An AI system may help interpret images or other health data. Its performance also depends on the data used to develop it: if some people are poorly represented, errors may be distributed unevenly.
This is why WHO highlights external validation: checking a system on data beyond those used during development. Its intended task must be clear, and its performance needs to be assessed in the setting where it will be used, with relevant populations and conditions of use.
A promising result therefore begins an evaluation rather than providing a universal guarantee. WHO also recommends human oversight and checks during use: clinical value requires safety, accountability and attention to differences between people.