20 Mishaps That Might Have Started Accidental Nuclear War
by Alan F. Phillips, M.D. [This version revision April, 2002]

Ever since the two adversaries in the Cold War, the U.S.A. and the U.S.S.R., realized that their nuclear arsenals were sufficient to do disastrous damage to both countries at short notice, the leaders and the military commanders have thought about the possibility of a nuclear war starting without their intention or as a result of a false alarm. Increasingly elaborate accessories have been incorporated in nuclear weapons and their delivery systems to minimize the risk of unauthorized or accidental launch or detonation. A most innovative action was the establishment of the "hot line" between Washington and Moscow in 1963 to reduce the risk of misunderstanding between the supreme commanders.

Despite all precautions, the possibility of an inadvertent war due to an unpredicted sequence of events remained as a deadly threat to both countries and to the world. That is the reason I am prepared to spend the rest of my life working for abolition of nuclear weapons.

One way a war could start is a false alarm via one of the warning systems, followed by an increased level of nuclear forces readiness while the validity of the information was being checked. This action would be detected by the other side, and they would take appropriate action; detection of the response would tend to confirm the original false alarm; and so on to disaster. A similar sequence could result from an accidental nuclear explosion anywhere. The risk of such a sequence developing would be increased if it happened during a period of increased international tension.

On the American side many "false alarms" and significant accidents have been listed, ranging from trivial to very serious, during the Cold War. Probably many remain unknown to the public and the research community because of individuals' desire to avoid blame and to maintain the good reputation of their unit or command. No doubt there have been as many mishaps on the Soviet side. One has been reported in which a Russian officer, on 23 September 1983, decided on his own initiative not to report an apparently grave warning on his computer screen, in the correct belief that it was a false warning. He may have saved the world, but was disgraced for failing to follow his orders; his career was ruined, and he suffered a mental breakdown.

Working with any new system, false alarms are more likely. The rising moon was misinterpreted as a missile attack during the early days of long-range radar. A fire at a broken gas pipeline was believed to be enemy jamming by laser of a satellite's infrared sensor when those sensors were first deployed.

The risks are illustrated by the following selections of mishaps. If the people involved had exercised less caution, or if some unfortunate coincidental event had occurred, escalation to nuclear war can easily be imagined. Details of some of the events differ in different sources: where there have been disagreements, I have chosen to quote those from the carefully researched book, "The Limits of Safety" by Scott D. Sagan. Sagan gives references to original sources in all instances.

These examples represent only a fraction of the false alarms that have been reported on the American side. Many on both sides probably remain unreported, or are hidden in records that remain classified.

1956, Nov.5: Suez Crisis coincidence.

British and French Forces were attacking Egypt at the Suez Canal. The Soviet Government had suggested to the U.S. that they combine forces to stop this by a joint military action, and had warned the British and French governments that (non-nuclear) rocket attacks on London and Paris were being considered. That night NORAD HQ received messages that: (i) unidentified aircraft were flying over Turkey and the Turkish air force was on alert (ii) 100 Soviet MIG-15's were flying over Syria (iii) a British Canberra bomber had been shot down over Syria (iv) the Soviet fleet was moving through the Dardanelles. It is reported that in the U.S.A. General Goodpaster himself was concerned that these events might trigger the NATO operations plan for nuclear strikes against the U.S.S.R.

The four reports were all shown afterwards to have innocent explanations. They were due, respectively, to: (i) a flight of swans (ii) a routine air force escort (much smaller than the number reported) for the president of Syria, who was returning from a visit to Moscow (iii) the Canberra bomber was forced down by mechanical problems (iv) the Soviet fleet was engaged in scheduled routine exercises.

1961, Nov.24: BMEWS communication failure.

On the night of 24 November 1961, all communication links went dead between SAC HQ and NORAD. The communication loss cut off SAC HQ from the three Ballistic Missile Early Warning Sites (BMEWS) at Thule (Greenland,) Clear (Alaska,) and Fylingdales (England,). There were two possible explanations facing SAC HQ: either enemy action, or the coincidental failure of all the communication systems, which had redundant and ostensibly independent routes, including commercial telephone circuits. All SAC bases in the United States were therefore alerted, and B-52 bomber crews started their engines, with instructions not to to take off without further orders. Radio communication was established with an orbiting B-52 on airborne alert, near Thule. It contacted the BMEWS stations by radio and could report that no attack had taken place.

The reason for the "coincidental" failure was that the redundant routes for telephone and telegraph between NORAD and SAC HQ all ran through one relay station in Colorado. At that relay station a motor had overheated and caused interruption of all the lines.

