If you've ever gotten into a discussion about nuclear weapons with a physicist, at some point it was probably remarked upon at some point that non-proliferation is hard, because at the end of the day, once a country has the requisite weapons-grade fuel, a few grad students could likely build one. As it turns out, this is true, and the US government has experimentally verified it.
I ran across this article in the Guardian that describes a classified project in the '60s, in which two physics PhDs with no knowledge of nuclear weapons were given only access to public libraries and made head researchers of a simulated weapons lab. In under two years, they had produced an (also simulated) working plutonium bomb that could be built in a machine shop. This is not especially surprising, really -- in reality, the biggest risk would be that they'd have an accident and die of radiation poisoning before they finished.
As an aside, doesn't this sound like a fascinating RPG? Manhattan Project RPG: "I redesign the firing pin to be 2 mm longer and have the trigger re-machined." The GM replies, "You didn't replace the beryllium housing. 1d4 technicians have perished in a neutron burst."