On a rainy morning in April, Curtis Rogers woke up and scrolled through e-mails on his phone—pausing on one from a user of his free ancestry-searching Web site, GEDmatch. The message stunned him. It said law enforcement officials had used his site to help find the Golden State Killer.

Rogers immediately told his wife, then started sifting through the e-mail deluge that soon followed. “I had seen news reports the day before saying that this guy had been caught using genealogy DNA,” Rogers says. “I had said to myself, ‘I wonder if we were involved.’ But then I forgot about it until that e-mail came in.” By the time the 79-year-old got to his office in Lake Worth, Fla., news vans were choking the street around it. A woman in a nearby, unrelated office had someone knock on her door looking for Rogers and saying, “there’s been some news about a serial killer.”

“Oh my god! I’ve been in his office,” she said, apparently thinking for a moment that Rogers was a potential suspect.

“My poor neighbor,” Rogers says.

Since that day law enforcement officials—working with independent genealogy experts—have used the free Web site on at least six cases, according to Rogers’ tracking. Most recently The New York Times reported arrests in long-dormant cases in Washington State and Pennsylvania, after authorities again turned to GEDmatch for help. The online DNA profile warehouse hosts genetic data from roughly one million users who voluntarily uploaded this intimate information in the hopes of identifying possible relatives also using the site. Law enforcement officials’ ability to upload a profile and search for any matches—for little cost and usually without a warrant—has reignited hopes for closing a lot of stubborn cold cases. But that means site users may now find themselves unwittingly offering information to criminal investigations. For Rogers this has led to some soul searching about how his site is used. For privacy and genetics researchers it has fueled concerns as to whether data on our genomes can ever be kept truly secure or anonymous.

Rogers and co-founder John Olson created GEDmatch in 2010 as a free repository where people could upload their raw ancestry or health data from genetic-testing services like 23andMe. Although the site is named after The Church of Jesus Christ of Latter-Day Saints’ genealogical file format GEDCOM—which stands for “Genealogical Data Communication”—it has no religious affiliation, and Rogers is not Mormon. (Genealogical history is important to many Mormons, who believe they should identify their ancestors so families can be together in the afterlife.) Rogers, an international businessman, says the project originally stemmed from his own desire to simply find people who share his surname. He and Olson, who handles the technical side of the site, later developed it to offer free and $10 per month premium algorithm services that scour uploaded genetic data to identify relatives. The Web site has been especially popular among adopted individuals seeking members of their birth families.

With the recent flurry of law enforcement activity on his Web site, Rogers—who handles the business side of GEDmatch—has been grappling with how to respond to this unintended use. First, he and Olson tried to make users more aware of such activity. The site’s policy initially mentioned in passing the data could be used for a variety of purposes; it now spells out, at the top of its sign-in page, all the information can be used by law enforcement. Rogers also e-mailed all users who had not been on the site for a while, detailing this possibility. “Some people took their data down, and that’s okay,” he says. But a handful of others e-mailed him to explicitly say they want their DNA used to help solve crimes. One woman wrote her father had been a serial killer, and she hoped her data could help bring closure to victims’ families. “That e-mail made me cry,” Rogers says. On balance, over the last couple of months, “I have come to accept that as long as our people know and are educated about what [their information] could be used for, then I am comfortable with it,” he says. “And I don’t think we have any other option.”

All humans are distantly related, so some genetic similarity may exist just due to chance. But with more than one million people’s data available on GEDmatch, if someone has two second cousins on the Web site—one on each side of the family—that increases the likelihood of finding a more substantive connection. As genetics professor Jim Evans at the University of North Carolina School of Medicine says, one’s first-degree relative—a sibling, parent or child—will likely have half of the same types of genetic variations (formally called variants) on the genome. That person will share one eighth of those same variants with a first cousin, one thirty-second with a second cousin, and so on. This can help a genetic genealogist extrapolate relationships.

Someone familiar with these intricacies and working with law enforcement officials could upload into GEDmatch raw data obtained from analyzing a genetic sample at a crime scene. The user uploading that information would likely make sure the data looked like any other entry provided to a user of a consumer genetics company such as 23andMe. If GEDmatch finds matches from a distant relative, investigators could potentially use that information as a starting point to triangulate a suspect’s identity—perhaps finding other possible relatives by using information from public records outside the Web site. If law enforcement officials then think they have identified a suspect—say by marshaling all their resources to conclude a relatively close match on GEDmatch has five cousins and any of them could be the person they are looking for, but four of them live far from the crime site—they can then legally sift through that suspect’s abandoned trash without a warrant, hunting for a discarded water bottle or other item bearing a DNA sample to confirm they have found the right person.

Yet this approach is still not the final word on definitely identifying a criminal. “DNA doesn’t lie, but if people are sloppy in applying its power, it can lead to unjust outcomes,” Evans says, who also advises judges on the limitations of genetics as evidence. Poor chain of custody practices with a sample collected by law enforcement officials can be a problem, he notes. So can the fact a cigarette butt found at a crime scene, for example, could have been thrown there by an innocent individual before or after a crime. There is, Evans says, much more to definitely fingering a culprit than identifying someone’s DNA from a crime scene. He feels torn about using GEDmatch to investigate crimes. “I think privacy is increasingly becoming a thing of the past,” he says. “But I also think some awfully good things can come out of these big databases—for example, catching these horrible criminals.” All that can be done at this point, he notes, is enhancing public education about the power of genetics and what people’s data can be used for.

This new frontier in criminal investigations could also change the landscape for genealogical genetics—and there is a lot of room for growth. Until now there has been no real competition for GEDmatch, and Rogers has done little to promote its offerings. He says the Web site makes very little in profits because its only source of revenue is the monthly fee it charges for its most advanced services; it never sells anyone’s data or allows advertising. Only a small number of people, Rogers notes, have real expertise in genealogical research with DNA. One of them—CeCe Moore, a genetic genealogist currently with forensic consulting firm Parabon—was involved in three of the most recent criminal cases that used GEDmatch data. Rogers predicts that as the field takes off, law enforcement agencies will likely start creating their own experts. “There’s no doubt in my mind that soon the FBI will be training their own people to use our data,” he says. “I don’t think it will take that long.”