In my last post, I discussed why it is so important for platforms to be able to speak about the discovery demands they receive, seeking to unmask their anonymous users. That candor is crucially important in ensuring that unmasking demands can’t damage the key constitutional right to speak anonymously, without some sort of check against their abuse.
The earlier post rolled together several different types of discovery instruments (subpoenas, warrants, NSLs, etc.) because to a certain extent it doesn’t matter which one is used to unmask an anonymous user. The issue raised by all of them is that if their power to unmask an anonymous user is too unfettered, then it will chill all sorts of legitimate speech. And, as noted in the last post, the ability for a platform receiving an unmasking demand to tell others it has received it is a critical check against unworthy demands seeking to unmask the speakers behind lawful speech.
The details of each type of unmasking instrument do matter, though, because each one has different interests to balance and, accordingly, different rules governing how to balance them. Unfortunately, the rules that have evolved for any particular one are not always adequately protective of the important speech interests any unmasking demand necessarily affects. As is the case for the type of unmasking demand at issue in this post: a federal grand jury subpoena.
Grand jury subpoenas are very powerful discovery instruments, and with good reason: the government needs a powerful weapon to be able to investigate serious crimes. There are also important constitutional reasons for why we equip grand juries with strong investigatory power, because if charges are to be brought against people, it’s important for due process reasons that they have been brought by the grand jury, as opposed to a more arbitrary exercise of government power. Grand juries are, however, largely at the disposal of government prosecutors, and thus a grand jury subpoena essentially functions as a government unmasking demand. The ability to compel information via a grand jury subpoena is therefore not a power we can allow to exist unchecked.
Which brings us to the story of the grand jury subpoena served on Glassdoor, which Paul Levy and Ars Technica wrote about earlier this year. It’s a story that raises three interrelated issues: (1) a poor balancing of the relevant interests, (2) a poor structural model that prevented a better balancing, and (3) a gag that has made it extraordinarily difficult to create a better rule governing how grand jury subpoenas should be balanced against important online speech rights.
Glassdoor is a platform focused on hosting user-provided information about employers. Much of the speech it hosts is necessarily contributed anonymously so that the speakers can avoid any fallout from their candor. This is the sort of fallout that, if they had to incur it, would discourage them from contributing information others might find valuable. The seriousness of these sorts of consequences is why the district court decision denying Glassdoor’s attempts to resist the grand jury subpoena seeking to unmask their users reflects such a poor balancing of the relevant interests. Perhaps if the subpoena had been intended to unmask people the government believed were themselves guilty of the crime being investigated, the balance might have tipped more in favor of enforcing it. But the people who the subpoena was seeking to unmask were simply suspected as possibly knowing something about the crime that others were apparently committing. It is not unreasonable for the government to want to be able to talk to witnesses, but that desire to talk to them is not the only interest present here. These are people who were simply availing themselves of their right to speak anonymously, and who, if this subpoena is enforced, are going to be shocked to suddenly find the government on their doorstep wanting to talk to them.
This sort of unmasking is chilling to them and anyone else who might want to speak anonymously because it means that there’s no way they ever will be able to speak should their speech happen to ever somehow relate (however tangentially) to someone else’s criminal behavior. It is also inconsistent with the purported goal of fighting crime because it will prevent criminal behavior from coming to light in the first place, for few will want to offer up information if it will only tempt trouble for them at some point in the future.
This mis-balancing of interests is almost a peripheral issue in this case, however. The more significant structural concern is why such a weak balancing test was used. As discussed previously, in order to protect the ability to speak anonymously online, it is important for a platform to be able to resist demands to unmask their users in cases where the reason for the unmasking does not substantially outweigh the need to protect people’s right to speak anonymously online. But the district court denied Glassdoor’s attempt to resist the subpoena when it chose to apply the test from Branzburg v. Hayes, a Supreme Court case focused on the ability to resist a grand jury subpoena. Branzburg, however, has nothing to do with the Internet or Internet platforms. It is a case from the 1970s that was solely focused on whether the First Amendment gave journalists the right to resist a grand jury subpoena. Ultimately it decided that they generally had no such right, at least so long as the government was not shown to be acting in bad faith, which, while not nothing, is not a standard that is particularly protective of anonymity. It also barely even addressed the interests of the confidential sources themselves, dismissing their interest in maintaining anonymity as a mere “preference,” and one the Court presumed was being sought only to shield themselves from prosecution for their own criminal culpability.
The upshot of Branzburg is that the journalist, as an intermediary for a source’s information, had no right to resist a grand jury subpoena. Unfortunately, Branzburg simply can’t be extended to the online world where, for better or worse, essentially all speech must be intermediated by some sort of platform or service in order to happen. The need to let the platforms resist grand jury subpoenas therefore has less to do with whether an intermediary itself has a right to resist them and everything to do with the the right of their users to speak anonymously, which, far from being a preference, is an affirmative right the Supreme Court, after Branzburg, subsequently recognized.
A better test, and one that respects the need to maintain this critical speech right, is therefore needed, which is why Glassdoor appealed the district court’s ruling. Unfortunately, its appeal has raised a third issue: while there is often a lot of secrecy surrounding a grand jury investigation, in part because it makes sense to keep the subject of an investigation in the dark, preserving that level of secrecy does not necessarily require keeping absolutely everything related to the subpoena under seal. Fortunately the district court (and the DOJ, who agreed to this) recognized that some information could safely be released, particularly related to Glassdoor’s challenge of the subpoena’s enforcement generally, and thanks to that limited unsealing we can tell that the case involved a misapplication of Branzburg to an Internet platform.
Unfortunately the Ninth Circuit didn’t agree to this limited disclosure and sealed the entirety of Glassdoor’s appeal, even the parts that were already made public. The effects of this sealing included that it became impossible for potential amici to weigh in in support of Glassdoor and to argue for a better rule that would allow platforms to better protect the speech rights of their users. While Glassdoor had been ably litigating the case, the point of amicus briefs is to help the court see the full implications of a particular ruling on interests beyond those immediately before it, which is a hard thing for the party directly litigating to do itself. The reality is that Glassdoor is not the first, and will not be the last, platform to get a grand jury subpoena, but unless the rules governing platforms’ ability to resist are stronger than what’s afforded by Branzburg, the privacy protection speakers have depended on will continue to evaporate should their speech ever happen to capture the interest of a federal prosecutor with access to grand jury.
For all we know, of course, the Ninth Circuit might have seen its point and quashed the subpoena. Or maybe it upheld it and maybe the FBI has now unpleasantly surprised those Glassdoor users. We may never know, just as we may never know if there are other occasions where courts have used specious reasoning to allow grand jury subpoenas to strip speakers of their anonymity. Even if the Ninth Circuit indeed fixed the problems with this questionable attempt at unmasking, by doing it in secret it’s missed an important opportunity to provide guidance to lower courts to help ensure that they don’t allow other questionable attempts to keep happening to speakers in the future.