There is a need to redefine the standard model of AI. In this model we aim to build machines able to reach the objectives stated. This can be done in several different ways, for example, by learning a policy that maximizes the expected discounted sum of rewards of the subject, who is a human. There is however a mistake in the standard model deriving from transferring to machines a definition of intelligence perfectly reasonable for humans. In fact, machines are not entitled to pursue their own objectives. A more sensible definition of AI would have machines pursuing our objectives. A new model is needed for AI development in which the machine’s uncertainty about the true objective leads to qualitatively new modes of behavior that are more robust, controllable, and deferential to humans.