[NOTE: Long after I wrote this, a reader informed me that he was a technician at Plattsburgh Air Force Base at the time. The order reached that Base as an "Alpha" alert, the highest level, at which nuclear-armed bombers were to fly direct to their targets and bomb, without waiting at the failsafe point for further orders. Before any bomber could take off the correction arrived making it a third-level "Cocoa" alert, at which the bombers stayed on the runway with engines running and waited for further orders. If even one bomber had taken off, it might have been very difficult to recall it or stop it.]


1962, Aug.23: B-52 Navigation Error.

SAC Chrome Dome airborne alert route included a leg from the northern tip of Ellesmere Island, SW across the Arctic Ocean to Barter Island, Alaska. On 23 August 1962, a B-52 nuclear armed bomber crew made a navigational error and flew a course 20 degrees too far towards the north. They approached within 300 miles of Soviet airspace near Wrangel island, where there was believed to be an interceptor base with aircraft having an operational radius of 400 miles.

Because of the risk of repetition of such an error, in this northern area where other checks on navigation are difficult to obtain, it was decided to fly a less provocative route in the future. However, the necessary orders had not been given by the time of the Cuban missile crisis in October, so throughout that crisis the same northern route was being flown 24 hours a day.

Aug.-Oct.1962: U2 flights into Soviet airspace.

U2 high altitude reconnaissance flights from Alaska occasionally strayed unintentionally into Soviet airspace. One such episode occurred in August 1962. During the Cuban missile crisis on October of 1962, the U2 pilots were ordered not to fly within 100 miles of Soviet airspace.

On the night of 26 October, for a reason irrelevant to the crisis, a U2 pilot was ordered to fly a new route, over the north pole, where positional checks on navigation were by sextant only. That night the aurora prevented good sextant readings and the plane strayed over the Chukotski Peninsula. Soviet MIG interceptors took off with orders to shoot down the U2. The pilot contacted his U.S. command post and was ordered to fly due east towards Alaska. He ran out of fuel while still over Siberia. In response to his S.O.S., U.S. F102-A fighters were launched to escort him on his glide to Alaska, with orders to prevent the MIG's from entering U.S. airspace. The U.S. interceptor aircraft were armed with nuclear missiles. These could have been used by any one of the F102-A pilots at his own discretion.

1962, Oct.24: Russian satellite explodes.

On 24 October a Soviet satellite entered its own parking orbit, and shortly afterward exploded. Sir Bernard Lovell, director of the Jodrell Bank observatory wrote in 1968: "the explosion of a Russian spacecraft in orbit during the Cuban missile crisis... led the U.S. to believe that the USSR was launching a massive ICBM attack." The NORAD Command Post logs of the dates in question remain classified, possibly to conceal reaction to the event. Its occurrence is recorded, and U.S. space tracking stations were informed on 31 October of debris resulting from the breakup of "62 BETA IOTA."

1962, Oct.25: Duluth intruder.

At around midnight on 25 October, a guard at the Duluth Sector Direction Center saw a figure climbing the security fence. He shot at it, and activated the "sabotage alarm". This automatically set off sabotage alarms at all bases in the area. At Volk Field, Wisconsin, the alarm was wrongly wired, and the Klaxon sounded which ordered nuclear armed F-106A interceptors to take off. The pilots knew there would be no practice alert drills while DEFCON 3 was in force, and they believed World War III had started.

Immediate communication with Duluth showed there was an error. By this time aircraft were starting down the runway. A car raced from command centre and successfully signalled the aircraft to stop.

The original intruder was a bear.

1962, Oct.26: ICBM Test Launch.

At Vandenburg Air Force Base, California, there was a program of routine ICBM test flights. When DEFCON 3 was ordered all the ICBM's were fitted with nuclear warheads except one Titan missile that was scheduled for a test launch later that week. That one was launched for its test, without further orders from Washington, at 4 a.m. on the 26th.

It must be assumed that Russian observers were monitoring U.S. missile activities as closely as U.S. observers were monitoring Russian and Cuban activities. They would have known of the general changeover to nuclear warheads, but not that this was only a test launch.

1962, Oct.26: Unannounced Titan missile launch.

During the Cuba crisis, some radar warning stations that were under construction and near completion were brought into full operation as fast as possible. The planned overlap of coverage was thus not always available.

