Injection drug users (IDU), crack smokers, and commercial sex workers engage in illegal activities that place them at risk for HIV infection. The marginalized nature of these groups often limits use of customary sampling methods to assess HIV prevalence. We tested urine samples of recent arrestees to evaluate HIV prevalence of at-risk populations that are difficult to access using standard surveillance methods. We tested for HIV-1 antibodies in urine specimens of recent Los Angeles County (California, U.S.A.) arrestees as part of the Drug Use Forecasting (DUF) Program funded by the U. S. National Institute of Justice. Data are presented for 5 years of a serial cross-sectional study of arrestees. Results from 1991 through 1995 indicate a slight HIV prevalence increase among crack smokers (from 4% to 6%). Prevalence estimates were relatively stable for IDU (6%), male (3%) and female arrestees (3%), arrestees who share needles (9%), and commercial sex workers (6%). HIV status was independently associated with injection drug use, crack smoking, and ever having exchanged sex for money or drugs. Prevalence of HIV among arrestee subgroups may reflect prevalence in the community. However the benefit of using the DUF sample must be weighed against bias introduced from using nonrandom samples to estimate prevalence.