Imagine the world, like our Earth, maybe in late 2020th or 2030th. There is a country, that we will call Plutonia. Plutonia has a lot of nuclear and biological weapons, but a shrinking isolated economy. The government is authoritarian. To ensure its stability, the government explains that all problems in Plutonia are due to the enemies from the West that want to overthrow the government and conquer the country. The people in the government even may partially believe in it.
Because the economy is shrinking and modern technologies are not coming from the world outside Plutonia, the conventional Plutonian army is unable to win the real war. Thus, in all "gray zone" conflicts, where Plutonia used to won, now the Plutonian neighbors may try to get their territories back. The only way to not lose is to use (limitedly) nuclear weapons.
Somehow, this particular crisis is solved and is not turning into a full-scale nuclear war. However, the Plutonian government tasted the power of nuclear blackmailing. To avoid it, other countries plan to develop a system that would protect them from this threat. Being created, this system would leave PLutonia defenseless. The Plutonian government believes that then it will be easy to overthrow them. So, they now blackmail other countries to stop any research in sensitive areas.
There is a chance that eventually such a situation will be completely resolved peacefully. However, there is a significant chance of eventual catastrophe, especially if the missile launch will be more or less automatized and controlled by the decision of a very small number of people (maybe even one). Notice that such a fragile situation also creates higher risks for bioweapons catastrophe and for misaligned AGI (other countries may try to create AGI as soon as possible to avoid the threat of nuclear war; this rush may lead to neglecting safety protocols).
What is the probability of such a scenario? Roughly, it is the probability that there will be something like Plutonia times the probability of the catastrophe induced by Plutonia. The latter is very hard to estimate since such an event never happened in history. It should be higher than the probability of the Cold War turning into real WWIII - because the USA and USSR were comparable agents, able to rely on something but nuclear weapons. Plutonia will have nukes as its only power.
Toby Ord estimates that during the Caribbean crisis the chance of the actual nuclear war was an order of 50%. Hence, the total risk of eventually transforming the Cold War into WWIII was even higher. Basically, there were only two options - war and peaceful transformation, and the last was chosen. For the Plutonian government, the transformation option might be significantly harder, since they have almost nothing but nuclear weapons. Thus, I would say that the probability of Plutonia induced catastrophe can be higher than 50%.
What about the first probability - the probability of emergence of Plutonia? There are many options, some are more likely, some are less. In my opinion, Russia is seriously likely to turn into Plutonia in the next decade, and it was going in that direction last 20 years. The alternative would be a democratic transformation, and, looking at similar cases, I would estimate the chance less than 50%. The chance of a change of international politics/economical growth without democratic transformation I would estimate as very low, maybe less than 10% (looking at the current modus operandi). This gives the net likelihood to get Plutonia more than 40%, and combining with the previous, more than 20% probability of the catastrophe.
This is way bigger than was estimated here, and I described just one very particular scenario. It is very likely I am wrong and I strongly overestimated this probability. Thus, I am writing it here, hoping for help from the community. What would be the correct estimation of the probability? If it is indeed large, what can be done to decrease it?