The Self-Appointed Spies Who Use Google Earth to Sniff Out Nukes

Nuclear intelligence isn’t just for government agencies anymore. A motley crew of outside watchdogs has found creative ways to deter proliferation.

a satellite superimposed on a nuclear symbol
Alexander Steam / Timofeev Vladimir / Dean Drobot / Shutterstock / The Atlantic

About the author: Amy Zegart is a contributing writer at The Atlantic. She is a senior fellow at the Hoover Institution and the Freeman Spogli Institute at Stanford University, and the author of the forthcoming book Spies, Lies, and Algorithms: The History and Future of American Intelligence (Princeton University Press).

Tracking nuclear threats used to be the sole province of secret agents and analysts at high-powered government intelligence agencies. Not anymore.

Today, the world of new nuclear sleuths is straight out of the Star Wars bar scene.

Peering into the hidden nuclear activities of North Korea, Iran, and other suspected proliferators are journalists, hobbyists, professors, students, political-opposition groups, advocacy groups, nonprofit organizations, for-profit companies, think tanks, and former senior government officials with informal links to international weapons inspectors, American policy makers, and intelligence leaders.

Among this wildly eclectic mix of individuals and organizations, some are amateurs. Others have extensive expertise. Some are driven by profit, or political causes. Others are driven by a mission to protect the United States and reduce global nuclear risks. Nearly all harbor an obsessive interest in nuclear secrets and finding creative ways to unlock them. Together, these self-appointed watchdogs are transforming American nonproliferation efforts—and largely for the better. Yet they also create new challenges for the U.S. government, which once enjoyed a near-monopoly on detailed surveillance imagery of hostile countries with nuclear ambitions. American intelligence agencies must now operate in a world where highly revealing information is sitting out in the open, for anyone to see and use.

David Schmerler, part of a team at the James Martin Center for Nonproliferation Studies, goes by the nickname “Geolocation Jesus” because of his skills at pinpointing North Korean locations using far-ranging clues, such as Kim Jong Un’s public schedule, the number of skylights in a photographed room, Google Earth, and his knowledge gleaned from watching every North Korean missile propaganda video ever released. Frank Pabian, who works closely on a Stanford University team led by former Los Alamos Laboratory Director Siegfried Hecker, is one of the world’s leading imagery analysts and a former American weapons inspector. Then there’s Jacob Bogle, a coin dealer by day and North Korean mapping hobbyist by night who has created one of the world’s most detailed maps of North Korea from his home in Murfreesboro, Tennessee. In my own research, I’ve found 17 major groups or players actively tracking illicit nuclear activities around the world.

Not all of the work generated by this wide-ranging ecosystem is accurate, but much of it is pathbreaking. And all of it is unclassified.

For decades, the governments of great powers—and especially the United States—had cornered the satellite market, for good reason: Operating anything in space was technically demanding and inordinately expensive. The CORONA satellite, a project of the CIA and the U.S. Air Force, was the first to photograph large swaths of the planet in 1960, returning its film in a capsule that had to be parachuted down to Earth and captured in midair over the Pacific Ocean to be developed. The engineering challenge was so punishing that CORONA’s first 13 missions failed. But on the 14th attempt, it hit pay dirt, photographing more Soviet territory than all previous U-2 spy-plane flights combined. According to Albert Wheelon, the CIA’s first deputy director of science and technology, “It was as if an enormous floodlight had been turned on in a darkened warehouse.” The Soviets soon launched their own photoreconaissance satellite, Zenit-2, in 1962. It, too, was expensive and failed repeatedly before it finally returned usable imagery.

Since the early 2000s, however, commercial satellites have become common. According to the 2019 threat assessment issued by the Office of the Director of National Intelligence, the annual number of satellite launches has quadrupled in the past five years. In a single launch last year, the private firm SpaceX sent 64 small satellites from 17 countries, and a Florida middle school, into space. News reports note that in 2018 alone, 322 small satellites the size of a shoebox were hurled into orbit, and some analysts estimate that more than 8,000 small satellites will be launched in the next decade.

Spy satellites still offer better resolutions and capabilities. But today’s commercial satellites are narrowing the gap, offering image resolutions that are roughly 900 percent better than what they were just 15 years ago—sharp enough to distinguish different types of cars driving along a road and capture certain indicators of equipment used in nuclear-weapons programs. What’s more, constellations of small satellites can fly over the same location multiple times a day, identifying changes on the ground in near–real time. Already, a San Francisco start-up called Planet has more than 150 satellites in orbit. Seattle-based BlackSky, which launched in 2013, has 60 satellites and says it flies over major cities 40 to 70 times a day. Perhaps most important, the costs of acquiring satellite imagery have plummeted—just as computing and communication power has been radically democratized.

Today, more than half the world’s population uses the internet, and by next year, more people will have cellphones than running water. Connectivity is making everyone a potential intelligence officer. People can review photos posted on social media, record seismic activities on their cellphones, and use 3-D modeling apps to assess whether a suspicious facility could actually accommodate the kind of equipment used in nuclear-weapons development.

In recent years, expert groups in this ecosystem—teams led by Hecker, also my colleague, at Stanford; Jeffrey Lewis of the Center for Nonproliferation Studies; and David Albright at the Institute for Science and International Security—have made a number of breakthroughs. They have pinpointed the location of North Korea’s first two nuclear tests years before the North Koreans confirmed them. They have tracked the construction of a new nuclear reactor at Pyongyang’s Yongbyon complex and estimated its operational capability. They have identified the function, size, and capacity of Iran’s secret nuclear facility at Natanz. And they have quickly debunked false information, such as Kim Jong Un’s claim that he had successfully tested a submarine-launched ballistic missile in 2016.

