A) In transgenic laboratory populations expressing Cas9 (brown) from an unlinked locus such as another chromosome, the sgRNA-only drive (green) will be copied in every generation. For clarity, copying is assumed to occur when haploid cells combine to form a diploid. In our S. cerevisiae experiments, Cas9 was encoded on an episomal plasmid with imperfect inheritance that should produce a similar pattern. (B) If escaped organisms encoding an sgRNA-only drive mate with wild-type organisms, the cas9 gene quickly segregates away from the sgRNA-only drive, precluding exponential spread. Any organisms that do encode Cas9 will still exhibit drive, but the total number of copies is limited by the number of escaped organisms and therefore is dwarfed by the wild-type population. If one organism is released from the laboratory for every million wild-type organisms in the population, a perfectly efficient drive with no fitness cost will linearly increase in relative abundance by 2E-6 per generation. This tiny inheritance advantage is exceedingly unlikely to counterbalance the fitness cost of an actual split gene drive. (C) The episomal Cas9-expressing plasmid is unstable in the absence of active selection. With an average loss rate of ~3.8% per generation, more than 2/3 of yeast have lost the plasmid after a single round of asexual overnight growth (10 generations). While variable across independent mating experiments, the plasmid is typically lost at a rate of ~50% during meiotic sporulation, approximately equivalent to a chromosomal transgene. These high loss rates suggest there is minimal risk of Cas9 remaining available to bias the inheritance of the sgRNA-only cassette over generations. Indeed, mitotic loss suggests that the plasmid-encoded gene would likely be eliminated from the population more quickly than a chromosomally-integrated equivalent in the event of an accidental release.