What Are Acceptable Nuclear Risks?

When I read Eric Schlosser’s acclaimed 2013 book, Command and Control: Nuclear Weapons, the Damascus Accident, and the Illusion of Safety, I found a tantalizing revelation on pages 170-171, when it asked, “What was the ‘acceptable’ probability of an accidental nuclear explosion?” and then proceeded to describe a 1957 Sandia Report, “Acceptable Premature Probabilities for Nuclear Weapons,” which dealt with that question.

Unable to find the report online, I contacted Schlosser, who was kind enough to share it with me. (We owe him a debt of gratitude for obtaining it through a laborious Freedom of Information Act request.) The full report, Schlosser’s FOIA request, and my analysis of the report are now freely accessible on my Stanford web site. (The 1955 Army report, “Acceptable Military Risks from Accidental Detonation of Atomic Weapons,” on which this 1957 Sandia report builds, appears not to be available. If anyone knows of an existing copy, please post a comment.)

Using the same criterion as this report*, which, of course, is open to question, my analysis shows that nuclear terrorism would have to have a risk of at most 0.5% per year to be considered “acceptable.” In contrast, existing estimates are roughly 20 times higher.**

My analysis also shows, that using the report’s criterion*, the risk of a full-scale nuclear war would have to be on the order of 0.0005% per year, corresponding to a “time horizon” of 200,000 years. In contrast, my preliminary risk analysis of nuclear deterrence indicates that risk to be at least a factor 100 and possibly a factor of 1,000 times higher. Similarly, when I ask people how long they think we can go before nuclear deterrence fails and we destroy ourselves (assuming nothing changes, which hopefully it will), almost all people see 10 years as too short and 1,000 years as too long, leaving 100 years as the only “order of magnitude” estimate left, an estimate which is 2,000 times riskier than the report’s criterion would allow.

In short, the risks of catastrophes involving nuclear weapons currently appear to be far above any acceptable level. Isn’t it time we started paying more attention to those risks, and taking steps to reduce them?

Martin Hellman

* The report required that the expected number of deaths due to an accidental nuclear detonation should be no greater than the number of American deaths each year due to natural disasters, such as hurricanes, floods, and earthquakes.

** In the Nuclear Tipping Point video documentary Henry Kissinger says, “if nothing fundamental changes, then I would expect the use of nuclear weapons in some 10 year period is very possible” – equivalent to a risk of approximately 10% per year. Similarly, noted national security expert Dr. Richard Garwin testified to Congress that he estimate the risk to be in the range of 10-20 percent per year. A survey of national security experts by Senator Richard Lugar was also in the 10% per year range.