The AI's erroneous identification of a person as a thief highlights risks of racial bias in AI security.