Dark Mirror

by Barton Gellman

  As for living conditions, for Snowden’s purposes, Russia and Hawaii were more alike than different.

  “How many hours a day do you think you are online?” I asked.

  “All the ones I am awake. It’s really difficult to get me to leave the house. I just don’t have a lot of needs,” he said. “Occasionally there’s things to go do, things to go see, people to meet, tasks to accomplish. But it’s really got to be goal-oriented, you know. Otherwise, as long as I can sit down and think and write and talk to somebody [online], that’s really more about the meaning of life to me than coming out and looking at landmarks. I like learning. I like reading. That makes the internet pretty attractive.”

  From time to time, I pushed Snowden on mysteries left unresolved in the archive, or between the lines of things he had already told me. One claim in particular from his “README_FIRST” text file, the one that accompanied his first transmission of documents, was very much on my mind. It was a startling assertion about something Snowden said he had done in order to prove that ordinary analysts could gain access to any U.S. communications they liked. It was not a claim I would publish without evidence, and he had provided none.

  Here in Moscow in 2013, Snowden demurred. “That would be dead man material,” he said elliptically. He was referring, as he had done with me before, to a “dead man’s switch,” a device or arrangement according to which the most sensitive files in his possession would somehow come to light automatically in circumstances he did not specify. Julian Assange of WikiLeaks had made an explicit arrangement like this in 2011, distributing online an encrypted “insurance file” and threatening to release the decryption key if the U.S. government did anything to harm him or shut down WikiLeaks. Snowden never made this kind of threat himself, but Glenn Greenwald did so in a published interview. “Snowden has enough information to cause more harm to the U.S. government in a single minute than any other person has ever had,” Greenwald told the Argentinian daily La Nación. “The U.S. government should be on its knees every day begging that nothing happen to Snowden, because if something does happen to him, all the information will be revealed and it could be its worst nightmare.”

  In private, Snowden dismissed that whole idea as a “stupid revenge narrative.” Greenwald was trying to protect Snowden, it appeared, but he was freelancing. Snowden said it would be wildly irrational to arrange things as Greenwald described, because that would “incentivize every foreign intelligence agency to shoot me in the face.” Kill Snowden, release the secrets. At the same time, in those early months he did not want to give Washington too much reassurance. “If you categorically deny the existence of a dead man’s switch, you take away a cloud of uncertainty that restrains the government’s behavior,” he said. “They’ve been much more careful about how they pursue people like Glenn because they’re concerned that the alternative is worse.”

  When I pressed for answers to some kinds of questions, Snowden still resorted to the “dead man” reply. The evidence existed, he said, and it might appear one day, but he no longer had control of the material. He had not brought secret information to Russia, and possessed no decryption key. Maybe some subtler version of the doomsday device existed after all. “I’m not talking about it beyond what’s already been stated, but given how much thought appears to have gone into things to this point, I think it would be fair to give our friend credit for considering obvious incentive problems and designing around them,” he said. So was there a dead man’s switch or not? Were there more documents that might still emerge? Maddening ambiguities like this were an irreducible fact of life with Snowden.

  When we broke for the night, I walked into a hotel stairwell and down two floors, where I found an armchair in a deserted hallway. I had a chore to take care of before I returned to my room, where I assumed there would be eyes watching me. I might or might not be under surveillance right now, but this looked like my best chance to work unobserved.

  I moved the audio files from the memory card of my voice recorder to an encrypted archive on my laptop, packaging up the notes I had typed as well. I locked the archive in such a way that I could not reopen it without a private key that I had left in a safe place back in New York. I uploaded the encrypted archive to an anonymous server, then another, then a third. Downloading from the servers would require another private key, also stored in New York. Once I was sure that there were redundant copies online, I wiped the encrypted files from my laptop and cut the voice recorder’s unencrypted memory card into pieces. Russian authorities would find nothing on my machines. When I reached the U.S. border, where anyone can be searched for any reason, where the warrant requirement of the Fourth Amendment does not apply, I would possess no evidence of this interview. Even under legal compulsion I would be unable to retrieve the recordings and notes in transit. I hoped to God I could retrieve them when I got home.

  * * *

  —

  Since the first round of NSA stories, readers had been asking me to explain what kind of information I held back from publication. After a while I came up with a parable. Imagine that the NSA supplies mind-reading earrings to the mistress of the emperor of Mars. The emperor’s Earth invasion plans are discovered and stopped just in time. This is blockbuster stuff, a scoop for the ages. If I told the tale, on the other hand, a sensible reader’s response might come in four parts:

  Boy, was that a close call.

  Didn’t know the NSA could do that.

  Sure am glad they did that.

  Now they can’t do it anymore, you son of a bitch, because you went and told the Martians all about it.

  (Subsequent reports from the Red Planet describe the mistress’s beheading, earrings and all.)

  The analysis would have to go deeper than this, of course. Secret mind-reading tech would be a dangerous and controversial thing for the government to acquire. The public would have a strong claim to know about that development, even if I kept quiet about the operation on Mars. But the parable serves its purpose here, I think. It is supposed to be an easy case. Most people can probably agree that it would be wrong to blow the cover of the mistress and her earrings, especially with planetary conquest at stake.

