The law normally advances by small and cautious steps—by the gradual extension of established precedents and rules to novel problems and fact patterns. Little wonder, then, that tricky questions of law often amount to conflicts between competing metaphors. Is a hard drive like a closed briefcase whose contents are all fair game for police once the "container" is legitimately opened? Or is it more like a warehouse containing hundreds or thousands of individual closed containers? If the latter, what are the "containers"? Directories? Individual files?
A similar metaphor war figures in the FBI's effort to expand its authority to acquire information from Internet Service Providers using National Security Letters, which are issued by agents without judicial oversight, and typically forbid providers from disclosing anything about the demand for records. The Bureau had long assumed that the NSL statutes gave them broad authority to get "electronic communications transaction records"—information about your online communications, though not the contents of the communications themselves—as long as they certified that those records would be "relevant" to a national security investigation, a far lower standard than the Fourth Amendment's "probable cause." But in a 2008 opinion, the Bush administration's Office of Legal Counsel rejected this interpretation, finding that NSLs could only be used to obtain the particular types of records specified in the statute, including "toll billing records." For Internet accounts, this meant the FBI could only get "information parallel to... toll billing records for ordinary telephone service."
The obvious question is what, exactly, constitutes information "parallel to" a toll billing record in the online context. The FBI would prefer to resolve the ambiguity by simply amending the law to give them blanket authority to acquire transaction records. In particular, according to The Washington Post, government lawyers think they can obtain "the addresses to which an Internet user sends e-mail; the times and dates e-mail was sent and received; and possibly a user's browser history." On its face, this sounds like a reasonable reading. An important 1979 Supreme Court case, Smith v. Maryland, held that the information contained in telephone "toll billing records"—the itemized list of calls placed and received you'd find on a standard phone bill—didn't enjoy Fourth Amendment protection, and so unlike the contents of phone conversations themselves, could be obtained by the government without a full probable cause warrant. Surely the obvious equivalent in the online context is the list of e-mail addresses in an Internet user's inbox and outbox? At a second glance, though, there are some problems with that metaphor, of two central kinds.
First, there's a problem with the formal analogy. The Court in Smith supported their finding of a diminished privacy interest in toll billing records on numerous grounds. For one, the Court noted that because one's itemized phone bill did contain these numbers, no reasonable person could be unaware that this information was "exposed" to employees of the phone company and retained as a matter of course among the company's business records. Of course, it's now increasingly common for phone companies to charge a flat rate rather than billing by individual calls, and so the legislative history of the NSL statutes makes clear that by "toll billing records" they mean information that could be used to assess a charge, even if a company happened not to charge that way.
The analogy gets pretty strained when we come to Internet services, though. At the time the laws in question here were written, ISPs almost universally charged people for the amount of time they were connected, not by the number of individual e-mails sent. Now it's much more common to simply play a flat monthly fee for broadband connection, though you also sometimes see plans where there's a charge by the megabyte above a certain threshold of bandwidth usage. Your ISP, of course has technical access to the list of e-mail addresses you've communicated with—just as they have the ability to access the e-mails themselves—but no major service, as far as I know, has ever actually kept this list as a separate billing record.
But maybe that's not the right way to apply the metaphor. Maybe what's important is whether those to/from e-mail records are substantively "parallel to" the kind of information you'd traditionally find in telephone toll billing records. As the Smith Court observed, a list of phone numbers was far less revealing and sensitive than the actual conversation—it revealed nothing of the "purport" of the communication itself, or even who was on the call. But as soon as we start to think more carefully about how we actually use e-mail in the real world, it becomes clear that the analogy is far from perfect.
One thing lots of people do with e-mail, after all, is participate in mailing lists and discussion groups. Records of this sort, then, are likely to reveal the membership in potentially controversial social, political, or religious groups—and the Supreme Court has also found that such membership lists enjoy First Amendment protection as a component of freedom of association. But they'd also reveal much more than that. The closest telephone analogue to a mailing list discussion is probable a conference call. An investigator who obtained toll billing records for such a call would, at most, have learned that a certain number of people called in for a certain amount of time; they'd learn nothing about who spoke in response to whom, or how much, and who remained silent. Someone getting e-mail transaction records would have a much more detailed picture of who was vocal and who was silent, the order and frequency with which participants spoke, and so on. And more generally, people in practice do not use e-mail like traditional letters: They tend to have exchanges in which each individual e-mail is more like a piece of the longer conversation.
There are also many common uses of e-mail that don’t really have close analogies in the telephonic context. If I make a purchase from Amazon, win an Ebay auction, make an OpenTable restaurant reservation, register for a conference at a local think tank, or place a Craigslist ad, that will typically generate an automatic confirmation e-mail from the site, and the e-mail address from which the site comes will often reveal something about the nature of the transaction. (My inbox has messages from auto-confirm, order-update, ship-confirm, and store-news @amazon.com—inherently more revealing than the mere fact that I called some mail-order vendor.) It's not a particularly big deal in those cases, but such e-mails could also reveal that I had opened or closed or modified an account at a particular politically, sexually, or religiously oriented Web site, or subscribed to a specific publication.
For an example of just how sensitive and revealing such task-specific e-mail addresses can be, consider Craigslist in particular. The site—which for those who haven't used it is the vast online equivalent of the newspaper's classified section—generates an individual anonymized e-mail addresses for each ad placed, so that users don't have to expose their own contact information to the world. Yet while this provides anonymity against the general public, it also makes those mere e-mail addresses much more revealing to the government agent who obtains transaction records. That's because each ad can be linked to a particular e-mail address, so if you've sent a message to pers-1234567-ABCD@craigslist.com, the government may not know exactly who you've written, but they can determine why you're writing: To respond to an ad offering a handgun for sale, say, or one soliciting a foot fetishist for a "casual encounter."
The point is not just that investigators shouldn't be able to get e-mail transaction records without a probable cause warrant—though I happen to think that would be a reasonable standard. It's that metaphors can mislead us: We need to look past the easy equivalencies between new technologies and more traditional forms of communication, and drill down to see the full range of privacy interests implicated given the real-world practices of ordinary people who use those technologies.