We should be careful not to put the NSA in an impossible position. Of course, we should be vigilant against the administrative state in all of its tangled tendrils, especially its collection of taxes (the IRS scandal) and enforcement of the laws (Obama’s refusal to enforce Obamacare and immigration law). The problem here, however, is that we are placing these kinds of domestic law-enforcement standards on a foreign intelligence function. With domestic law enforcement, we want the Justice Department to monitor one identified target (identified because other evidence gives probable cause that he or she has already committed a crime) and to carefully minimize any surveillance so as not to intrude on privacy interests.
Once we impose those standards on the military and intelligence agencies, however, we are either guaranteeing failure or we must accept a certain level of error. If the military and intelligence agencies had to follow law-enforcement standards, their mission would fail because they would not give us any improvement over what the FBI could achieve anyway. If the intelligence community is to detect future terrorist attacks through analyzing electronic communications, we are asking them to search through a vast sea of e-mails and phone-call patterns to find those few which, on the surface, look innocent but are actually covert terrorist messages. If we give them broader authority, we would have to accept a level of error that is inherent in any human activity. No intelligence agency could perform its mission of protecting the nation’s security without making a few of these kinds of mistakes. The question is whether there are too many, not whether there will be any at all.
Domestic law enforcement makes these errors too. Police seek warrants for the wrong guy, execute a search in the wrong house, arrest the wrong suspect, and even shoot unarmed suspects. We accept these mistakes because we understand that no law-enforcement system can successfully protect our communities from crime with perfection. The question is the error rate, how much it would cost to reduce it, the impact on the effectiveness of the program, and the remedies we have for mistakes. Consider those questions in the context of the NSA surveillance program. The more important question is not the top of the fraction but the bottom — not just how many mistakes occurred, but how many records were searched overall. If there were 2,000 or so mistakes, as the Washington Post suggests, but involving billions of communications, the error rate is well less than 1 percent. Without looking at the latest figures, I suspect that is a far lower error rate than those turned in by domestic police on searches and arrests.
To end the NSA’s efforts to intercept terrorist communications would be to willfully blind ourselves to the most valuable intelligence sources on al-Qaeda (now that the president won’t allow the capture and interrogation of al-Qaeda leaders). The more useful question is whether there is a cost-effective way to reduce the error rate without detracting from the effectiveness of the program, which, by General Keith Alexander’s accounting, has been high. Increasing judicial oversight might reduce errors — though I am dubious — but in a way that would seriously slow down the speed of the program, which is all-important if the mission is to stop terrorists. And perhaps Congress should think about ways to remedy any privacy violations in the future. But to end the program because it does not have an error rate of zero is to impose a demand on the NSA that no other government program, foreign or domestic, military or civilian, could survive.