When senior executives were asked to identify appropriate controls for five AI risks, only 12 percent, on average, answered correctly