Deep State

Home > Other > Deep State > Page 29
Deep State Page 29

by Marc Ambinder


  A few weeks after the programs began, Deitz called Addington’s office and asked to see the Justice Department’s legal opinion on the special programs. Addington refused to provide it. The president was entitled to private legal advice, and the OLC’s opinions were not designed to be shared even within the executive branch. Deitz asked him for a summary, so Addington gave Deitz the gist: the president had inherent authority under Article II of the Constitution to—well, it was Article II über alles. He could do anything he wanted. Deitz did not agree.

  We don’t know what John Yoo wrote exactly, because his initial opinion and many subsequent memoranda remain classified. But there are clues in a September 25, 2001, memo prepared by Yoo for David Kris, who was then an associate attorney general trying to figure out whether FISA could be interpreted to allow collection in cases where foreign intelligence was merely a purpose, rather than the purpose. The change of an article might allow FISA orders to be issued for the purposes of terrorism cases, Kris reasoned. What he did not know was that the creators of the program had already decided against asking Congress to modify FISA in any substantial way.22 Yoo responded to Kris in such a way as to suggest that Congress and the courts wouldn’t mind a little participle switcheroo. “The factors favoring warrantless searches for national security reasons may be even more compelling under current circumstances,” concluded Yoo after analyzing circuit court decisions. “After the attacks on September 11, 2001, the government interest in conducting searches related to fighting terrorism is perhaps of the highest order—the need to defend the nation from direct attack.”23

  During interviews about the topic years later, Yoo basically confirmed that he had argued in favor of inherent presidential authority to protect the nation and that the congressional authorization to fight al-Qaeda at least implicitly authorized aggressive surveillance during this war. In any event, FISA was simply not applicable to the post-9/11 terrorist threat.24 He argued, in other words, that FISA was okay until existential events rendered it irrelevant.

  The NSA’s initial opinion on the “Presidential Authorization for Specified Electronic Surveillance Activities during a Limited Period to Detect and Prevent Acts of Terrorism within the United States” was much more cautious.∗ The opinion remains classified, and Deitz would not discuss its contents with the authors. Several who have read it say that it did not draw the same conclusion as Yoo did. Instead, it read in to FISA as an applied exemption that would permit surveillance without a court order if it had direct bearing on imminent threats—particularly if the type of war incorporated the United States as a battlefield, and if a combatant or terrorist was receiving orders from another country. The NSA lawyers viewed the Congressional Authorization for the Use of Military Force against al-Qaeda and its affiliates as a safe harbor. Intelligence collection was clearly a part of the process of going to war.

  Deitz added one thing that displeased the White House: he insisted that the analysts stick to the probable cause threshold, and not simply a “reasonable suspicion” threshold. In practice, it might not have altered the decision chain. The analyst would decide independently and the collection would begin immediately. NSA lawyers and the attorney general could stop it if they didn’t think the threshold was actually met.

  Had Congress been aware that the NSA and the Department of Justice (DOJ) had different, somewhat contrary, legal justifications for the program, they might have had greater reservations. But Addington argued that the OLC opinion was tantamount to internal deliberations and could be—indeed should be—withheld. It is possible that he was aware of how the two opinions conflicted and knew that their commingling would create problems. Regardless, he argued that if FISA didn’t apply, it didn’t apply. The NSA could collect on phone calls and read e-mails if it wanted to, he believed.25

  That said, the threshold to tap into any phone was fairly inviolable when it came to U.S. persons: the government had to show probable cause that a U.S. person was connected to terrorism before any interception could begin. Anticipating the objection that FISA expressly allowed for emergency wiretapping, both the NSA’s and the Justice Department’s legal briefs argued that as a practical matter, because the probable cause standard as the court would define it could not possibly be met to the satisfaction of a court, all wiretap activities would be halted by the FISC as soon as the exemptions ran out, which would be disastrous if the NSA were in the middle of tracking someone.

  Context is crucial here: as all of this was playing out, lower Manhattan was still covered in ash. The lawyers were trying to figure a way to allow the executive branch to discover whether any additional plots were imminent or whether conspirators were working on U.S. soil. A secondary priority was to learn whether other al-Qaeda sleeper cells were preparing for later plots. Neither the White House nor the NSA viewed the surveillance as a police investigator would. Instead, this was an intelligence operation that by design and urgency had to be carried out pursuant to the laws governing a covert operation.

  And that’s basically how the White House played it. Within a week of this determination—on October 1, 2001—Hayden went to Capitol Hill and gave a briefing to the Gang of Eight on the program. He did so over the objections of Addington and Cheney, which itself provides some evidence that, contrary to a shared belief among the program’s critics, the two men did not make all of the decisions about it and were occasionally overruled.26 There were no notes taken during that briefing, so it is not clear whether the congressional leadership was forthright in later complaints that what they were told was less egregious than what the NSA actually did (or possibly just described to them as technically as possible in order to obscure the egregiousness).27 Associates of Hayden’s say that the target analysis component of the surveillance program—identified as STELLAR WIND—had not yet been set up when he first went to Congress, though engineers were working on it.28

  Although terrorism was on the minds of NSA managers before September 11, 2001, the agency was not on a war footing. It had few linguists capable of translating intercepted phone calls, much less a good system of figuring out how to determine which calls to intercept. (Its SIGINT directorate tried to ascertain how many skilled Urdu and Pashto speakers it had on staff; the NSA’s own internal count was off by a factor of three.)

