Quiz: De-identification (and its limits) for genomics

Why can't genomic sequence data be fully de-identified using standard HIPAA Safe Harbor methods?