Roll-out of antiretroviral treatment (ART) raises concerns about the potential for unprotected sex if sexual activity increases with well-being, resulting in continued HIV spread. Beliefs about reduced risk for HIV transmission with ART may also influence behavior. From September 2003 to November 2004, 234 adults enrolled in a trial assessing the efficacy of modified directly observed therapy in improving adherence to ART. Unsafe sexual behavior (unprotected sex with an HIV-negative or unknown status partner) before starting ART and 12 months thereafter was compared. Participants were a mean 37.2 years (standard deviation [SD} = 7.9 years) and 64% (149/234) were female. Nearly half (107/225) were sexually active in the 12 months prior to ART, the majority (96/107) reporting one sexual partner. Unsafe sex was reported by half of those sexually active in the 12 months before ART (54/107), while after 12 months ART, this reduced to 28% (30/107). Unsafe sex was associated with nondisclosure of HIV status to partner; recent HIV diagnosis; not being married or cohabiting; stigma; depression and body mass index <18.5kg/m2. ART beliefs, adherence, and viral suppression were not associated with unsafe sex. After adjusting for gender and stigma, unsafe sex was 0.59 times less likely after 12 months ART than before initiation (95% confidence interval [CI] = 0.37-0.94; p = 0.026). In conclusion, although risky sexual behaviors had decreased, a considerable portion do not practice safe sex. Beliefs about ART's effect on transmission, viral load, and adherence appear not to influence sexual behavior but require long-term surveillance. Positive prevention interventions for those receiving ART must reinforce safer sex practices and partner disclosure.