  Still, pretty much any phone call originating from Afghanistan and from the tribal regions of Pakistan was suspicious in September 2001; very few people in Afghanistan had satellite phones capable of making calls, and those who did were either terrorists or drug dealers. “You could figure out that if someone turned on their Iridium, bounced a call off an Intelsat satellite to a number in the United States, they probably had something to do with terrorism,” says Matthew Aid, a former NSA signals intelligence analyst turned historian who wrote The Secret Sentry, a definitive history of the agency.

  By Afghanistan’s standards, Pakistan had a fairly modern technological infrastructure. The NSA had spent a significant amount of money on a secret program to tap into the private cell phone network used by the country’s civilian, military, and intelligence leaders. The priority of U.S. intelligence was to collect as much information as possible about the Pakistani nuclear program; terrorism was a secondary concern. When thinking about the scope of potential domestic targets, the NSA ran into a problem of scale. In one of New York City’s “Little Pakistan” enclaves, about ten thousand immigrants speak a dialect of Urdu that the NSA could not properly decipher. Even if they had found some legal or technological way to monitor all international communications to telephone numbers and e-mail addresses associated with these Pakistani Americans, they could do nothing with the intercepts.

  One of the NSA’s foundational secrets is the result of two historical accidents. First, early worldwide telephone treaties made it much less expensive to route calls through the United States than through other, smaller countries. This made the telecommunications infrastructure an American creation—the product of American engineers and American equipment.
Second, the U.S. Department of Defense created the Internet. Americans like to think of ourselves as central to the smooth operation of the world, and so far as communications are concerned, we are.

  For the most part, telecommunication “switches,” or central hubs, where packets of data stream through are either physically located in the United States or built in foreign countries by U.S. companies.29 This is highly inefficient for someone in Latin America trying to send an e-mail to a friend in Europe, but it is a boon to the National Security Agency.30 According to NSA officials, before 9/11 as much as 85 percent of the world’s telecommunications traffic (cell phone calls, satellite calls, Internet traffic) coursed through a fiber optic cable inside the United States at some point during its transit. When it “hit the wire,” the NSA had the right to intercept it, provided there was no reason to believe that it was a purely domestic transmission. The NSA therefore found it fairly easy to convince U.S. telecom providers to allow them access to the international portion of that traffic, although, as we have noted, the technology to segregate the data was not very mature. Where it had gaps, it negotiated secret agreements with friendly countries like India, which contained the largest telecom switch for e-mail and phone calls for the Middle East.31

  The NSA also has a daring unit of secret wiretappers known as the Special Collection Service (SCS), or internally as “F6,” which operates in the field with CIA officers. In 1999, to give one example, the NSA tasked SCS with secretly tapping the communications network used by the Pakistani military to communicate with its civilian leadership about nuclear weapons.32

  This fiber optic idyll for American eavesdroppers contains many hidden thorns, however:

  How do you separate packets involving U.S. persons from packets that don’t, if it all flows through the same node?

  Many countries don’t want the United States to be able to listen in on their traffic, or don’t want their citizens to know that the United States has such easy access to it. As a result, these countries are often difficult to deal with.

  Telecommunications companies might not wish to cooperate. Because no law compels them to, there are gaps in data.

  The volume of raw signals generated by the world each day is increasing, and not even Google has the computing power to sample more than a fraction of it. How is the NSA supposed to know what is important?

  Routing calls and emails is an art of economy. Telecoms don’t operate according to the protocols of an intelligence agency. Rather, they’re primarily concerned about cost. Communications, and especially email, are commonly routed in groups without distinction, which makes it difficult at times to separate the domestic from the international.

  Because of the way email works, locating the origin of an email by IP address is difficult. Using the old Information Superhighway metaphor, email is like a motorcade. The computer disassembles files into a series of discrete vehicles, each of which contains a portion of the original entourage. (These vehicles are called packets.) But no worries: every car in this motorcade has the same map so no one will get lost. Each packet includes directions about where it needs to go; Internet hardware reads this metadata and determines the easiest route. Still, the motorcade often gets split up. There may be too much traffic on one road, forcing some packets to divert to another. As long as the packets don’t lose their metadata maps, the computer waiting for their arrival will be able to reassemble everything in the right order. The process of travel often makes it difficult to determine precisely where in the world the email came from, because each packet collects a different souvenir from all the hubs and switches it transits through. Spammers have figured out how to spoof the unique twelve-digit addresses that serve to identify the original computer.33 The NSA has an ISP Geolocation Cell that does nothing but track down the physical locations associated with IP addresses. Its computers are unable to do this automatically.

