In breast or ovarian cancer (BC/OC) patients with evocative personal and/or family history, multigene panel sequencing is performed on blood to diagnose hereditary predispositions. Additionally, BRCA1/BRCA2 testing can be performed on tumor sample for therapeutic purpose. The accuracy of multigene panel tumor analysis on BC/OC to detect predisposing germline pathogenic variants (gPV) has not been precisely assessed. By comparing sequencing data from blood and fresh-frozen tumor we show that tumor genomic instability causes pitfalls to consider when performing tumor testing to detect gPV. Even if loss of heterozygosity increases germline signal in most cases, somatic copy number variants (CNV) can mask germline CNV and collapse point gPV variant allele frequency (VAF). Moreover, VAF does not allow an accurate distinction between germline and somatic pathogenic variants.