  Snowden and I had a conversation in Moscow about drawing lines like this. He brought up two kinds of encrypted communication. The NSA could decrypt one kind but not the other. We spoke of them elliptically, knowing that our conversation might be overheard. I said I would not be writing either of those stories.

  “Why?” he asked, more curious than argumentative. He had seldom expressed an opinion about what should be published.

  “Because it’s—generally speaking, I don’t like the idea of pointing to vulnerabilities or blind spots,” I said.

  “I see your point. You can’t tell everybody without telling the bad guys.”

  “Right. And in the other case you have a particular encryption standard favored by certain parties who pretty much anyone would regard as legitimate foreign intelligence targets, and I know what the capabilities are of the NSA against those, and I’m not the least bit interested in publishing that.”

  I will not be specific here about the stories I chose not to write, but some of the NSA archive, I strongly believed, should not see the light of day. As I’ve mentioned, there were photographs of operational personnel in the field. Names of adversaries under surveillance. Particulars of the networks on which they talked. The NSA had acquired precious information that anyone who wished my country well would want the government to know. I did not consider revealing those things, which is not to say I never found hard cases.

  Other people, no doubt, would draw the lines differently. “The damage is incalculable by anyone not a professional in the field. Sorry but you do not know what is damaging and what isn’t,” former NSA chief scientist George Cotter wrote to me. “Your view that the public has a right to know is seriously offset by an author’s limited and (regrettably) non-professio
nal judgment.” Bill McRaven, Bob Litt, James Clapper—they and many of their peers believed without second thought that I should publish nothing classified, ever, full stop. The proposition seemed obvious to them. Proper authorities drew lines of secrecy, and I should defer to them. Advocates for radical transparency, on the other hand, objected in principle when I held anything back. And plenty of mainstream critics wrote that I should reveal somewhat more than I did—additional details from the black budget, for example, in service of a better-informed public debate on spending priorities.

  Criticism from all sides did not mean I got the balance right. I have always thought that a fatuous argument. Logically it could also mean I made the wrong choice every time. Weighing the value of news against potential harms was bound to be a contested exercise. There was no avoiding the responsibility, in my view. Ethical journalists do not publish every secret they learn, but neither can they accept the government’s judgment as final.

  A lot of people are surprised to hear that we have any choice in the matter. They assume it is illegal to publish classified information, or believe that it should be so. That is not in fact the state of the law. Under prevailing legal practice, a person entrusted with classified information by the government can be charged with espionage for giving that information to someone who lacks such a clearance, even if the recipient is a reporter and not a foreign intelligence service. Reporters, however, have not been held criminally liable for publishing those secrets. The First Amendment is widely thought to prevent such a prosecution, given the central role of freedom of speech and of the press in our constitutional system. No law, for much the same reason, forbids publication of false information (with narrow exceptions such as libel). Likewise, publishing information stolen by someone else does not qualify legally as trafficking in stolen goods.

  There was, as the Post lawyers had warned me, one catch. The government had never tried to prosecute a journalist or publisher under the Espionage Act of 1917, which is so broadly drafted that it could be read as banning any news story about “national defense information,” whether classified or not. No one knew for certain how such a prosecution would fare under First Amendment analysis. In mid-2019, the Trump administration decided to test the question. In an indictment unsealed against Julian Assange, the founder of WikiLeaks, the Justice Department charged him with sixteen counts of espionage, three of them based entirely on his communication of secrets “to all the world by publishing them on the Internet.” Whether you wish to call Assange a “journalist” or not, the distinction is not important legally. The elements of the crime alleged against him—disclosure of secret information to the public—are very hard to distinguish from what I did with Snowden’s NSA archive, even if I was more selective about it. If Assange is extradited, tried, and convicted for publishing classified information, and if the conviction withstands a constitutional appeal, then the climate for investigative journalism on national security will change pretty starkly. At this writing, that day has yet to come.

  By the time the Snowden story arrived I had done a lot of thinking about these questions. My master’s thesis at Oxford explored the foundations in democratic theory for a “right to know” in the context of national security. I had twice taught a Princeton class called “Secrecy, Accountability, and the National Security State.” Most of my guest speakers—including Mike Levin, a former NSA chief of information security who had once called me a “traitor” on film—agreed by the end of our three-hour seminars that there might be circumstances in which disclosing classified facts would be the right thing to do.

  What if the U.S. government deliberately exposed American troops to nuclear radiation in order to learn more about the medical effects? That really happened after World War II. “It is desired that no document be released which refers to experiments with humans and might have adverse effect on public opinion or result in legal suits,” wrote the responsible official at the Atomic Energy Commission in 1947, in a memo that remained classified until 1994. “Documents covering such work field should be classified ‘secret.’” If reporters had known the truth then, should they have suppressed it?

  What if the U.S. government deliberately infected commercial sex workers in Guatemala with gonorrhea and syphilis? That happened, too, in wildly unethical experiments between 1946 and 1948. When the sex workers failed to transmit the diseases quickly enough for scientific purposes, “the research approach changed to direct inoculation of soldiers, prisoners, and mental hospital patients,” the government acknowledged in an apology in 2010.

