Genotoxicity models are extremely important to assess retroviral vector biosafety before gene therapy. We have developed an in utero model that demonstrates that hepatocellular carcinoma (HCC) development is restricted to mice receiving nonprimate (np) lentiviral vectors (LV) and does not occur when a primate (p) LV is used regardless of woodchuck post-translation regulatory element (WPRE) mutations to prevent truncated X gene expression. Analysis of 839 npLV and 244 pLV integrations in the liver genomes of vector-treated mice revealed clear differences between vector insertions in gene dense regions and highly expressed genes, suggestive of vector preference for insertion or clonal outgrowth. In npLV-associated clonal tumors, 56% of insertions occurred in oncogenes or genes associated with oncogenesis or tumor suppression and surprisingly, most genes examined (11/12) had reduced expression as compared with control livers and tumors. Two examples of vector-inserted genes were the Park 7 oncogene and Uvrag tumor suppressor gene. Both these genes and their known interactive partners had differential expression profiles. Interactive partners were assigned to networks specific to liver disease and HCC via ingenuity pathway analysis. The fetal mouse model not only exposes the genotoxic potential of vectors intended for gene therapy but can also reveal genes associated with liver oncogenesis.