The evolutionary race between diseases and their hosts may lead to attenuation of the disease agent, increasing resistance in the host, or both. This is an undesirable outcome when the disease is being used as a biocontrol agent but a desired outcome when the host is valued by people. Introduced wild rabbits Oryctolagus cuniculus are a pest to agriculture and biodiversity values in New Zealand’s grasslands, particularly on the drier eastern sides of both islands. The costs to manage them using conventional control could not be sustained by landowners who since the 1980s have proposed the introduction of the viral biocontrol agents myxomatosis and then rabbit haemorrhagic disease virus (RHDV). Myxomatosis failed to establish but RHDV did establish and spread following its illegal introduction in 1997. However, since 1997, rabbit haemorrhagic disease (RHD) has become less effective for biocontrol of rabbits in New Zealand. Three lines of evidence from our four study sites in the Mackenzie Basin support this claim. First, the proportion of rabbits of all ages with antibodies to RHDV has increased in samples of rabbits shot each year since 1997. Taken alone this may simply reflect an accumulation in cross-sectional samples of seropositive older rabbits that have been exposed to infection but survived successive epizootics. Second, the proportion of young rabbits, sampled at an age when they have been exposed to a single epizootic event, that have antibodies to RHDV has also increased since 1997. This is strong evidence that something has changed in the rabbit–virus interaction. The cause of this effect remains unknown but is reflected in the third line of evidence, that the abundance of rabbits as indexed by standardised spotlight counts has increased since 1997. The rate of increase has, however, been much slower than that seen in the same populations as they recovered from conventional control before the arrival of RHD. Thus, we conclude that RHD is still an effective biocontrol but its efficacy is waning.