Emergent viral diseases remain a critical obstacle to welfare across landscapes and species, encompassing humans, wildlife, and agriculture. Following a jump to a novel host, the severity of disease resulting from infection is a critical determinant of the overall emergent pathogen threat. Conventional wisdom posits that virulence, defined here as host mortality, attenuates to intermediate levels as a pathogen adapts to a novel host, but this is largely based on data from just one system, myxoma virus, which was intentionally introduced as a biocontrol agent in rabbits (Oryctolagus cuniculus) in mid-1900s Australia. In this study, we demonstrate that infectious hematopoietic necrosis virus (IHNV), which made a host jump from sockeye salmon (Oncorhynchus nerka, ancestral host) to rainbow trout (O. mykiss, novel host), has not conformed to classical theory. We quantified virulence in the ancestral and novel hosts using common garden in vivo experiments with 16 archival IHNV isolates collected from 1972-2017, which span the period from shortly after the host jump and the subsequent 45 years of host adaptation. These virus isolates also represent two distinct phylogenetic genogroups, each associated with either the ancestral or novel host. The experiments were replicated across two research facilities, two challenges dosages, and two temperatures. While isolates from the ancestral genogroup showed no temporal change in virulence in either host, isolates from the novel viral genogroup displayed a significant increase in virulence over time in the novel host. Some possible indication of a virus temperature adaption after the host jump was also present. Potential drivers of virulence evolution are discussed. This represents one of only a handful of systems in which the evolution of increased virulence has been empirically characterized after a host jump and subsequent adaptation. It contributes to a growing body of evidence that contradicts the classical case study of myxoma virus attenuation after adaptation.