  What if a classified military investigation found “numerous incidents of sadistic, blatant, and wanton criminal abuses” against foreign detainees, in violation of the Geneva Conventions and the Uniform Code of Military Justice? That happened at the Abu Ghraib Confinement Facility in 2003. Major General Antonio Taguba’s investigation was classified SECRET//NOFORN. Even oversight committees in Congress did not receive an unredacted copy until an unauthorized leak made it public. Much the same sequence of events, with classification stamps employed to conceal information that public officials could not or did not wish to justify, took place after the government tortured al Qaeda suspects in secret prisons, authorized warrantless surveillance of U.S. citizens, and lied about intelligence on weapons of mass destruction in Iraq. These were history-making events, full of political and legal repercussions, but they were hidden from public scrutiny until news stories broke through barriers of classification.

  My point here is fairly modest, I think. Classification stamps do not capture the whole of the public interest when the question is whether or not to keep something secret. Even if secrecy served legitimate purposes in these examples, it also thwarted accountability for consequential choices. “Cold War secrecy became a habit,” wrote Mary Graham, author of the authoritative history of executive secrecy. “Presidents expanded surveillance in order to catch spies and uncover enemies’ secrets. But they also hid their mental or physical illnesses, their ethical lapses, their intrusions into the privacy of ordinary citizens, and their illicit efforts to weaken their political opponents.” The Moynihan Commission, the best known of the many bipartisan panels that have studied secrecy since the 1950s, found that the classification system “is used too often to deny the public an understanding of the policymaking process, rather than for the necessary protection of intelligence activities and other highly sensitive matters.”

  Consider a simple question. How often does the NSA break its own privacy rules? The agency keeps internal statistics of those “compliance incidents.” The statistics are classified Confidential, which is supposed to mean that disclosure would damage national security. Why should the mere number, a simple count of the errors, be treated as a state secret? Quite a lot harder to justify: the Justice Department, which prepares a similar compliance report for Congress and the FISA Court, classified exactly the same statistics as TOP SECRET//SI. That had a very practical impact. High-level clearances are rare among members of congressional staffs. Most offices had nobody eligible to read the compliance reports. You might suspect that someone preferred it that way.

  Members of Congress often express frustration at their impotence to oversee secret bureaucracies in the executive branch. It is difficult, even with constitutional authority, to induce a person to tell you what you do not know how to ask. “You have to start just spitting off random questions,” I once heard Representative Justin Amash say, explaining why Congress failed in its oversight of surveillance. “Does the government have a moon base? Does the government have a talking bear? Does the government have a cyborg army? If you don’t know what kind of things the government might have, you just have to guess and it becomes a totally ridiculous game of twenty questions.”

  Sometimes an official secret is merely trivial, classified by force of habit. Steven Aftergood of the Federation of American Scientists once turned up a Navy laundry and dry cleaning manual that was stamped Secret. In the Snow
den documents I found another choice example. Here is a Top Secret, compartmented paragraph in its entirety from a presentation prepared in 2003:

  (TS//SI) On 4 November 1979, a mob of Iranian students stormed the U.S. Embassy in Tehran and took the diplomatic staff hostage. The students actions were soon sanctioned by Iran’s revolutionary government, and a standoff developed between the United States and Iran over the issue of the 52 captive American diplomats. The standoff lasted for almost two years before the hostages were released in January 1981 (Inauguration day for Ronald Reagan).

  Few events in modern history were more thoroughly covered in the news media than the hostage crisis in Iran. Still, someone at the NSA decided—more than twenty years after the fact—that mentioning the episode publicly would do “exceptionally grave damage to national security.” That is the legal standard for a Top Secret classification under Executive Order 13526.

  These are neither rare, wacky exceptions nor representative samples of the classified universe. There are plenty of legitimate secrets, if by “legitimate” we mean that their disclosure could bring a foreseeably bad result. But not all classified secrets meet that test, and sometimes the workings of cause and effect are hypothetical, conditional, or unclear. Sometimes the asserted harm is controversial. Does it count as harm if disclosure of classified U.S. government conduct offends an ally, provokes a lawsuit, builds pressure for legislation, shrinks the NSA’s pool of recruits, or leads a private company to encrypt its email servers? All those things happened after the Snowden disclosures, and any of them may have caused a loss of collection, but if so the result was a feature, not a bug, in the systems that govern us. Diplomacy, law, politics, and free markets worked as intended.

  None of this is to dismiss the idea that stories about the Snowden documents, including some of my own, resulted in genuine damage to NSA operations. I do not know the particulars, because damage of that sort would itself be classified, but I refuse to pretend that so many secrets could have spilled without any consequence. “You don’t delude yourself, I hope,” former NSA analyst Alan Tu told me. “Despite your efforts, you have negatively impacted some valuable work people have done, a lot of work actually. You can say it’s the cost of ripping away corrosive lack of transparency in a hostile environment, but you should still know there is a cost to what you have done.”