For the first time ever AI systems can directly harm patients. Are we doing enough to prevent a medical AI tragedy, the equivalent of a thalidomide event?