The United States Food and Drug Administration is not known for its prose. So the warning letter that it sent to 23andMe, the direct-to-consumer genetic-testing company, on November 22nd, was a surprise in more ways than one. “It reads like the letter of a jilted lover,” Misha Angrist, a former genetic counsellor who writes about personal genomics and teaches at Duke University, said. “ ‘We went on fourteen dates! We exchanged all these e-mails! We held hands in the park! Now you’re telling me, “Fuck you,” and kicking me to the curb.’ ”

The letter was sent to the 23andMe co-founder and C.E.O. Anne Wojcicki after the breakdown of negotiations about how to regulate the company’s delivery of genetic health-risk information. “Since July of 2009,” the F.D.A. writes, the agency’s Office of In Vitro Diagnostics and Radiological Health had worked patiently with 23andMe, sharing “more than 14 face-to-face and teleconference meetings, hundreds of email exchanges, and dozens of written communications.” Then the company had not only failed to coöperate with the agency but had completely ignored it for six months. After listing everything that they’d talked about, and all of the ways in which the company had rebuffed its entreaties, the F.D.A. ordered 23andMe to stop selling its ninety-nine-dollar DNA tests, which the agency says constitute medical advice and therefore require F.D.A. approval. If 23andMe didn’t cease selling the test and resume negotiations, the F.D.A. said, it would take corrective actions that might include, “but are not limited to, seizure, injunction, and civil money penalties.”

As of Wednesday morning, five days later, the company is still selling its spit kits: you can mail your saliva to the company and have it analyzed for information about your personal genome, from your caffeine sensitivity to your maternal haplogroup. The company’s Web site continues to showcase a story about how “a simple DNA test led to an unexpected diagnosis for a mother and daughter, changing their lives forever”—precisely the kind of marketing that the F.D.A. has objected to. Initially, the company’s only public response to the F.D.A.’s letter was three sentences of boilerplate. On Tuesday night, Wojcicki wrote a short blog post, acknowledging, “We are behind schedule with our responses” to the F.D.A. The company then began to e-mail customers. While the e-mail largely re-states Wojcicki’s blog post, it also apologizes “for the limited response to the questions many of you have raised regarding the letter and its implications for the service, and warns, “We don’t have the answers to all of those questions yet.”

Many people who have tracked the rise of what’s known as direct-to-consumer (D.T.C.), or personal genetic testing, have been expecting a conflict between a company in the field and the F.D.A. since at least 2010, when the agency first announced that it considered the spit kits medical devices and consumer genetic-testing services medical interventions, which require regulation. Advocates for easy access to personal genetic information feared that the F.D.A., and the medical establishment vis-à-vis the agency, would use this regulation to gain tight control of genetic data—which has become increasingly sought after, for both medical and personal reasons. Those who advocate D.T.C. genetic testing, even as they fume at 23andMe for letting things go so sour with the F.D.A., suspect that they’re now watching a government power grab.

Aggravating the conflict is a cultural clash between Silicon Valley, with its devotion to libertarianism and self-knowledge, and a ponderous old-school bureaucracy. But as the veteran journalist and F.D.A. watchdog Matthew Herper notes in one of the sharpest pieces written on this debacle, the agency’s investment into at least four years of negotiations suggests that it was trying to be accommodating. Herper reluctantly concluded that 23andMe is “either … deliberately trying to force a battle with the F.D.A., which I think would potentially win points for the movement the company represents but kill the company itself, or it is simply guilty of the single dumbest regulatory strategy I have seen in 13 years” of covering the F.D.A.

Nearly everyone remains mystified that 23andMe stopped communicating with the F.D.A., even as the company launched new marketing campaigns and claims. One theory is that the company, convinced that the F.D.A. is incompetent, is trying to provoke agency to overreach so it can then outmaneuver it. Herper told me that he thinks this would be foolish, given that he has never seen the F.D.A. overreach, and that it is not an agency to trifle with. As the genomics researcher and blogger known as Mike the Mad Biologist put it, the F.D.A. “is like those humongous tractors used to move space rockets: they’re slow, but crush everything in their path.”

Another possibility is that the company simply dropped the ball. Linda Avey, who founded 23andMe with Anne Wojcicki, in 2006, and left in 2009, still owns shares and keeps in touch with some employees, says she doesn’t know what created the rift between the company and the F.D.A. “It surprised me,” she said. But she pointed out that 23andMe’s general counsel, whom she understands was leading the negotiations with F.D.A., left the company this summer; perhaps it fell through the cracks.

“The whole time I was there,” Avey told me on the phone yesterday, “we were in an outreach mode with the F.D.A. I can’t imagine there was that much of a cultural shift since then. It might be they weren’t paying close attention.” She admits this sounds strange, but thinks that it is no stranger than any other explanation.

