For that reason, the authors of the report lay out exactly how many and what kind of off-target effects might be acceptable. They put that threshold at no more than the average rate of new mutations an embryo spontaneously acquires. DNA replication isn’t perfect, and most people are born with a few dozen mutations that don’t exist in either of their biological parents’ genomes. Gene editing shouldn’t introduce any more genetic variations than occur naturally, the authors concluded, and the types of changes should be carefully studied in the lab to make sure they don’t lead to adverse outcomes.

The trouble is, though, that right now there aren’t any good methods for assessing off-target effects in embryos. Doing so requires collecting large amounts of DNA, which can only be done by sacrificing a number of cells from the embryo for genetic sequencing. In addition to being unreliable, these methods harm the viability of the embryo, making it less likely to result in a pregnancy. It could take years for better methods of evaluation to be developed, says commission member Haoyi Wang, a reproductive biologist at the Institute of Zoology and Institute for Stem Cell and Regeneration at the Chinese Academy of Sciences. “From the genome editing to the genome sequencing of a single embryo, there are still many gaps to be filled,” Wang told reporters at a press briefing Thursday.

The commission was more narrowly focused on addressing these sorts of scientific gaps, while other authorities, like the WHO, will look more broadly at how societies might decide to accept human germline editing and how governments will regulate the technology. Developing ethical frameworks can’t just be about autonomy, privacy, and justice, says commission member Bartha Maria Knoppers, who directs the Centre for Genomics and Policy and serves as the Canada Research Chair in Law and Medicine at McGill University in Montreal. “For me, scientific quality and safety are primordial ethical considerations; they’re not peripheral,” she says. “I think this report reflects the emphasis on getting those aspects right.”

Brightening the lines between good science and bad is especially important to prevent anyone intent on operating outside of established regulatory frameworks from causing undue harm, says Dana Carroll, a commission member and professor of biochemistry at the University of Utah School of Medicine. “These standards have to be so high because actually initiating a pregnancy with an edited embryo is going to lead us into territory we essentially have no experience with,” he says.

Carroll, whose lab pioneered some of the earliest work with older (pre-Crispr) generations of genome editors, was in the audience at a summit in Hong Kong in November, 2018, tensely watching as He Jiankui presented data describing his experiments creating the world’s first Crispr’d children. But that data never was never published. And importantly, says Carroll, no follow-up information on the status of the children has ever become public. So while Jiankui opened the door for other researchers to follow, he didn’t leave a lot in the way of better understanding what happens when you stick Crispr inside a human embryo and then stick that embryo inside a human uterus.

“We’ve heard rumblings that people are waiting in the wings, ready to proceed,” says Carroll, referring to other researchers intent on booting up their own Crispr baby experiments. “So we wanted to make sure that when they do, that the standards they need to meet for safety and efficiency and specificity are clear.”

In addition to providing a list of all the kinds of preclinical work that scientists will have to accomplish before moving to human trials, the international commission also made recommendations for who would be eligible to participate in them. After conducting listening sessions with a number of patient and disability advocacy groups, the commission settled on a very narrow set of indications for which it would be deemed ethical—in other words, for the benefits to outweigh the risks—to apply human germline gene editing, at least at first. Should a country decide to move forward with the technology, they wrote, gene editing should only be used to treat serious monogenic diseases—that is, caused by a mutation in a single gene that causes severe morbidity or premature death. Examples include cystic fibrosis, sickle cell anemia, and Tay-Sachs disease.