Abstract: In many classical compact settings, entropy is upper semicontinuous, i.e., given a convergent sequence of invariant probability measures, the entropy of the limit measure is at least the limsup of the entropies of the sequence. There are only a few results of this type for non-compact cases since both mass and entropy can escape. In this talk I will describe how this can happen in the context of countable Markov shifts and give continuity results recently proved with Godofredo Iommi and Anibal Velozo.