Beware! Medical AI systems are easy targets for fraud and error

You can fake diagnoses with adversarial examples

Medical AI systems are particularly vulnerable to attacks and have been overlooked in security research, a new study suggests."