{
  "id": 16085176,
  "title": "Ghana’s hospitals are adopting AI, but the machines were never trained on Ghanaian patients",
  "url": "https://urgent.news/2026/10/22/ghanas-hospitals-are-adopting-ai-but-the-machines-were-never-trained",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-22T17:41:37.000Z",
  "source": {
    "name": "MyJoyOnline Ghana",
    "slug": "myjoyonline-ghana",
    "url": "https://www.myjoyonline.com/ghanas-hospitals-are-adopting-ai-but-the-machines-were-never-trained-on-ghanaian-patients/"
  },
  "original_language": "en",
  "account": "Artificial intelligence is being introduced into Ghana's healthcare system, but the machines have not been trained on Ghanaian patients. This gap means that algorithms can make mistakes when applied to local patients, as they are based on reference populations from elsewhere. Despite this, AI tools are already being used in Ghanaian hospitals, such as computer-aided detection for tuberculosis screening and smartphone-based ultrasound support for midwives. However, the adoption of these tools has outrun the necessary governance. The people who are absorbing the risk are often those with the least ability to see it. The investigation revealed that there are no rules governing how AI enters a hospital, who validates it, or what happens when it makes a mistake. This leaves clinicians making judgments that should have been made beforehand. The questions that need answering include who approved the AI tool, who is accountable when it is wrong, and whose data is used to train the next version. While there are efforts from Ghanaian researchers and founders to build models using local data, this is a small effort against a global pipeline with significant resources. The solution lies in insistent procurement rules that ask what data a tool was trained on, a regulator that can validate performance locally, and data governance that informs Ghanaians about the use of their scans. Ghana's response cannot be to reject the technology, but to insist on these questions being asked.",
  "summary": "An algorithm that reads chest X-rays learns by looking at tens of thousands of them. It learns what a normal lung looks like, what tuberculosis looks like, what a tumour looks like, by being shown examples until the pattern sticks. The question nobody asks at the point of purchase is whose lungs it was shown. If the answer is Boston and Berlin and Seoul, and the lung now in front of it belongs to…",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}