Automated work stands with the robots are widely used in the industry. The undisturbed work of the robot demands application of different types of safety systems. The classical safety systems, like fixed barriers, interlocked guards, light barriers, pressure sensing mats or laser scanners, have different disadvantages. In particular, they are sensitive for a random small objects appearing in the surrounding of a robot which do not cause a danger situation, or the workers can not approach to the robot, even in the sector far from the robot arm. In the present work the neuralsystem for safety control at the robot work stand, with the advanced recognition of danger situations, is presented. The system can reacts intelligently to the situation occurring in the robot’s surrounding and decides whether the situation is dangerous or not. When an object with the defined features appears in the safety field (SF) surrounding the robot, it is not stopped, only the arm is slowed down. When new object appears in SF near the arm, i.e. in the restricted field (RF), the robot is stopped. The system definesdynamically RF, according to the current position of the robot arm and takes into account information about new objects appearing in the neighborhood of the robot, in particular their real size and position.