Abstract

The providers of network services (and the makers of network devices) know an enormous amount about our lives. Because they do, these network intermediaries are being asked with increasing frequency to assist the government in solving crimes or gathering intelligence. Given how much they know about us, if the government can secure the assistance of these intermediaries, it will enjoy a huge increase in its theoretical capacity for surveillance—the ability to learn almost anything about anyone. This has the potential to create serious social harm, even assuming that the government continues to adhere to ordinary democratic norms and the rule of law.

One possible solution to this problem is for network intermediaries to refuse government requests for aid and attempt to sustain those refusals in court. Although this proposal has received an enormous amount of attention, there is substantial cause for skepticism about how well it can work. Congress has given the government wide authority to demand information and assistance through tools like subpoenas, the Stored Communications Act, and Title III. Even when the government does not have specific statutory authorization, courts have interpreted the All Writs Act to authorize a great deal of open-ended aid, consistent with the well-settled Anglo-American history of third-party assistance in law enforcement. It is also far from unheard of for the executive to read restrictions on its surveillance authority narrowly, and its own inherent powers broadly, to engage in surveillance that is quasi- or extra-legal.

A superior (or at least complementary) response to the problem is to restrict network intermediaries themselves by limiting how much they can learn about us and how long they can retain it. This approach treats enhanced state surveillance as a problem created by the intermediaries’ stockpiling of data, and proposes to solve it at the root—which would, as a useful side effect, solve a number of other problems created by that stockpiling, too.