If a tanker splits its hull and dumps oil into the sea, trained teams show up with specialized gear to begin the process of stanching the flow and cleaning up the spill. Today, there’s no equivalent team or tools for resolving a “spill” of genetic material into the environment, but that could soon change.

Over the next four years a new program in the Pentagon’s Defense Advanced Research Projects Agency (DARPA) plans to cultivate, among other things, a kind of cleanup crew for engineered genes deemed harmful to or undesirable in an ecosystem. The initiative, called Safe Genes, comes at a time when so-called “gene drive” systems, which override the standard rules of gene inheritance and natural selection, are raising hopes among some scientists that the technology could alter or suppress populations of disease-carrying insects or other pests in as few as 20 generations.

The Bill and Melinda Gates Foundation sees so much promise in gene drive technology that it plans to double spending on its Target Malaria initiative, which aims to create systems for driving genes in two species of malaria mosquitoes, to $70 million. Yet without careful precautions, a gene drive released into the wild could spread or change in unexpected ways. Kevin Esvelt, head of the Sculpting Evolution lab at MIT Media Lab, which is applying for Safe Genes funding in collaboration with eight other research groups, predicts that eventually, perhaps around 15 years from now, an accident will allow a drive with potential to spread globally to escape laboratory controls. “It’s not going to be bioterror,” he says, “it’s going to be ‘bioerror.’”

DARPA itself has been one of the largest public funders of synthetic biology research in the U.S. in recent years, upping its spending on synthetic biology projects to more than $100 million in 2014 from nothing in 2010, according to one analysis. The agency announced its Safe Genes program in September 2016 and plans to award funding to multiple research teams by the first half of 2017. “If we’re going to be really bullish about genome engineering,” says DARPA program manager Renee Wegrzyn, “we need to be just as aggressive with tools to reverse those changes.”

Normally, a parent organism with a given trait passes that genetic code to offspring about half the time. Recent advances combining the gene-editing tool CRISPR–Cas9 (for, clustered, regularly interspaced, short palindromic repeats with a guiding enzyme called Cas9) are now making it easier for scientists to modify a genome such that nearly all offspring inherit the desired trait.

Wegrzyn’s team is now evaluating proposals from researchers outside DARPA to develop “radically different” approaches to remove, replace or inhibit unwanted genetic changes and their effects, whether the culprit involves a gene drive mechanism or a different type of editing tool. "If we get this right [and are] prepared for these accidental situations,” Wegrzyn says, "we’re going to be 80 percent there to deal with it in a nefarious situation."

The search for ways to eliminate engineered genes from a variety of species and habitats is only one of three focus areas for DARPA’s new program. A second area challenges teams who receive funding from Safe Genes to design systems for controlling and reversing gene editing tools like CRISPR–Cas9. This could mean designing an editing system to work only in one type of tissue, for example. Or, as Esvelt’s lab suggested in June 2016, drive components could be scattered throughout a genome so the driving mechanism fades from the gene pool as generations pass. Lastly, researchers will be tasked with developing small molecules, antibodies or other means of arming organisms to rebuff genome editors at a molecular level as well as novel ways to deliver the most promising of these inhibitors to cellular or animal hosts—orally, perhaps, or by way of a specially designed virus.

To receive funding from Safe Genes, teams focused on gene-drive technology must work on controls plus at least one of the program’s other focus areas—remediation or inhibitors. According to DARPA’s detailed proposal guidelines, Safe Genes expects all projects supported by the program to begin with mathematical modeling and testing in contained systems that increase in complexity and scale only after problems are found and fixed in the less risky environment. “I don’t think you let a car go out of the garage without brakes or seat belts,” Wegrzyn says. “You really need to understand every component of how it works and how you fix the system if anything goes awry.”

In addition to DARPA environmentalists, biosafety experts and leading gene drive researchers say a new approach to mitigation and control is needed for the technology to advance safely. According to a recent report by a panel of 15 experts assembled by the National Academies of Sciences, Engineering and Medicine, traditional containment tools will eventually fall short for systems hardwired to spread genetic information.

The idea of “reversal drives,” which would overwrite mutated genes with the original sequence, often arises as a promising solution to gene drives gone wild. But the National Academies report warns the technology is an incomplete safeguard that may produce its own unintended effects and fail to undo the ecological impact of the original drive. “A reversal drive,” Esvelt says, “will always play catch-up as the unwanted drive spreads through the wild population.”

Another option is to create what’s called an immunizing reversal drive in a kind of offense-as-defense move that would overwrite the gene drive and also arm the wild population against it. This, Esvelt says, “could solve the problem but would of course spread through the entire wild population itself.” Original traits might be restored but at the genetic level the population would still bear components from the fix-it drive.

Suppressing one species may unleash a surge by a competitor that is equally capable of spreading a disease like dengue or malaria. Disrupting a population could also produce ripple effects in other corners of the ecosystem, for example taking a food source from aquatic predators that feed on mosquito larvae. “You couldn’t go back in time,” says Jason Delborne, a professor of science, policy and society at North Carolina State University who helped produce the National Academies report, and is part of a team that is putting together a proposal for Safe Genes funding. “We shouldn’t move forward being emboldened [by the idea] that we can release a new trait and have things go back to the way it was.”

Evolutionary geneticist Austin Burt, who leads Target Malaria’s research at Imperial College London and has no affiliation with Safe Genes, concurs. The prospect of remediation, he says, “shouldn’t give us a cavalier attitude.” Instead, the goal should be to do the incremental work to anticipate and prevent problems. “We have the precedent of biological control,” he says, "where if you have an invasive pest that is destroying your crop, you can release a parasitoid wasp,” which kills its host. “They do a very careful assessment. They don’t have something in their back pocket,” to delete errors.

That said, Target Malaria’s researchers are already thinking about how they might stop a gene drive in its tracks. In a response to recommendations in the National Academies report, the organization notes two possible strategies for harnessing natural selection to incapacitate a gene drive that is diminishing a population. One route, first suggested by Burt in 2003, is to release a sequence that is resistant—effectively unrecognizable—to the guiding enzyme that finds cuts of DNA in a gene drive. “If that was functional, then that would spread through the population and the population would rebound,” Burt explained. An alternative, proposed earlier this year in Nature Biotechnology, involves a suppressor construct, which would essentially program a secondary guiding enzyme to target the guiding component of the original drive.

For now, Wegrzyn wants to keep the door “wide open” to new ideas that can keep pace with advances in genome editing. “There’s a lot of speculation going on in the field right now, and you can continue to speculate indefinitely,” she says. “But you need to start to collect the data to really see how these systems are going to work.”