  Even before 9/11, collecting foreign intelligence required a massive amount of post facto deletion of data inadvertently collected about U.S. persons. And that’s okay. Congress accepted that in the course of daily SIGINT hunting, domestic traffic would creep into the bins. Provided that the NSA got rid of it quickly—unless there was some emergency that threatened someone’s life—eavesdroppers were none the worse for wear.

  But then events changed the NSA’s mission, and accordingly, its mind-set. Technology simply wasn’t ready for these changes, its progress being dependent on, but agnostic to, the goals of human beings. The agency lacked linguists, equipment, policy, and legal guidance. And it was under extreme pressure. The White House needed product the NSA could not produce. The SIGINT directorate would have to create databases of its own and massively reorient its collection program.

  After 9/11, The White House asked Hayden to treat any communication that terminated in Afghanistan or Pakistan as potentially intelligence-bearing. That meant the NSA had to figure out how to sift those communications from all the rest. It had to determine how best to geolocate email senders using Internet Protocol addresses, something that to this day it has trouble doing.34

  Email was easier in theory. The NSA could program its computers to search through electronic messages and maintain them with much greater ease than phone calls, which required significant server space. Immediately after 9/11, an NSA assessment concluded that terrorists would likely communicate with one another by embedding data in PDF files. So it proposed an expanded program to determine whether emails passing through its systems had PDF attachments. Its analysts would focus on those items first.

  The work was divided into compartments—perhaps as many as a dozen, each given their own classified code word.35 A technical team would figure out how to modify computers and equipment to allow for the type of collection that was required. One team of analysts would review all pocket litter (that is, things found in a terrorist’s possession) coming back from the battlefield. Another would try to use data-mining programs and statistical methods to search patterns of telephone calls and emails of specific targets. Suspicious sets were given probability scores, which, if high enough, triggered an interception. Another team of engineers worked to draw a map of the world’s telecommunications pipes to see whether there were any access points inside the United States that the NSA was not yet able to monitor.

  The NSA still had to sanitize the content of messages and phone calls—there was no way around U.S. law here.36 Only after the interception standard had been reached could the actual calls and emails themselves be monitored.

  The compartment of the program revealed by the New York Times allowed the NSA to intercept conversations between a U.S.-based target and a person overseas, provided that there was probable cause to believe that at least one side of the communication was involved in terrorism. Usually, the NSA knew the foreign target; it didn’t know the domestic target and used the bulk data analysis to figure out who it might be. (FISA, of course, required that if one side terminated in the United States, an order had to be issued, with few exceptions.) The collection had to be directed at a target, which under the new standard could be almost anyone or anything—an unknown person living in Detroit whose identity (but not a name) was collected from an al-Qaeda detainee in Afghanistan; a group of people; even people whose behavioral patterns resembled those of a terrorist. In other words, the “other side” of this war on terrorism could be anywhere, and that meant the program’s definition of a target could be anything that reasonably resembled an enemy.∗

  Regardless, it turned out to be difficult to accurately pipe exact conversations to the NSA analytical teams. Ironing out these technical wrinkles, which a former senior NSA official likened to a building engineer trying to figure out which valves to open to properly heat a building, meant that a lot of unrelated data was sent to the analysts. They ignored it (or “minimized” it), which is what they’re supposed to do when they encounter a U.S. person unrelated to the target.

  Because of technological limitations, getting to the target required t
he collection and conflation of many unconnected and innocent phone numbers and email or text message metadata. Some intercepted communications originated and terminated inside the United States.37 To the NSA, this was all right because nothing was done with those conversations once the correct “valve mix” was determined. To those later read in to the program, intention was now irrelevant; what mattered was the very fact of collection.

  President Bush acknowledged only one part of the program—the Terrorist Surveillance Program—after the New York Times publicly disclosed it. The president did not reveal the “other intelligence programs,” as a government report later called it, that he had ordered.

  It would be reasonable to assume that the TSP itself prompted hand-wringing and objections and almost, in 2004, led to the near resignations of the director of the FBI and the entire top turret of the Justice Department. But it was not. Something about the way these “other intelligence programs” were presented to Justice was the real problem.

  When acting attorney general James Comey was read in to the program for the first time, he relied on the advice of the OLC head, Jack Goldsmith, who had growing concerns about Yoo’s legal analysis, particularly with regard to the Other Intelligence Activities (OIAs)—not the TSP.38 One of the OIAs, for example, directly contravened a statute of Congress. (The statute in question was not FISA.) Comey has never publicly disclosed exactly what he objected to, but people briefed on the program and who have spoken to Comey say it was the legal rationale giving the NSA quick access to telecom-collected metadata that “drove him bonkers. There was just no way to justify this.”

 

‹ Prev