A normal test launch of a Titan-II ICBM took place in the afternoon of 26 October, from Florida towards the South Pacific. It caused temporary concern at Moorestown Radar site until its course could be plotted and showed no predicted impact within the United States. It was not until after this event that the potential for a serious false alarm was realized, and orders were given that radar warning sites must be notified in advance of test launches, and the countdown be relayed to them.

1962, Oct.26: Malmstrom Air Force Base.

When DEFCON 2 was declared on 24 October, solid-fuel Minuteman-1 missiles at Malmstrom Air Force Base were being prepared for full deployment. The work was accelerated to ready the missiles for operation, without waiting for the normal handover procedures and safety checks. When one silo and the first missile were ready on 26 October no armed guards were available to cover transport from the normal separate storage, so the launch enabling equipment and codes were all placed in the silo. It was thus physically possible for a single operator to launch a fully armed missile at a SIOP target.

During the remaining period of the Crisis the several missiles at Malmstrom were repeatedly put on and off alert as errors and defects were found and corrected. Fortunately no combination of errors caused or threatened an unauthorized launch, but in the extreme tension of the period the danger can be well imagined.

October 1962: NATO Readiness.

It is recorded that early in the crisis, in order to avoid provocation of the U.S.S.R., British Prime Minister Harold Macmillan and NATO Supreme Commander General Lauris Norstad agreed not to put NATO on alert. When the U.S. Joint Chiefs of Staff ordered DEFCON 3, Norstad was authorized to use his discretion in complying, and Norstad did not order a NATO alert. However, several NATO subordinate commanders did order alerts to DEFCON 3 or equivalent levels of readiness at bases in West Germany, Italy, Turkey, and United Kingdom. This seems to have been largely due to the action of General Truman Landon, CINC U.S. Air Forces Europe, who had on his own initiative started alert procedures on 17 October in anticipation of a serious crisis over Cuba.

October 1962: British Alerts.

When the U.S. SAC went to DEFCON 2, on 24 October, the British Bomber Command was carrying out an unrelated readiness exercise. On 26 October, Air Marshall Cross, CINC of Bomber Command, decided to prolong the exercise because of the Cuba crisis, and later increased the alert status of British nuclear forces, so that they could launch in 15 minutes.

It seems likely that Soviet intelligence would perceive these moves as part of a coordinated plan in preparation for immediate war. They could not be expected to know that neither the British Minister of Defence nor Prime Minister Macmillan had authorized them.

It is disturbing to note how little was learned from these errors in Europe. McGeorge Bundy wrote in "Danger and Survival" (New York: Random House 1988), "the risk [of nuclear war] was small, given the prudence and unchallenged final control of the two leaders."

1962, Oct.28: Moorestown false alarm.

Just before 9 a.m. on 28 October the Moorestown, New Jersey, radar operators informed the national command post that a nuclear attack was under way. A test tape simulating a missile launch from Cuba was being run, and simultaneously a satellite came over the horizon. Operators became confused and reported by voice line to NORAD HQ that impact was expected 18 miles west of Tampa at 9:02 a.m. The whole of NORAD was alerted, but before irrevocable action had been taken it was reported that no detonation had taken place at the predicted time, and Moorestown operators reported the reason for the false alarm.

During the incident overlapping radars that should have been available to confirm or disagree, were not in operation. The radar post had not received routine information of satellite passage because the facility carrying out that task had been given other work for the duration of the crisis.

1962, Oct.28: False warning due to satellite sighting.

At 5:26 p.m. on 28 October, the Laredo radar warning site had just become operational. Operators misidentified a satellite in orbit as two possible missiles over Georgia and reported by voice line to NORAD HQ. NORAD was unable to identify that the warning came from the new station at Laredo and believed it to be from Moorestown, and therefore more reliable. Moorestown failed to intervene and contradict the false warning. By the time the CINC, NORAD had been informed, no impact had been reported and the warning was "given low credence."


1962, Nov.2: The Penkovsky False Warning.

In the fall of 1962, Colonel Oleg Penkovsky was working in Russia as a double agent for the CIA He had been given a code by which to warn the CIA if he was convinced that a Soviet attack on the United States was imminent. He was to call twice, one minute apart, and only blow into the receiver. Further information was then to be left at a "dead drop" in Moscow.

The pre-arranged code message was received by the CIA on 2 November 1962. It was not known at the CIA that Penkovsky had been arrested on 22 October. Penkovsky knew he was going to be executed. It is not known whether he had told the KGB the meaning of the code signal or only how it would be given, nor is it known exactly why or with what authorization the KGB staff used it. When another CIA agent checked the dead drop he was arrested.

1965, November: Power failure and faulty bomb alarms.

