This review article delineates how artificial intelligence approaches to medical diagnosis are vulnerable to adversarial attacks. The authors suggest that regulatory review of devices that employ machine learning include an assessment of resilience against such attacks.