Nongovernmental nuclear sleuths have taken one another on, too. When an Iranian opposition group called the National Council of Resistance of Iran tried to derail the Iran nuclear deal in 2015 by announcing that a company called Maritan was secretly housing a nuclear facility in the basement of its Tehran office, Jeffrey Lewis’s team showed within a week that the purported evidence was false. Maritan was a real company. It even had employees on LinkedIn. But it had nothing to do with nuclear enrichment. It specialized in making secure documents like national identification cards.

Analyzing satellite imagery, Lewis’s team found no construction activity at the Maritan office site during the alleged nuclear-facility-construction time frame or obvious signatures of nuclear-enrichment activities found at other known Iranian sites—such as ventilation systems or electrical substations to power nuclear centrifuges. Using 3-D modeling, it showed that the alleged facility was in fact too small to fit the necessary nuclear machinery and infrastructure. And Lewis found that the group’s photograph of a lead door—which supposedly proved there must be radioactive substances inside—was actually just copied from a commercial Iranian website. Lewis’s team also found that none of the known Iranian sites ever used lead doors because radiation leakage had never been a concern. Perhaps most incredibly, it used crowdsourcing, social media, and a GPS location app to find someone who had actually been to Maritan. The team contacted this person by email and verified who he was—identifying his hobbies, marital status, and even getting his photograph in the process, all from open-source detective work. From him, it learned that Maritan regularly brought foreign contractors to the office, making it highly unlikely that the company would put a secret nuclear facility in its basement.

As these examples suggest, nongovernmental nuclear sleuths provide more hands on deck for intelligence agencies to validate or disprove nuclear developments. And because nongovernmental organizations and individuals operate in the unclassified world, their findings can be shared within governments and between them. That’s a major shift. Their findings can be publicized across agencies and borders, galvanizing attention to an issue. In addition, open-source intelligence makes possible more input and analysis from a broader array of experts than information collected by traditional intelligence agencies. Particularly because nuclear threats are so dangerous, intelligence about them is almost always highly classified. While the siloing of this information has benefits, it also comes with serious drawbacks. Chief among them is the risk that the information will not be sufficiently subjected to independent or competing perspectives. The more classified something becomes, the fewer people get to see it. Going black runs the risk of going dark, leaving bits of intelligence underdeveloped and under-considered.

To be sure, open-source nuclear sleuths raise the risk that errors could go viral and that adversaries could be tipped off that they need to hide their nuclear activities better. Amateurism has its limits, especially when it comes to analyzing images from space. Even a seemingly unmistakable landmark can be hard to pick out. From directly overhead, St. Louis’s Gateway Arch does not appear archlike at all. Identifying telltale indicators of weapons proliferation is a very subtle art; imagery analysts need to understand the nuclear-fuel cycle so they know which visual clues to look for. To the untrained eye viewing objects from unfamiliar angles, a road can look like a railroad track, a dried-up stream bed can look like a tunnel, a massive elevator can look like a missile launch pad, a livestock pen can resemble an Indian nuclear test site, a cylindrical foundation for a hotel can look like the beginnings of a hidden nuclear facility. These aren’t hypothetical mistakes. Amateur nuclear detectives have made these errors, which were publicized before they were corrected.

In 2011, a group of Georgetown students and their professor even sparked congressional hearings and a flurry of activity inside the Pentagon when their amateur analysis suggested that China was hiding thousands more nuclear weapons in underground tunnels than American intelligence officials had estimated. Their analysis turned out to be incorrect, but not before it captured national headlines and generated pressure for officials to waste time double-checking and justifying the accurate assessments they already had.

And that’s what happens with well-meaning mistakes. Nefarious actors could inject deliberate deceptions, raising the risk that falsehoods will be believed, truth will be doubted, and intelligence agencies will be tied up serving as “verifiers of last resort” rather than advancing their own intelligence collection and analysis priorities.

Clever nuclear sleuthing could also tip off adversaries, alerting them to weaknesses in their camouflage, concealment, and deception techniques that they didn’t know existed and causing them to take new measures that make monitoring by everyone more difficult. Some public evidence already suggests that the detail of images available on Google Earth has prompted new Chinese efforts to conceal military facilities from more frequent satellite-shooting intervals. And after Dave Schmerler was able to measure the size of North Korea’s first nuclear device and locate the building where it was photographed by using objects in the room as clues, the next North Korean photo of a warhead was taken in an otherwise empty room. Whether Schmerler’s research prompted the change remains a mystery. But history suggests that hiding and seeking go hand in hand: Whenever new monitoring capabilities are revealed to an adversary, countermeasures are likely to follow.

Yet despite these risks, the democratization of nuclear-threat intelligence is likely to be a boon to the cause of nonproliferation. Aspiring nuclear states have always gone to great lengths to conceal their atomic ambitions and activities. But dark programs can quickly spiral into global dangers—as Americans saw in October 1962, when the Soviets’ determination to surprise the United States with a nuclear fait accompli in Cuba brought the world to the brink of total nuclear war. Thanks to the new nuclear sleuths, estimating nuclear dangers isn’t just for governments anymore. For would-be proliferators like Iran and North Korea and future regimes that might consider following in their footsteps, hiding the evidence is going to get a whole lot harder.