts a company out of reach.
- Mandatory disclosure: Kenyan companies will have to tell customers when they’re interacting with an AI system, giving individuals a formal right to know when an algorithm, not a person, is acting on them.
- Shared liability: developers, deployers, operators, vendors, and users will all carry legal responsibility for an AI system’s harms, with future legislation to spell out how liability, insurance, and redress get allocated across that chain.
- Two new institutions: the Kenya AI Safety Institute (KAISI) will handle model testing, red-teaming, benchmarking, and incident tracking, alongside a public AI and Emerging Technologies Registry and mandatory registration for high-risk AI applications.
- A compute lifeline for locals: the National AI Compute Access Programme is designed to stop local startups, researchers, and public institutions from being priced out of AI, including a push to reclassify GPUs and AI accelerators as tax-exempt educational equipment across the East African Community.
- The uncomfortable backdrop: Microsoft and G42’s roughly $1 billion Kenya data-centre project, announced back in May 2024, remains stalled over unresolved negotiations on government payment guarantees for capacity.
KINI BIG DEAL?
If you build, sell, or deploy AI touching Kenyan users, the compliance clock just started. Foreign AI providers, OpenAI and Meta included, now face real registration, disclosure, and liability requirements instead of operating on guidance alone, while Kenyan startups get a genuine lever to compete: cheaper GPU access through the new compute programme rather than having to out-fund Big Tech.
One catch worth watching: the policy assumes local compute infrastructure that doesn’t fully exist yet. Kenya’s own $1 billion Microsoft-G42 data centre has sat stalled for over a year on funding talks, so “accessible compute” is still a promise more than a delivered reality.
