The widespread dissemination of smart things, the penetration of robots and automatic mechanisms into almost all areas of human activity causes many people to have an anxious feeling and fear of a possible robotic uprising, more than once described in science fiction and shown in the movies. Leaders in the development of artificial intelligence, Google DeepMind and Open AI (founder – Elon Musk), presented a new look at training robots to provide greater human security and increased control over machines with intelligence. A new vision presented by the companies in a joint research article, indicates the need to oblige AI to consult with a person when solving a fundamentally new task. Such an approach can limit the independence of the activity of robots, whose rational decision can in any way bring discomfort or even harm to humans. The main problem in this case is the determination of the balance between rational and effective actions and the transfer of instructions to the machine. For example, the most effective way of cleaning a room, from the robot’s point of view, may be to completely remove things, furniture, and workarounds from the room, which doesn’t suit the owner of the room and things. For the first time, the principle “The robot cannot harm humans” was formulated by the brilliant science fiction writer Isaac Asimov back in 1942. In the work \However, the great science fiction showed that the literal adoption of these laws can also lead to unforeseen consequences. The film “I am a robot”, recently shot based on Azimov’s works, shows a situation where robots perceived the ban on harming humans as a command to completely restrict freedom of action and movement of people. To expand the ability to control AI decisions, it was proposed to provide an incentive for the machine to find the right one for a person response to resolve the problem. Thus, AI even taught to play elementary computer games. Despite the increase in time spent on training machines using the consulting method, the safety of using robots is increasing and control over artificial intelligence is increasing.