Great efforts are taken to ensure that weed biological control agents are highly host-specific prior to their release into a new country. The genetic stability of the agent over the long term is often questioned, especially in relation to possible shifts in host-range, but rarely scientifically tested. Here, we use the Puccinia chondrillina (rust fungus) and Chondrilla juncea (skeleton weed) pathosystem in Australia to explore such issues. Three forms of C. juncea exist in Australia (narrow, intermediate and broad leaf forms), with P. chondrillina isolates specific to the narrow and intermediate forms previously released as biological control agents. One isolate, IT32, released in 1971 to target the narrow-leaf form showed exceptional impact in controlling this form. Three isolates targeting the intermediate-leaf form were subsequently released: TU21 in 1980, recorded soon after as not having persisted; IT36 in 1982, recorded as having a moderate impact; and TU788 in 1996, for which the impact is unknown. DNA sequencing of these released isolates, using DNA extracted from stored spores, and of isolates collected on both narrow and intermediate leaf forms in 2007 and 2016 is being used to determine the lineage of the contemporary isolates and identify genetic changes that may have occurred over time. Concurrently, each contemporary isolate is being tested on the three C. juncea forms to confirm its pathogenicity phenotype. We hypothesise that P. chondrillina has changed little since its introduction into Australia and that this pathosystem remains stable thereby providing evidence for the safety of biological control programs using host-specific plant pathogens.