Martin Hellman is a professor at Stanford, one of the co-inventors of public-key cryptography, and the creator of NuclearRisks.org. He has recently published an excellent essay about the risks of failure of nuclear deterrence: Soaring, Cryptography and Nuclear Weapons. (also available on PDF)
I highly recommend that you read it, along with the other resources on NuclearRisks.org, and also subscribe to their newsletter (on the left on the frontpage).
Update: Here’s a Martin Hellman quote from a piece he wrote called Work on Technology, War & Peace:
You have a right to know the risk of locating a nuclear power plant near your home and to object if you feel that risk is too high. Similarly, you should have a right to know know the risk of relying on nuclear weapons for our national security and to object if you feel that risk is too high. But almost no effort has gone into estimating that risk. To remedy that lack of information, this effort urgently calls for in-depth studies of the risk associated with nuclear deterrence.
While this new project may seem to have a much more modest goal than Beyond War, there is tremendous hidden potential: My preliminary analysis indicates that the risk from relying on nuclear weapons is thousands of times greater than is prudent. If the results of the proposed studies are anywhere near my preliminary estimate, those studies then become merely the first step in a long-term process of risk reduction. Because many later steps in that process seem impossible from our current vantage point, it is better to leave them to be discovered as the process unfolds, thereby removing objections that the effort is not rooted in reality.