Just when you thought nuclear holocaust had gone away...

Nuclear war accidents: Minutemen missiles in silos should be abolished. - Slate Magazine

Ros Rosenbaum argues that the risk of an accidental nuclear exchange is still high, particularly from the land based Minuteman missiles.   Computers are now a large part of the problem:

The silo-based missiles have been "detargeted," we've been told, but can easily be retargeted in seconds with a burst of code containing—say, Moscow's GPS coordinates. They've been detargeted but not de-alerted (something I called for back 2008 in Slate). They're still ready to go, at a moment's notice, vulnerable to hackers despite the claims by some security experts they're "air gapped from the internet"—but think USB sticks.

Nobody has yet explained the incident in late 2010 when 50 nuclear missiles in Wyoming stopped responding to the C3I system for a frightening period of time. The Pentagon hastened to say, hey, no problem, don’t worry that our computer system isn't capable of error-proof command and control of 50 nuclear missiles for just a little while. The episode was not confidence-inducing, and it reminded us how much world-destroying power is entrusted to glitch-prone computer architecture. The fact that this sensational story was virtually ignored is further evidence of head-in-the sand consciousness we have about nuclear power. 

When I discussed the matter of our "launch posture" with a Pentagon general who specialized in nuclear affairs, he refused to say whether these missiles could still be launched on the basis of "dual phenomenology." This obfuscatory euphemism means that once we receive signals from two electronic tracking systems—radar and satellite—that something on the screens looks like an attack, we'd face the "use it or lose it" choice. We'd have to "launch on warning" before we knew that the warnings were not "false positives" like the flock of geese. Fortunately, it's unlikely there could be two simultaneous false positives in our dual phenomenology, but a statistician's analysis has argued that eventually it would happen. One prominent statistics expert, Martin Hellman, one of the inventors of the "trap door" and "public key" methods of encryption for the Internet, calculated a 1-in-10 chance of a nuclear exchange in the next decade, and Scientific American put it at a not-so-reassuring 1-in-30 for the same period.