I think Orin's hit the nail on the head: It's all about the filters. Whether and to what extent the NSA practices are desirable, and constitutional, depends on what the filters are designed to do, and how well they work.
I think we all agree with Orin that, in one way or another, the NSA has "figured out settings that accurately identify foreign traffic," in the sense of being able to identify calls and e-mails that have a foreign address on at least one end. And I don't really have much of a problem, in theory anyway, with letting the NSA computers freely take a first cut to separate those international communications from purely domestic communications, without the need for a FISA court order.
The concern, however, is that then the NSA finds itself in possession of not only a slew of wholly foreign communications, but also a bunch of foreign-to-domestic communications—indeed, perhaps virtually all of our international phone calls and e-mails.
What's the next-level filter for those? Can we really evaluate the program, Orin asks, "without knowing the filter and how accurate it is?"
No, we can't.
But we can evaluate the statute based on how accurate it requires the NSA filters to be. And that's the problem with the new law. Because according to the Protect America Act, it doesn't matter how "accurate" the next filter is or even whether the NSA uses any filter at all. So long as the monitoring is "directed at" one party who is reasonably believed to be overseas, all bets are off.
This is about as far as one can imagine from Orin's hypothetical perfect filter that "passes on only calls and e-mails involving terrorists discussing their plans." Of course, he's right that we can't insist on such unattainable precision. But the Protect America Act goes to the opposite extreme: There's no requirement that the surveillance have anything to do with terrorism—or any crime for that matter. No requirement that the surveillance even be related to national security. No requirement for what the NSA must demonstrate before it listens in on the U.S. end of the calls and e-mails. No limitation on sharing the information with other government agencies. No restriction on maintaining files on Americans based on what the NSA discovers in our international calls and e-mails.
So, for example, what if the government decides that it doesn't much appreciate a human rights organization in London that is having a significant influence on public debate about U.S. interrogation practices? And so the NSA intercepts all international calls and e-mails involving that organization—including all calls and e-mails to or from the United States. Obviously, this surveillance will uncover a great deal of information conveyed by U.S. persons who have frequent contacts with the London organization.
Under the Protect America Act, what's to stop the NSA from collecting and using all that it learns about the Americans who are parties to those phone calls and e-mails? Surveillance under the Protect America Act is just about the polar opposite of Orin's Little Nicky Scarfo example. In that case, the government's monitoring was limited to a single computer keyboard, because a court had been convinced that there was a mobster at that keyboard—and the government was required to jettison the information that it found that was immaterial to its organized-crime investigation.
Under the new law, by contrast, the NSA can indiscriminately monitor all domestic-to-foreign communications for any reason relating to U.S. foreign policy, without any showing to a court, without any court supervision, and without any meaningful constraints on what it can do with the information about U.S. persons that it obtains.
How's that for filtering?