This talk presents two results on the interplay between causality and quantum information flow. First I will discuss about the task of switching the connections among quantum gates in a network. In ordinary quantum circuits, gates are connected in a fixed causal sequence. However, we can imagine a physical mechanism where the connections among gates are not fixed, but instead are controlled by the quantum state of a control system. Such a "quantum switch" mechanism is consistent with quantum theory but cannot be described with in the standard model of causally ordered circuits, where it would be equivalent to a deterministic time travel and hence would violate the causality principle. With respect to the standard circuit model, the quantum switch is a new primitive that enables new information-processing protocols, such as the perfect discrimination between two classes of channels that are not perfectly distinguishable in a single query by any ordinary quantum circuit. Second, I will discuss about the probabilistic simulation of impossible channels that take an input in the future and produce an output in the past. In this case, I will show that the maximum probability of success in such a simulation is determined by causality and is inversely proportional to the amount of information that the channel can transmit.