In short, the group tested the CRISPR/Cas9 genome engineering tools in non-viable human zygotes.
They discovered that their genome editing occurred at low efficiency, their intended target suffered repair from a endogenous sequence (a similar sequence on another part of the human genome) rather than using their donor oligos (synthetic DNA supplied to direct the intended mutation), and off-target mutation was observed. This is the first published article on genome editing in a model that is directly applicable to human health. This technology has been used in other organisms. The problems describe highlight the need to improve efficiency and specificity if this technology is ever going to have practical use.