Rachel Dunscombe and Jane Rendall examine what needs to happen to make sure extremely important AI algorithms can be introduced within healthcare rapidly, and in a way that is clinically safe for populations.