What We Can and Can’t Know About NSA Spying: A Reply to Prof. Cordero

Georgetown Law professor Carrie Cordero—who previously worked at the Department of Justice improving privacy procedures for monitoring under the Foreign Intelligence Surveillance Act—attended our event with Sen. Ron Wyden (D-OR) on the FISA Amendments Act last week.  Perhaps unsurprisingly, she’s rather more comfortable with the surveillance authorized by the law than our speakers were, and posted some critical commentary at the Lawfare blog (which is, incidentally, required reading for national security and intelligence buffs). Marcy Wheeler has already posted her own reply, but I’d like to hit a few points as well. Here’s Cordero:

Since at least the summer of 2011, [Wyden and Sen. Mark Udall] have been pushing the Intelligence Community to provide more public information about how the FAA works, and how it affects the privacy rights of Americans. In particular, they have, in a series of letters, requested that the Executive Branch provide an estimate of the number of Americans incidentally intercepted during the course of FAA surveillance. According to the exchanges of letters, the Executive Branch has repeatedly denied the request, on the basis that: i) it would be an unreasonable burden on the workforce (and, presumably, would take intelligence professionals off their national security mission); and ii) gathering the data the senators are requesting would, in and of itself, violate privacy rights of Americans.

The workforce argument, even if true, is, of course, a loser. The question of whether the data call itself would violate privacy rights is a more interesting one. Multiple oversight personnel independent of the operational and analytical wings of the Intelligence Community – including the Office of Management and Budget, the NSA Inspector General, and just last month, the Inspector General of the Intelligence Community, have all said that the data call requested by the senators is not feasible. The other members of the SSCI appear to accept this claim on its face. Meanwhile, Senator Wyden states he just finds the claim unbelievable. That there must be some way it can be done, he says, if even on a sample basis. Maintaining that position puts him in an interesting place, however: is the privacy advocate actually advocating for violating the privacy rules, to appease a Congressional request? Assuming that he would not actually want to advocate that the rules be waived at the request of a politician, a question then arises as to whether the Intelligence Community has adequately explained exactly how the data call would work and why it would conflict with existing privacy rules and protections, such as minimization procedures.

I’ll grant Cordero this point: as absurd as it sounds to say “we can’t tell you how many Americans we’re spying on, because it would violate their privacy,” this might well be a concern if those of us who follow these issues from the outside are correct in our surmises about what NSA is doing under FAA authority. The only real restriction the law places on the initial interception of communications is that the NSA use “targeting procedures” designed to capture traffic to or from overseas groups and individuals. There’s an enormous amount of circumstantial evidence to suggest that initial acquisition is therefore extremely broad, with a large percentage of international communications traffic being fed into NSA databases for later querying. If that’s the case, then naturally the tiny subset of communications later reviewed by a human analyst—because they match far narrower criteria for suspicion—is going to be highly unrepresentative. To get even a rough statistical sample of what’s in the larger database, then, one would have to “inspect”—possibly using software—a whole lot of the innocent communications that wouldn’t otherwise ever be analyzed. And possibly the rules currently in place don’t make any allowance for querying the database—even to analyze metadata for the purpose of generating aggregate statistics—unless it’s directly related to an intelligence purpose.

A few points about this.  First: assuming, for the moment, that  this is the case, why can’t NSA and DOJ say so clearly and publicly? Because it would somehow imperil national security to characterize the surveillance program even at this highest level of generality, without any mention of particular search parameters or targets? Would it “help the terrorists” if they answered a more recent query from a bipartisan group of senators, asking whether database searches (as opposed to initial “targeting”) had focused on specific American citizens?  Please.

A  more plausible hypothesis is that they recognize that an official, public acknowledgement that the government is routinely copying and warehousing millions of completely innocent communications—even if they’re only looking at the “suspicious” minority— would not go over entirely smoothly with the citizenry. There might even be a demand for some public debate about whether this is the kind of thing we’re willing to countenance. Legal scholars might become curious whether whatever arguments support the constitutionality of this practice hold up as well in the light of the day as they do when they’re made unopposed in closed chambers. Even without an actual estimate, any meaningful discussion of the workings of the program would be likely to undermine the whole pretense that it only “incidentally” involves the communications of innocent Americans, or that the constraints on “targeting”constitute a meaningful safeguard.  The desire to avoid the whole hornet’s nest using the pretext of national security is perhaps understandable, but it shouldn’t be acceptable in a democracy. Yet everyone knows overclassification is endemic—even the government’s own former “classification czar” has blasted the government’s use of inappropriate secrecy as a weapon against critics.

Second, transparency at this level of generality is an essential component of privacy protection. To the extent that the rules governing  access to the database preclude any attempt to audit its aggregate contents—including by automated software tallying of identifiers such as area codes and IP addresses—then they should indeed be changed, not because a senator demanded it, but because they otherwise preclude adequate oversight. An online service that keeps no server logs would be somewhat more protective of its users privacy… if  its database were otherwise perfectly secure against intrusion or misuse. In the real world, where there’s no such thing as perfect security, such a service would be protecting user privacy extremely poorly, because it would lack the ability to detect and prevent breaches. If it is not possible to audit the NSA’s system in this way, then that system needs to be altered until it is possible. If giving Congress a rough sense of the extent of the agency’s surveillance of Americans falls outside the parameters of the intelligence mission (and therefore the permissible uses of the database), it’s time for a new mission statement.

Finally, Cordero closes by noting the SSCI has touted its own oversight as “extensive” and “robust,” which Cordero thinks “debunks” the  suggestion embedded in our event title that the FAA enables “mass spying without accountability.”  (Can I debunk the debunking by lauding the accuracy and thoroughness of my own analysis?)  Unfortunately, the consensus of most independent analysts of the intelligence committees’ performance is a good deal less sanguine—which makes me hesitant to take that self-assessment at face value.

As scholars frequently point out, the overseers are asked to process incredibly complex information with a limited cleared staff to assist them, and often forbidden to take notes at briefings or remove reports from secure facilities. When you read about those extensive reports, recall that in the run-up to the invasion of Iraq only six senators and a handful of representatives ever read past the executive summary of the National Intelligence Estimate on Iraq’s WMD programs to the far more qualified language of the  full 92-page report. You might think the intel committees would need to hold more hearings than their counterparts to compensate for these disadvantages, but UCLA’s Amy Zegart has found that they consistently rank at the bottom of the pack, year after year. Little wonder, then, that years of flagrant and systemic misuse of another controversial surveillance tool—National Security Letters—was not uncovered by the “extensive” and “robust” oversight of the intelligence committees, but by the Justice Department’s inspector general.

In any event, we seem to have at least 13 senators who don’t believe they’ve been provided with enough information to perform their oversight role adequately. Perhaps they’re setting the bar too high, but I find it more likely that their colleagues—who over time naturally grow to like and trust the intelligence officials upon whom they rely for their information—are a bit too easily satisfied. There are no  prizes for expending time, energy, and political capital on ferreting out civil liberties problems in covert intelligence programs, least of all in an election year. It’s far easier to be satisfied with whatever data the intelligence community deigns to dribble out—often with heroic indifference to statutory reporting deadlines—and take it on faith that everything’s running as smoothly as they say. That allows you to write, and even believe, that you’re conducting “robust” oversight without knowing (as Wyden’s letter suggests the committee members do not) roughly how many Americans are being captured in NSA’s database, how many purely-domestic communications have been intercepted,  whether warrantless “backdoor” targeting of Americans is being done via the selection of database queries. But the public need not be so easily satisfied, nor accept that meaningful “accountability” exists when all those extensive reports leave the overseers ignorant of so many basic facts.