The article deals with the problems of preventing nuclear threats by electronic systems and humans, which have not yet been studied in socio-political science. The article discusses in more detail the issues of prevention of nuclear threats by electronic systems and humans: artificial intelligence and intelligence in preventing a nuclear catastrophe; unreliability in the early warning system; contextual knowledge; suggestions for overcoming doubts; about the possibility of greater security based on incomplete information of electronic systems; which of the surprise attack control systems is more reliable. The article notes that the presence of nuclear weapons even among equivalent powers is not a guarantee of a nuclear catastrophe. The guarantee of the possibility of a new nuclear strike as soon as possible is one of the decisive conditions for preventing nuclear aggression. The computer early warning system and electronic decision-making systems are aimed at timely detection of a nuclear attack.

In times of peace and during the period of defusing tensions between nuclear powers, the risk is relatively small that a computer signal of an attack immediately leads to a retaliatory nuclear strike. In such a situation, the role of a person in making a decision on a dubious nuclear attack signal increases dramatically. The situation changes dramatically in times of crisis exacerbations of relations between nuclear Powers, up to the expression of mutual threats.