A workshop was convened to discuss safety issues for traditional-approach HIV vaccines, especially inactivated vaccines. The topics included issues pertaining to (1) cell substrates used for production and (2) vaccine virus inactivation. The use of cell substrates such as tumor-derived continuous cell lines (TCLs) or virus-transformed. CLs may be the most feasible approach to provide commercial-scale virus yields. However, especially because of concerns about tumorigenicity, TCLs have not been used to produce preventive vaccines for human trials with healthy subjects in the United States. Residual TCL material (e.g., DNA, cellular proteins, viruses) may not be removed during purification of intact HIV virions to the same extent achievable for a recombinant protein. Manufacturing processes, e.g., physicochemical methods of destroying DNA, could decrease tumorigenicity risk. Methods to assess potential for tumorigenicity may need further development. Another potential substrate for viral production that merits further study is human peripheral blood mononuclear cells (PBMCs). Regardless of the cell substrate used, extensive testing for adventitious agents (including non-HIV retroviruses) is needed. Vaccine virus inactivation can be considered in statistical terms, i.e., the probability of a surviving infectious particle. One formula to determine a "safety margin" (SM) is reduction of titer in log10 for all inactivation steps minus initial viral infectivity in log10. Calculations for appropriate SMs should include all sources of variability (e.g., lot-to-lot differences). Ensuring a specified SM, as the lower bound of the 95% confidence interval, for production lots was discussed. Sensitivity and specificity of infectivity assays may present limitations.