Steph: This is an excerpt from Permanent Record by Edward Snowden, beginning on page 175.

It was only later, long after I’d forgotten about the missing Inspector General report, that the classified version came skimming across my desktop, as if in proof of that old maxim that the best way to find something is to stop looking for it. Once the classified version turned up, I realized why I hadn’t had any luck finding it previously: it couldn’t be seen, not even by the heads of agencies. It was filed in an Exceptionally Controlled Information (ECI) compartment, an extremely rare classification used only to make sure that something would remain hidden even from those holding top security clearance. Because of my position, I was familiar with most of the ECIs at the NSA, but not this one. The report’s full classification designation was TOP SECRET//STLW//HCS/COMINT//ORCON/NOFORN, which translates to: pretty much only a few dozen people in the world are allowed to read this.

I was most definitely not one of them. The report came to my attention by mistake: someone in the NSA IG’s office had left a draft copy on a system that I, as a sysadmin, had access to. Its caveat of STLW, which I didn’t recognize, turned out to be what’s called a “dirty word” on my system: a label signifying a document that wasn’t supposed to be stored on lower-security drives. These drives were being constantly checked for any newly appearing dirty words, and the moment one was found I was alerted so that I could decide how best to scrub the document from the system, but before I did, I’d have to examine the offending file myself, just to confirm that the dirty word search hadn’t flagged anything accidentally. Usually I’d take just the briefest glance at the thing. But this time, as soon as I opened the document and read the title, I knew I’d be reading it all the way through.

Here was everything that was missing from the unclassified version. Here was everything that the journalism I’d read had lacked, and that the court proceedings I’d followed had been denied: a complete accounting of the NSA’s most secret surveillance programs, and the agency directives and Department of Justice policies that had been used to subvert American law and contravene the US Constitution. After reading the thing, I could understand why no IC employee had ever leaked it to journalists, and no judge would be able to force the government to produce it in open court. The document was so deeply classified that anybody who had access to it who wasn’t a sysadmin would be immediately identifiable. And the activities it outlined were so deeply criminal that no government would ever allow it to be released unredacted.

One issue jumped out at me immediately: it was clear that the unclassified version I was already familiar with wasn’t a redaction of the classified version, as would usually be the practice. Rather, it was a wholly different document, which the classified version immediately exposed as an outright and carefully concocted lie. The duplicity was stupefying, especially given that I just dedicated months of my time to deduplicating files. Most of the time, when you’re dealing with two versions of the same document, the differences between them are trivial—a  few commas here, a few words there. But the only thing these two particular reports had in common was their title.

Whereas the unclassified version merely made reference to the NSA being ordered to intensify its intelligence-gathering practices following 9/11, the classified version laid out the nature, and scale, of that intensification. The NSA’s historic brief had been fundamentally altered from target collection of communications to “bulk collection,” which is the agency’s euphemism for mass surveillance. And whereas the unclassified version obfuscated this shift, advocating for expanded surveillance by scaring the public with the specter of terror, the classified version made this shift explicit, justifying it as the legitimate corollary of expanded technological capacity.

The NSA IG’s portion of the classified report outlined what it called “a collection gap,” noting that existing surveillance legislation (particularly the Foreign Intelligence Surveillance Act) dated from 1978, a time when most communications signals traveled via radio or telephone lines, rather than fiber-optic cables and satellites. In essence, the agency was arguing that the speed and volume of contemporary communication had outpaced, and outgrown, American law—no court, not even a secret court, could issue enough individually targeted warrants fast enough to keep up—and that a truly global world required a truly global intelligence agency. All of this pointed, in the NSA’s logic, to the necessity of the bulk collection of Internet communications. The code name for this bulk collection initiative was indicated in the very dirty word that got it flagged on my system: STLW, an abbreviation of STELLARWIND. This turned out to be the single major component of the PSP that had continued, and even grown, in secret after the rest of the program had been made public in the press.

STELLARWIND was the classified report’s deepest secret. It was, in fact, the NSA’s deepest secret, and the one that the report’s sensitive status had been designed to protect. The program’s very existence was an indication that the agency’s mission had been transformed, from using technology to defend America to using technology to control it by redefining citizens’ private internet communications as potential signals intelligence.

Such fraudulent redefinitions ran throughout the report, but perhaps the most fundamental and transparently desperate involved the government’s vocabulary: STELLARWIND had been collecting communications since the PSP’s inception in 2001, but in 2004—when Justice Department officials balked at the continuation of the initiative—the Bush administration attempted to legitimize it ex post facto by changing the meanings of basic English words, such as “acquire” and “obtain.” According to the report, it was the government’s position that the NSA could collect whatever communications records it wanted to, without having to get a warrant, because it could only be said to have acquired or obtained them, in the legal sense, if and when the agency “searched for and retrieved” them from its database.

This lexical sophistry was particularly galling to me, as I was well aware that the agency’s goal was to be able to retain as much data as it could for as long as it could—for perpetuity. If communications records would only be considered definitively “obtained” once they were used, they could remain “unobtained” but collected in storage forever, raw data awaiting its future manipulation. By redefining the terms “acquire” and “obtain”—from describing the act of data being entered into a database, to describing the act of a person (or, more likely, an algorithm) querying that database and getting a “hit” or “return” at any conceivable point in the future—the US government was developing the capacity of an eternal law-enforcement agency. At any time, the government could dig through the past communications of anyone it wanted to victimize in search of a crime (and everybody’s communications contain evidence of something). At any point, for all perpetuity, any new administration—any future rogue head of the NSA—could just show up to work and, as easily as flicking a switch, instantly track everybody with a phone or a computer, know who they were, where they were, what they were doing with whom, and what they had ever done in the past.


Permanent Record by Edward Snowden

Published by Metropolitan Books, Henry Holt and Company New York ( 2019)

Leave a Reply

Your email address will not be published. Required fields are marked *