That is decidedly not the case. Even dictator is backed by a decent numbers of allies within his country, and he will need to get their agreement to actually do

That is decidedly not the case. Even dictator is backed by a decent numbers of allies within his country, and he will need to get their agreement to actually do the disarmament, even if he already agreed to outside parties. And in country like China (authoritarian but not quite dictator yet), the top guy is backed by a huge power pyramid, there are tens of thousands of politicians and strategists funnel their plans and ideas upward for discussions. The good news is that it goes both way too: even dictator can’t unilaterally decide to go ham and nukes everywhere. Sure, nuke your direct enemy is still an option, but anything more than that, or risks escalation more than that will probably be hard to be decided by just one or few guys. If you want to be optimistic, just realize that any time in the past where things might go wrong, actual human beings had stepped up to prevent fullscale nuclear war to happen. From Bayesian point of view, the odd is still on our side. I know that tactical nuke could be unfortunately escalated to larger scale, but then again, this needs people to agree to escalate in some sense. And there is still the obvious Murphy’s law of anything that can go wrong will go wrong. But we are trying to be optimistic, aren’t we?