Special bomb alarms were installed near military facilities and near cities in the U.S.A., so that the locations of nuclear bursts would be transmitted before the expected communication failure. The alarm circuits were set up to display a red signal at command posts the instant that the flash of a nuclear detonation reached the sensor, and before the blast could put it out of action. Normally the display would show a green signal, and yellow if the sensor was not operating or was out of communication for any other reason.

During the commercial power failure in the NE United States in November 1965, displays from all the bomb alarms for the area should have shown yellow. In fact, two of them from different cities showed red because of circuit errors. The effect was consistent with the power failure being due to nuclear weapons explosions, and the Command Center of the Office of Emergency Planning went on full alert. Apparently the military did not.

1968, Jan.21: B-52 crash near Thule.

Communication between NORAD HQ and the BMEWS station at Thule had 3 elements: 1. Direct radio communication. 2. A "bomb alarm" as described above. 3. Radio Communication relayed by a B-52 bomber on airborne alert.

On 21 January 1968, a fire broke out in the B-52 bomber on airborne alert near Thule. The pilot prepared for an emergency landing at the base. However the situation deteriorated rapidly, and the crew had to bale out. There had been no time to communicate with SAC HQ, and the pilotless plane flew over the Thule base before crashing on the ice 7 miles miles offshore. Its fuel, and the high explosive component of its nuclear weapons exploded, but there was no nuclear detonation.

At that time, the "one point safe" condition of the nuclear weapons could not be guaranteed, and it is believed that a nuclear explosion could have resulted from accidental detonation of the high explosive trigger. Had there been a nuclear detonation even at 7 miles distant, and certainly if one happened nearer the base, all three communication methods would have given an indication consistent with a successful nuclear attack on both the base and the B-52 bomber. The bomb alarm would have shown red, and the other two communication paths would have gone dead. It would hardly have been anticipated that the combination could have been caused by accident, particularly as the map of the routes for B-52 airborne flights approved by the President showed no flight near to Thule. The route had been apparently changed without informing the White House.

1973, Oct.24-25: False alarm during Middle East crisis.

On 24 October 1973, when the U.N. sponsored cease fire intended to end the Arab-Israeli war was in force, further fighting started between Egyptian and Israeli troops in the Sinai desert. U.S. intelligence reports and other sources suggested that the U.S.S.R. was planning to intervene to protect the Egyptians. President Nixon was in the throes of the Watergate episode and not available for a conference, so Kissinger and other U.S. officials ordered DEFCON 3. The consequent movements of aircraft and troops were of course observed by Soviet intelligence. The purpose of the alert was not to prepare for war, but to warn the U.S.S.R. not to intervene in the Sinai. However, if the following accident had not been promptly corrected then the Soviet command might have made a more dangerous interpretation.

On 25 October, while DEFCON 3 was in force, mechanics were repairing one of the Klaxons at Kinchole Air Force Base, Michigan, and accidentally activated the whole base alarm system. B-52 crews rushed to their aircraft and started the engines. The duty officer recognized the alarm was false and recalled the crews before any took off.

1979, Nov.9: Computer Exercise Tape.

At 8:50 a.m. on 9 November 1979, duty officers at 4 command centres (NORAD HQ, SAC Command Post, The Pentagon National Military Command Center, and the Alternate National Military Command Center) all saw on their displays a pattern showing a large number of Soviet Missiles in a full scale attack on the U.S.A. During the next 6 minutes emergency preparations for retaliation were made. A number of Air Force planes were launched, including the President's National Emergency Airborne Command Post, though without the President! The President had not been informed, perhaps because he could not be found.

With commendable speed, NORAD was able to contact PAVE PAWS early warning radar and learn that no missiles had been reported. Also, the sensors on the satellites were functioning that day and had detected no missiles. In only 6 minutes the threat assessment conference was terminated.

The reason for the false alarm was an exercise tape running on the computer system. U.S. Senator Charles Percy happened to be in NORAD HQ at the time and is reported to have said there was absolute panic. A question was asked in Congress. The General Accounting Office conducted an investigation, and an off-site testing facility was constructed so that test tapes did not in the future have to be run on a system that could be in military operation.

1980, June 3-6: Faulty Computer Chip.

The Warning displays at the Command Centers mentioned in the last episode included windows that normally showed:

0000 ICBMs detected
0000 SLBMs detected

At 2:25 a.m. on 3 June 1980, these displays started showing various numbers of missiles detected, represented by 2's in place of one or more 0's. Preparations for retaliation were instituted, including nuclear bomber crews starting their engines, launch of Pacific Command's Airborne Command Post, and readying of Minutemen missiles for launch. It was not difficult to assess that this was a false alarm because the numbers displayed were not rational.

