Technology’s answer to invasive species and disease-spreading insects is looking riskier than ever in its current form.

Aside from astonishing cures, one of CRISPR’s most tantalizing uses could be so-called gene drives. As we’ve reported in the past, the technique can be used to easily insert fertility-reducing genes into the DNA of disease-carrying insects or invasive species (such as the humble starling in the U.S., were we so inclined). That would systematically wipe them out. Applied to mosquitoes, the technology could fight malaria and Zika. Targeted at interloping animals, it’s been suggested as a means of preserving fast-declining species in New Zealand.

But the trick is obviously not without its risks. As we’ve explained before, the wholesale removal of a species from a geographic region might be useful, but it’s easy enough to imagine the technique getting out of control, spreading further than intended, and transforming entire ecosystems in ways it wasn't meant to. Indeed, one of the early pioneers of gene drives, MIT researcher Kevin Esvelt, began voicing concerns about their safety several years ago, when the science began to edge toward becoming an applied technology.

Esvelt and several other researchers have published a paper on bioRxiv that argues the technique isn’t safe enough to use in its current, basic form. As he explains along with coauthor Neil Gemmell of the University of Otago in an accompanying commentary in PLOS ONE, his team’s modeling suggests that without further safeguards in place, edited genes could spread to locations where species aren’t invasive at all. “The bottom line is that making a standard, self-propagating CRISPR-based gene drive system is likely equivalent to creating a new, highly invasive species,” the two write. “Both will likely spread to any ecosystem in which they are viable, possibly causing ecological change.”

Still, there is some hope: Esvelt and Gemmell point out that it’s perfectly possible to tweak gene drives so that they don’t snowball out of control in this way. But while researchers have already postulated several ways to do that, none of the ideas are yet mature enough to be used in practice. Until they are, we may be wise to hold off on releasing bucketloads of engineered animals and insects into the wild. “The likely cost of impatience is simply too high,” Esvelt and Gemmell conclude.