Underreporting of HIV/AIDS cases is a common problem in HIV epidemiology which often skews epidemiologic projections on which public health policy decisions are often based, especially in the cases of low HIV prevalence countries or in early phases of an emerging epidemic when the HIV incidence is still low, but might be growing rapidly. In this work, we propose a simple mathematical model with groups of known and undetected HIV-positives. Using this model with the annual HIV incidence data of new HIV cases and new AIDS cases detected at onset of symptoms, we are able to obtain an estimate for the number of undetected HIV-positives. Moreover, using Taiwan data of 1993-2000, we are able to predict the number of new cases in the next two years within 5% accuracy. We also give a approximate ratio of underreporting which approximates the magnitude of underreporting of HIV cases in low HIV prevalence settings. The procedure is illustrated with the HIV data of Taiwan and Cuba. The result shows that underreporting in Cuba is low, probably due to its intense contact tracing program. For Taiwan, the level of underreporting is higher, but has improved slightly since 1999. The method is useful as a simple tool to gauge the immediate impact of an emerging epidemic, as well as for the purpose of public health policy planning and short-term future projections.