While the cause of that false alarm was still being investigated 3 days later, the same thing happened and again preparations were made for retaliation. The cause was a single faulty chip that was failing in a random fashion. The basic design of the system was faulty, allowing this single failure to cause a deceptive display at several command posts.


The extreme boredom and isolation of missile launch crews on duty must contribute to occasional bizarre behaviour. An example is reported by Lloyd J.Dumas in Bulletin of the Atomic Scientists vol.36, #9, p.15 (1980) quoting Air Force Magazine of 17 Nov. 71. As a practical joke, a silo crew recorded a launch message and played it when their relief came on duty. The new crew heard with consternation what appeared to be a valid launch message. They would not of course have been able to effect an actual launch under normal conditions, without proper confirmation from outside the silo.

Launch on Warning

There are still thousands of nuclear weapons deployed. At the time of writing (December 2001) Russia and the U.S.A. still have the policy of "Launch on Warning": that is to say, they plan to launch a salvo of nuclear-armed rockets if the warning systems show that a missile attack appears to be on the way. The retaliatory salvo would be launched before any of the incoming missiles arrived, so it could be launched as a result of a false warning. Thus a nuclear war could start for no military or political reason whatever.

The following event could have caused the final disaster if, for example, the rocket guidance system or the radar tracking had malfunctioned:

1995, Jan.25: Norwegian Rocket Incident.

On 25 January, 1995, the Russian early warning radars detected an unexpected missile launch near Spitzbergen. The estimated flight time to Moscow was 5 minutes. The Russian Defence Minister and the Chief of Staff were informed. The early warning and the control and command systems switched to combat mode. President Yeltsin was handed the "nuclear suitcase" ready to give the launch signal.

Within 5 minutes, the radars determined that the missile's impact point would be outside the Russian borders. The missile was carrying instruments for scientific measurements. On 16 January Norway had notified 35 countries including Russia that the launch was planned. Information had apparently reached the Russian Defence Ministry, but failed to reach the on-duty personnel of the early warning system. (See article in Scientific American, November 1997, by Bruce G. Blair, Harold A. Feiveson and Frank N. von Hippel.)

Comment and Note On Probability

The probability of actual progression to nuclear war on any one of the occasions listed may have been small, due to planned "fail-safe" features in the warning and launch systems, and to responsible action by those in the chain of command when the failsafe features had failed. However, the accumulation of small probabilities of disaster from a long sequence of risks add up to serious danger. There is no way of telling what the actual level of risk was in these mishaps but if the chance of disaster in every one of the 20 incidents had been only 1 in 100, it is mathematical fact that the chance of surviving all 20 would have been 82%, i.e. about the same as the chance of surviving a single pull of the trigger at Russian roulette played with a 6 shooter. With a similar series of mishaps on the Soviet side: another pull of the trigger. If the risk in some of the events had been as high as 1 in 10, then the chance of surviving just seven such events would have been less than 50:50. [Note that there is no attempt here to calculate an actual probability. This is merely an example to illustrate the cumulative effect of any low-probability risk that is taken repeatedly, or accepted continuously, over a period of time.]


BMEWS: Ballistic Missile Early Warning Site
CIA: Central Intelligence Agency
CINC: Commander in Chief
DEFCON: Defence Readiness Condition (DEFCON 5 is the peacetime state; DEFCON 1 is a maximum war readiness).
HQ: Headquarters
ICBM: Intercontinental Ballistic Missile (land based)
KGB: Komitet Gosudarstvennoi Bezopaznosti (Soviet Secret Police and Intelligence)
NORAD: North American Aerospace Defense Command
PAVE PAWS: Precision Acquisition of Vehicle Entry Phased-Array Warning System
SAC: Strategic Air Command
SIOP: Single Integrated Operational Plan
SLBM: Submarine Launched Ballistic Missile

Principal Sources

Britten, Stewart: The Invisible Event, (London: Menard Press, 1983).
Calder, Nigel: Nuclear Nightmares, (London: British Broadcasting Corporation, 1979)
Peace Research Reviews, vol. ix: 4, 5 (1984);
vol. x: 3, 4 (1986) (Dundas, ON.: Peace Research Institute, Dundas).
Sagan, Scott D.: The Limits of Safety, (Princeton, N.J.: Princeton University Press, (1993).

Alan F. Phillips M.D., 11 January, 1997; revised April, 2002.