In the unfortunate case that the sequence appears more than once in the genome, together with a PAM, Cas9 might cut in the wrong place, which generates an off-target hit (= a cut in a location where it is not supposed to be).
When CRISPR/Cas9 gene editing was first introduced back in 2012 there was a lot of concern about off targets with CRISPR/Cas9 and indeed the dual nickase is a novel strategy. But there have been several reports now showing that off targets are actually quite rare. For example see this very recent article: No unexpected CRISPR-Cas9 off-target activity revealed by trio sequencing of gene-edited mice.