Even with ‘big’ data about a given social network, it is very hard to predict what behaviors might emerge in the future. However this problem becomes even more challenging for clandestine networks in which adaptation to survive combined with limited hierarchical control can generate a highly active temporal evolution. In this talk I analyze three complementary examples of such dark temporal networks — online activity surrounding the civil unrest ‘Springs’ in Latin America during 2013-2014, offline networks generating anti-government violence during the ‘Troubles’ in Northern Ireland during the period 1970-1995, and the recent explosion of networks of networks in the dark pools associated with high-speed financial trading. In each case, I examine the effect that operating in the dark seems to have on the network dynamics, and discuss consequences for the prediction of future behaviors in — and threats from — such systems.