In response to a lawsuit first filed eight years ago by the Electronic Privacy Information Center, the  government has finally released a less-heavily-redacted version of Jack Goldsmith’s memorandum approving the legality of the STELLARWIND surveillance program authorized by President Bush shortly after the 9/11 attacks. Though it is less reminiscent of Swiss cheese than a previously released version, however, Charlie Savage writes in the New York Times that the remaining redactions leave some big mysteries in place:

However, the government continued to redact crucial portions of the memo that would answer a primary remaining question about the history of Stellarwind: What prompted the Justice Department to conclude in early 2004 that one aspect of the program, which collected records about Americans’ emails in bulk, was illegal — even though it permitted other aspects, like warrantless wiretapping and the bulk collection of Americans’ phone records, to continue?

Pace Charlie, I think that it’s actually pretty easy to infer from the new memorandum precisely what the problem must have been. In fact, I laid out my theory in a piece for Ars Technica last summer without relying on the Goldsmith memo, but this newly released version  makes me fairly certain that my analysis there was basically correct.  In brief: The problem was that the bulk collection of Internet metadata fell within FISA’s definition of “electronic surveillance,” which the bulk collection of telephony metadata did not. And while Goldsmith argued that the targeted collection of content in violation of FISA could be justified by a combination of the president’s inherent constitutional authority and the Authorization for the Use of Military Force (AUMF) passed by Congress, his arguments to this effect are all conspicuously limited to surveillance directed at “the enemy,” whereas STELLARWIND’s collection of metadata was not targeted in this way. This component therefore fell within the statutory prohibition on “electronic surveillance” under color of law outside the FISA structure, but beyond the scope of the president’s supposed authority—which Goldsmith derives from both the AUMF and Article II —to ignore those prohibitions when targeting “the enemy” during wartime.

I lay out this argument in greater detail in the Ars Technica piece, but I’ll recapitulate it here drawing mainly on evidence from the Goldsmith memo, and proceeding by process of elimination.

We know STELLARWIND collected four main categories of information: phone content, Internet content, phone metadata, and Internet metadata. We know that the Internet metadata component was the controversial part, which James Comey objected to so strenuously that he was prepared to resign if it was not modified. The memo tells us that the basis of this objection cannot have been Fourth Amendment considerations, because Goldsmith (wrongly) believes that “Acquisition of Meta Data Does Not Implicate the Fourth Amendment,” explicitly opining that e-mail metadata is categorically beyond the protection of the Fourth Amendment—a mistaken argument I’ll return to in a subsequent post.

Therefore Goldsmith’s objection here must have been statutory. As he observes, FISA makes “electronic surveillance under color of law” a felony unless the procedures of FISA or the Wiretap Act are adhered to. The telephony metadata program ran no risk of running afoul of this prohibition, because it involved only the production of call records created by phone companies in the ordinary course of business, and not the “installation or use” of any “device”—meaning it did not constitute “electronic surveillance” as defined by FISA.  But Internet backbone providers, unlike phone companies or e-mail providers, don’t have any need to process e-mail metadata, let alone create any business records containing that information. The only way to obtain e-mail metadata from the Internet backbone,  then, would be via “electronic surveillance” of Internet traffic, triggering FISA’s prohibition.  So the Internet metadata component of STELLARWIND (unlike the phone metadata component) would be presumptively illegal under FISA unless it fell within the scope of either the AUMF passed by Congress or the president’s core constitutional powers as commander-in-chief, which Goldsmith claims cannot be constrained by statute.

As Goldsmith observes in a footnote, John Yoo had earlier offered a far more ambitious argument, effectively asserting that the Fourth Amendment is a dead letter in wartime, and that the president enjoys near plenary discretion to ignore Congress while wearing his Commander-in-Chief hat.  As we now know, this was an argument his successors did not find terribly impressive.  The issue with STELLARWIND’s e-mail metadata component, then, arose with the rejection of Yoo’s radical theories and the shift to reliance on the AUMF combined with a slightly more restrained view of the president’s preclusive wartime powers. To identify the problem, then, we just need to ask what differences between the content and metadata components would account for these sources of authority overriding FISA’s prohibition on electronic surveillance in the case of the former, but not the latter.

Goldsmith explicitly argues that both the AUMF and the president’s Article II powers are sufficient to support the content collection components of STELLARWIND—but the unredacted portions of the memo conspicuously limit his arguments to those two portions. He does not, in other words, ever argue in those unredacted portions that either source of authority would permit the president to order STELLARWIND’s metadata collection. If nothing else, the fact that the two types of content collection are lumped together, yet treated seperately from the two types of metadata collection, tells us that Goldsmith thought they required distinct analysis: The same arguments did not apply to all the components in the same way.

A close look at Goldsmith’s arguments along these lines makes it obvious why that would be the case.  The AUMF empowers the president to use “all necessary and appropriate force” against “those nations, organizations, or persons he determines planned, authorized, committed, or aided the terrorist attacks” of 9/11. At the time Goldsmith wrote his memo, STELLARWIND’s content collection was targeted against such parties; but the metadata programs were not. Supposing that “necessary and appropriate force” includes domestic surveillance, the AUMF still could not easily be read to authorize surveillance of persons with no suspected connection to the 9/11 attacks.   Similarly, Goldsmith argues  specifically that FISA would be unconstitutional if construed to limit the president’s powers “in the narrow context of interception of enemy communications in the midst of an armed conflict,” and that “congressional efforts to prohibit the President’s efforts to intercept enemy communications through STELLAR WIND would be an unconstitutional encroachment on the Commander-in-Chief power.”

Moreover, in attempting to distinguish STELLARWIND form historical cases where courts have held the president’s wartime power to be subject to congressional restrictions, Goldsmith stresses the targeted nature of content collection under the program.  He considers Little v. Barreme, involving a statute which authorized the president to seize American merchant ships bound to but not from French ports during the “Quasi War” with France of 1799. The Supreme Court held that a presidential order permitting seizure of ships both to and from French ports was therefore invalid under the statute. But Goldsmith rejects the relevance of that precedent because “the operative section of the statute in question restricted the movements of and granted authority to seize American merchant ships. It was not a provision that purported to regulate by statute the steps the Commander in Chief could take in confronting armed vessels of the enemy.” Goldsmith argued that this distinction was “particularly important when the content collection aspect of STELLAR WIND is under consideration, because content collection is directed solely against targeted telephone numbers of e-mails where there is a reason for believing that one of the communicants is an enemy.”  The obvious but unstated implication of this argument, however, is that surveillance not directed against “the enemy” in this way could not be similarly distinguished from Barreme.

Combining all these elements, the upshot of Goldsmith’s objection to STELLARWIND’s original e-mail metadata component is not too hard to infer: It constituted “electronic surveillance” prohibited by Congress (which the telephone program was not), and not targeted enough (indeed, it was not targeted at all) to fall within the scope of either the AUMF or an inherent presidential power to engage “the enemy.”  Mystery—I think—solved.