Last week, precisely 10 days after USA F-Redux — with its different formulas allowing for provider transparency –passed, Amazon released its first transparency report. In general, the report shows that Amazon either doesn’t retain — or successfully pushes back — against a lot of requests. For example, Amazon provided no or only partial information to a third of the 813 subpoenas it received last year.
Also of note, in a post accompanying the report, Stephen Schmidt claimed that “Amazon never participated in the NSA’s PRISM program,” which may not be all that surprising given that it has only received 25 non-national security search warrants.
As I’ve already suggested, I find the most interested detail to be the timing: given that Amazon has gotten crap as the only major company not to release a transparency report before, I suspect either that Amazon had a new application 2 years ago when everyone started reporting, meaning it had to wait until the new collection had aged under the reporting guidelines, or something about the more granular reporting made the difference for Amazon. Amazon reported in the 0-250 range (including both NSLs and other FISA orders), so it may just have been waiting to be able to report that lower number.
That said, Amazon received 13 non-national security court orders (aside from the one take down order they treat separately, which I believe has to do with an ISIL site), only 4 of which they responded fully to. I think this category would be where Amazon would count pen registers. And I’d expect Amazon to get pen registers in connection with their hosting services. If any of the 0 to 250 National Security orders are pen registers, it could be fairly intrusive.
Finally, Amazon clarified (sort of) something of particular interest. While Amazon makes clear that content stored in a customer’s site is content (self-evident, I know, but there are loopholes for stored content, which is a big part of why Amazon would be of interest (and was when Aaron Swartz was using them as a hosting service).
Non-content. “Non-content” information means subscriber information such as name, address, email address, billing information, date of account creation, and certain purchase history and service usage information. Content.
“Content” information means the content of data files stored in a customer’s account.
But Amazon doesn’t include “certain purchase history information” to be content.
As the country’s biggest online store, that’s where Amazon might be of the most interest. Indeed, in the legal filings pertaining to Usaamah Abdullah Rahim (the claimed ISIL follower whom Boston cops shot and killed on June 2) show they were tracking Rahim’s Amazon purchase of a knife very closely.
If you wanted to do a dragnet of purchase records, you’d include Amazon in there one way or another. And such a dragnet order might represent just one (or four) of the fewer than 250 orders Amazon got in a year.
It’s not surprising they’re treating (“certain”) purchase records as metadata. But it is worth noting.