At the center of this dispute is the F.D.A.’s designation of the spit kit and 23andMe service as a medical device. In its letter, the F.D.A. seems to argue that it’s not the actual DNA sequencing machines that it sees as the medical device but, rather, the combination of the assay and the company’s offer to tell customers what the assay means. The raw data out of the sequencer is a bunch of letters—“G”s, “T”s, “C”s, and “A”s—that to most people mean nothing. 23andMe takes that data, identifies the gene variants that it represents, and, by drawing on the scientific literature and on its own database, tells the customer about the assay’s medical implications. A 23andMe report might tell you that you carry, say, a gene variant that slightly elevates your risk of heart disease, another that protects you from celiac disease (if you have digestive problems, they probably stem from something else), or, in one of the company’s most laden tests, that you have either zero, one, or two copies of a gene variant that increases the risk for Alzheimer’s, and thus stand either a low, a modest, or a shockingly high chance of developing Alzheimer’s. The company delivers this news with material about how solid the associations are, links to the literature that they’re based on, links to a video series on how genetics work, and many, many caveats.

To 23andMe and to most of its customers, this is information that a person has every right to know. To the F.D.A., it is medical advice, which makes 23andMe’s delivery system a medical device. The F.D.A.’s letter specifically mentioned company’s inclusion of some risk variants of the so-called breast-cancer genes BRCA1 and BRCA2 as the sort of high-consequence information that might lead people to take radical medical action without adequate medical advice. What surgeon would do a mastectomy based on a 23andMe result? A reasonable answer seems to be none—until you remember that surgeons already do thousands of unnecessary operations every year, and that if you can’t find a willing surgeon here you can always go abroad.

Yet 23andMe delivers this information with little friction or guaranteed guidance. Kenneth Britten, a neurobiologist and a customer, learned he has one copy of the gene that increases Alzheimer’s risk, which raises his nominal risk to about one in seven. But he then did enough reading to learn that because neither of his parents developed Alzheimer’s, he could essentially erase this extra nominal risk if he started exercising regularly before he developed symptoms. (He says he now works out a lot.) But he’s a neurobiologist in his prime. A fifty-five-year-old who is confused and depressed and learns that he carries two copies of the risk gene and stands an eighty-per-cent chance of getting Alzheimer’s might reach for a gun, which is the kind of scenario that some genetic counsellors worry about.

My own 23andMe assay included the same Alzheimer’s result that Kenneth Britten’s did. I, too, learned, rather suddenly, that I carry one copy of the risk gene and thus have a nominal Alzheimer’s risk of around thirteen per cent. That news was given to me after I read a special, extra contextualizing statement that 23andMe attaches to such results—which took less than a minute—and clicking on a small square on my profile. Out popped little Pandora, provoking four quick reactions: first, a cold fearful shock; second, a rapid recognition, as my brain churned to contextualize this ugly surprise, that family history and personal health probably cut my tripled risk in half; third, a feeling of intense relief that I did not have two copies of the risk gene; and, finally, the thought, as I walked around the room letting my pulse drop, that it was hard to imagine that anyone could resist clicking that little square.

How do we provide the necessary counsel to people opening Pandora’s box without effectively locking it? The F.D.A. has a point: genetic-risk information needs to be reasonably solid, and its recipients need to be properly served; there’s a place for selective paternalism in genetic testing. Yet the agency seems like a poor candidate to strike this balance. The hope is that it will demand documentation confirming 23andMe’s method for the most laden results, and that it will require a better filter—one that’s thinner than a doctor’s visit but thicker than a click.

If the F.D.A. indeed insists on making 23andMe prove beyond doubt the validity of every single correlation, no genetic-testing service will be able to economically deliver medically relevant genetic information directly to consumers. It will destroy the industry and leave medical genetics in the hands of a medical establishment that has already failed to give people an easy way to obtain and use the elemental information in their own spit. Even now, many doctors who want medically relevant genetic information about a patient get that information by asking her to sign up with 23andMe. Meanwhile, this company’s cheap, widespread testing, which has gathered genetic, ancestry, and health data from nearly five hundred thousand people, has created a rich medical source of risk information.

If the agency instead forbids the company from “interpreting” the tests—that is, if the agency insists that the companies strip from the raw report of genotypes any information about health risks—it would, as the Cold Spring Harbor geneticist Gholson Lyon pointed out this week, threaten to make the data meaningless. Stopping 23andMe, writes Lyon, is not improving medicine. It is “delaying, or worse stopping, the revolution that today’s medicine desperately needs. If that happens, Lyon writes, we’re stuck with a slow, closed system whose constrictions prevent genetic information from helping even those who provide it. Two years ago, Lyon writes, he used genetic testing to identify the cause of Ogden Syndrome, a rare, fatal disease that killed a four-month-old child named Max. “But the rules and regulations for genetic testing in the U.S. … meant I could not share the results of the family’s genetic tests with them.”

The family laid Max to rest without knowing of the gene that killed him. When one of the family’s sisters became pregnant and asked Lyon what he knew, the rules forbade him from telling her. Under the rules of the system that 23andMe is hoping to replace, she would have to set about having any other children without knowing that her family carried a potentially fatal genetic combination that, if it were not tested for, could kill again.