On the night of the 26th of September 1983, the fate of the Earth hinged on a decision made by one man. Fortunately for us, Lieutenant Colonel Stanislav Petrov of the Soviet Air Defence Command Centre in Moscow made the right decision, and averted what could have become the greatest tragedy in the history of mankind. He chose to ignore the Soviet satellite surveillance system’s warning that five ICBMs had been launched by the USA in the direction of the Soviet Union. By doing so, Petrov saved an incalculable number of lives and the overall health of the planet. The false alarm was eventually traced to the satellite, which picked up the sun's reflection off the tops of clouds and mistook it for a missile launch. The computer program that was supposed to filter out such information had failed.
The Americans too had their own share of false alerts. On the 3rd of June 1980, the Early Warning System at the Strategic Air Command at Omaha, Nebraska, detected two Soviet SLBMs heading for the US mainland. Fortunately, those in charge at the North American Aerospace Defence Command (NORAD) had several reasons to doubt these warnings and it was eventually traced to a faulty IC chip in a computer that was part of a communications system.
On the 5th of October 1960, the Early Warning System at NORAD indicated that the USA was under attack by Soviet ICBMs with a certainty of 99.9%. It turned out to be a software failure at a radar installation in Greenland.
These incidents, where only human judgement stood between false alarms and nuclear catastrophe, give us an insight into the risks posed by the potential unreliability of computer systems used to control or detect nuclear weapons. This is a very relevant problem even in the 21st century. Due to the short warning times involved - measured at best in minutes - today's nuclear forces could not function without high-speed computers to automate the warning process, control communications, and, should it be deemed necessary, guide missiles to their targets. How reliable are the computers used in the command and control of nuclear weapons?
The standard of reliability required of military computer systems whose failure could precipitate a thermonuclear war must be higher than that of any other computer system, since the magnitude of possible disaster is so great.
Computer systems can fail because of incorrect or incomplete system specifications, hardware failure, hardware design errors, software coding errors, software design errors, and human error such as incorrect equipment operation or maintenance. Particularly with complex, normally highly reliable systems, a failure may be caused by some unusual combination of problems from several of these categories.
In hindsight, the blame for each of the above incidents can be assigned to individual component failures, faulty design, or specific human errors, as is almost always the case with such incidents. But the real culprit is simply the complexity of the systems, and our inability to anticipate and plan for all of the things that can go wrong.
There is clearly room for technical improvements in nuclear weapons computer systems. However, adding more and more such improvements cannot ensure that they will always function correctly. The problems are fundamental ones due to untestability, limits of human decision making during high tension and crisis, and our inability to think through all the things that might happen in a complex and unfamiliar situation. We must recognize the limits of technology. The threat of nuclear war is a political problem, and it is in the political, human realm that solutions must be sought.