Since 1947, nuclear scientists have been setting a “doomsday clock” to alert the public to the current risk of nuclear war. The current state is 100 seconds to twelve, closer to twelve than ever before.
False alarms in early warning systems
The risk of an accidental nuclear war results primarily from early warning systems. These are based on sensors, very complex computer systems and networks for predicting and evaluating possible attacks by nuclear missiles. This can lead to false alarms, which can have very different causes (e.g. hardware, software, operating errors or incorrect evaluation of sensor signals). In peacetime and phases of political relaxation, the risks that the evaluation of an alarm message leads to a nuclear attack are very low. In such situations, false alarms are assumed in case of doubt.
Political crises – several events
The situation can change drastically if there are political crisis situations, possibly with mutual threats, or if further events occur in a temporal context with a false alarm. For this purpose, an evaluation searches for causes, i.e. attempts are made to find causal relationships. If such causal connections are found and are logically plausible, there is a great danger that these are assumed to be valid, i.e. that the alarm message is assumed to be valid, even if independent events coincide at random.
The risks can be aggravated by alerting chains. As a result of an alarm signal from an early warning system, armed forces can be put on alert. Such activities are recognized by the enemy and can lead to an increased readiness for alert in conflict situations. This in turn has repercussions on their own assessment of the situation. If a false alarm with regard to attacking nuclear missiles occurs in crisis situations with mutual threats and events that are classified as hostile, then a chain reaction with ever higher alarm levels can be set in motion within minutes, which gets out of control.
Error-free software is not feasible
Errors can never be excluded in a complex system and can be caused by both humans and computers. With complex applications, it is technically impossible to produce error-free software. Even if a software is proven to be correct with techniques of program verification, such proofs are only possible on the basis of a formal specification, which may itself contain errors. An important method to reduce errors in software development is testing. But testing an early warning system under real conditions will hardly be possible.
Rare errors are particularly dangerous
Furthermore, the risks of a nuclear war by accident are not reduced if there are fewer false alarms, because rare errors are difficult to assess and are taken more seriously. If there have been a lot of alarms in the past and they all turned out to be wrong, there is a high probability that the next alarm will also be classified as a false alarm. In particular, errors caused by unclear sensor data have occurred frequently in the past and will hardly lead to an incorrect assessment, at least in peacetime. Rare or unusual errors, on the other hand, are much more difficult to evaluate and therefore much more dangerous.
If it is possible to improve early warning systems in such a way that false alarms only occur very rarely, security will not be increased. The rarely occurring alarm messages are then unusual and difficult to interpret. This significantly increases the risk that they will be taken seriously, i.e. as valid. This is particularly true in crisis situations or when there are other events that can be related to them.
The risk of nuclear war will rise significantly
some reasons: Climate change, increasing cyber attacks, ever shorter advance warning times, more and more states with nuclear weapons, new arms race (see next point under “Quick introduction to the topic”)
Translated with www.DeepL.com/Translator