Stable nuclear deterrence relies on the impossibility of one nation being able to successfully conduct a preemptive nuclear first strike on another. To this end, the United States has developed the capability to be able to respond while it is under attack, before the bulk of its nuclear forces, their command-and-control system, and even the president are destroyed.
But the intense time pressure involved in accurately diagnosing, and then ordering and executing a response to a nuclear attack, could easily backfire catastrophically. Specifically, the time pressure inherent in this launch-under-attack approach creates two unacceptable risks: that the president might fail to respond to a real attack, or that he or she might order a nuclear response to a false warning. To address these risks, the president should be able to order a delayed response—an option I call “decide under attack.”
The current command-and-control system is under stress
The United States fields three long-range nuclear strike capabilities, known as the triad. Intercontinental ballistic missiles (ICBMs) are the quickest-reacting leg, submarine-launched ballistic missiles (SLBMs) the most survivable, and bombers the most flexible leg. The triad is the foundation for strategic stability, because all three legs cannot be destroyed at once.
In the abstract world of academics, arms control negotiators, budgeteers, and politicians, the number and cost of these delivery vehicles and their warheads attracts most of the attention. But the challenges of actually controlling the machine, in the real and imperfect world, often gets less scrutiny. Yet, the president’s ability to gain and maintain situational awareness during a nuclear crisis, and to clearly direct an appropriate response under extraordinarily intense time pressure is vital. As such, the United States’ nuclear command and control, or NC2, system can be considered to be the triad’s essential nervous system—without which its legs could be paralyzed. The brain directing this nervous system is provided by the president, who has the sole authority to order the launch of nuclear weapons, or to rescind such an order.
Not least because the president could be a target, even a small-scale nuclear strike could threaten the U.S. command-and-control system. As a result, if an incoming nuclear attack is detected, time is of the essence. Yet, controlling nuclear forces during a nuclear attack involves a complex and interdependent set of decisions and actions. An enemy’s launch must be detected and accurately characterized. The president must be connected to the NC2 system, his identity verified, and connectivity maintained. He must be briefed on the adversary’s strategic intent, which may only become clear, if at all, gradually and imperfectly.
The president must then be briefed on options to counter the attack and on the consequences of these options, such as civilian casualties, fallout in neighboring nations, relations with allies, and the likely reactions of the attacking nation’s leadership, to name but a few. He may also wish to consult advisors, who may be hard to find.
The president must also consider the tricky issue of whether the attack is real, or whether the warning is the result of a system or procedural failure, or was generated by a malicious actor, perhaps even through a cyber attack. Internal failures have actually generated false warnings in the past. While many measures are in place to mitigate both risks, and the president would surely consider the prevailing geopolitical situation, it would be foolish to assume that an attack cannot be incorrectly reported.
If the president decides to direct an immediate nuclear response while enemy missiles are (believed to be) incoming, his or her orders must then be rapidly transmitted to the forces that will execute them. The pressure to get it right is amplified by the fact that such a decision cannot be reversed once missiles have been launched. The clock is running breathtakingly fast; there may be only a matter of minutes to get everything done. The pressure is further heightened by the reality that, if the president is killed before acting, getting an executable decision from a successor will be very difficult.
Responding to even a small-scale nuclear attack—let alone a large-scale one—is, therefore, an exceedingly difficult process, deeply enveloped in the fog of war. To make matters worse, any president will likely not possess a deep understanding of the system that will underpin the lowest probability but highest consequence decision anyone on the planet will ever make.
How the current system might fail
Because the NC2 system must perform perfectly under intense time pressure, it is a tempting target for an enemy seeking to upend strategic stability through a first strike.
Ultimately, two variables are in play. The first is whether the perceived attack is real, regardless of whether the president is certain that the information he or she is receiving is accurate. The second is whether the president orders an immediate response. There are thus four ways in which any given scenario could play out.
If the attack is not real, and the president is uncertain and does not launch, the outcome—as worrisome as it is—is merely a close call.
The opposite case occurs if the attack is real and the president overcomes any uncertainty and launches an immediate response. While nobody in their right mind wants a nuclear exchange, nobody wants to fail to respond to one either.
However, the other two cases are both disasters that ought to be preventable. Either the attack is real and the president does not launch due to uncertainty; or the attack is not real, but the president launches an unrecallable immediate response in spite of the uncertainty.
Unfortunately, intense time pressure and the fog of war, coupled with the mindset that the only two options are an immediate launch or doing nothing, renders these catastrophes more likely than someone who has never rehearsed the process may think.
How a decide-under-attack option would work
But what about providing an additional option for the president to decide-under-attack? Under this option, a response of SLBMs, bombers that took to the air right away, and any surviving ICBMs, would be delayed by a few hours—long enough to allow the order to be recalled if the attack were not real. In this case, since the NC2 system would still be intact, the president could use it to cancel the U.S. response while it was still possible to do so, that is, before the response had begun. If, however, the attack was real and the president did not survive, then the response would be executed.
The president would still retain the option of launching an immediate response if there were certainty regarding the attack. But in the event there was not, a valuable alternative would be available.
The decide-under-attack option would yield better outcomes in all four scenarios, including and especially the two of particular concern, principally by tempering the time crunch. If the attack was real but shrouded in uncertainty, the response would occur even if the president and the NC2 system were destroyed—and, importantly, an adversary would know this. But the U.S. president would also know that the response could be recalled if the attack was not real.
This option would not be foolproof. However, the only risk would be extremely remote: the highly unlikely combination of a false warning and a presidential launch order and either the loss of the entire NC2 system in peacetime or the untimely incapacitation of the president with no successor at hand. The odds of these three events occurring at exactly the wrong moment are miniscule, compared to what I assess to be, based on observing many exercises, the high likelihood of presidential uncertainty in the vital minutes that bound a nuclear decision, potentially leading to a tragically wrong choice.
There is, of course, another case in which the decide-under-attack approach would be useful. If the attack was real but the president and the NC2 system survived, he or she could decide to modify the response based on new information. Of course, there would be no guarantee that the president and NC2 system would survive, but the current approach allows no possibility of modifying a U.S. response that relies on immediate launches of ICBMs or SLBMs.
Some will argue that the decide-under-attack option is already provided by the bomber leg of the triad, because unlike the other two legs, pilots must fly them to the target and can be recalled. Perhaps, but this force is only a small portion of the United States’ nuclear strike capability, is rarely generated (that is, placed on alert), and its communications and survivability are not guaranteed. Others will suggest that this option eliminates the need for ICBMs, which would likely be destroyed in a first strike and unavailable for a delayed response. But ICBMs remain essential for providing redundancy in the event of a technical failure in the triad’s other delivery vehicles and warheads, and, because they can be launched more quickly than other delivery systems, provide the option of an immediate response.
Implementing the decide-under-attack option will require careful public diplomacy. For it to work best, the United States’ enemies must know it exists. It is also important to emphasize that, under this option, the decision would not be automatic—unlike the automated “Dead Hand” system Russia reportedly uses. The president would remain in control.
Given these stipulations, the decide-under-attack option has the potential to reduce the chances of a mistaken nuclear exchange and strengthen strategic stability. We should wring it out.
Admiral Winnefeld served as the commander of NORAD, and retired in 2015 as the ninth Vice Chairman of the Joint Chiefs of Staff. He now serves as a Distinguished Professor at Georgia Tech's Sam Nunn School of International Affairs.