AI Is Diagnosing Patients in Africa Without Their Knowledge — Here's the Risk
Your health data might be checked by a machine you never hear about. Should that worry you?
Imagine going to a clinic in Tanzania with a persistent cough. A machine analyzes your chest X-ray and quietly flags it as a likely tuberculosis case. A nurse reads the result and starts treatment. You never learn that the actual diagnosis came from an artificial intelligence system, not a human doctor. That's exactly what's happening across sub-Saharan Africa, according to a new paper from researchers Percy Brown and Kweku Yamoah.
The paper describes three real deployments: tuberculosis detection in Tanzania, diabetic retinopathy and TB screening in Zambia, and a health chatbot in Ghana. These AI agents (computer programs that act like doctors) make or support decisions about your health without requiring a human in the loop. The problem? Patients aren't told an AI is involved. In many places, formal rules for AI in healthcare barely exist, and the usual safeguards like explainable AI or doctor-centered accountability don't protect the actual person affected: you, the patient.
The researchers argue this creates an "informed consent trap." You can't consent to something you don't know about. If the AI gets it wrong, who is accountable? The developer? The hospital? The algorithm? Right now, the answer is fuzzy. And in low-resource settings, patients often have little power to ask questions or challenge a diagnosis.
The paper suggests three practical rules: AI involvement must be clearly disclosed before you're treated; a human must always have the power to override the AI; and any explanation an AI gives has to be adapted to local languages and cultural context, not a generic English paragraph. These aren't radical ideas, but in regions where formal AI regulation is still being built, they set a much-needed minimum standard.
- AI systems are already making TB and eye-disease diagnoses in Tanzania, Zambia, and Ghana without clearly telling patients.
- Without disclosure, patients can't give meaningful informed consent — and no one is clearly accountable if the AI errs.
- The paper proposes three fixes: tell patients the AI is involved, let humans override it, and make explanations local and understandable.
Why It Matters
As AI quietly enters wait rooms everywhere, knowing when a machine decides your care protects your rights and your health.