Skip to content

Thoughts on Gul

By
Sunday, July 24, 2011 at 8:34 AM

I have now had a chance to read Gul, the other D.C. Circuit case that came down on Friday. Gul establishes a proposition that, in my opinion at least, should be pretty obvious: that Guantanamo habeas jurisdiction does not survive the release or transfer of the detainee from Guantanamo. I’m not sure that many people had much doubt that this was the direction the doctrine would go. But the Supreme Court case law, which allows habeas jurisdiction to sustain itself sometimes based on the collateral consequences of a criminal conviction even after a release, left at least notional doubt on that point. The appeals court has now, absent a most-improbable cert grant, removed that doubt.

While the decision is no surprise, it is important, if only because a decision in the other direction would have been an earthquake. If habeas cases were not mooted by transfers, the government would potentially face hundreds of additional habeas suits from people no longer in custody but who wished to clear their names–or, more minimally, who didn’t mind if their lawyers kept litigating their cases. The result would potentially have been that bringing someone to Guantanamo would subject the government to years of litigation, even if the government relatively swiftly decided to let the person go. That would present another enormous disincentive to use the facility–or any similar U.S. detention site that might ever develop. I suspect that it would also function subtly as a disincentive for transfers themselves. Getting rid of litigation in which the government might not prevail, after all, is one of the reasons the government transfers people. If the government is going to have to litigate their cases anyway, why not get the marginal security benefit of the detention that’s at issue in the litigation?

With this clarification from the D.C. Circuit, by contrast, transfer simply makes a case go away, and perhaps just as important, it makes it go away without either side’s having to win or lose. The result of this rule in practice–for this has been the default rule since the habeas cases started–has been a large number of cases in which detainees have walked free but in which the government does not concede that they were unreasonably detained. And no mechanism to resolve whether their detention was reasonable or not exists. This is frustrating, no doubt, for detainees who want not merely freedom but some vindication of their claims that their detention itself was in error or otherwise unlawful. It is also frustrating intellectually for those of us who would like some means of accounting for what the real error rate has been in detention cases. But the frustration is a reasonable price, in my judgment anyway, for a system that encumbers transfers as little as possible and that does not needlessly burden the government with wartime litigation.

One other note unrelated note on Gul–and on Al Alwi, on which I commented yesterday. Both decisions are unanimous, continuing a long string of D.C. Circuit detainee cases in which the court has at least agreed about cases’ proper outcomes. There has been a lot of attention to the small number of issues in which the court has divided. And there has been a great deal of attention too to individual judges on the court–particularly A. Raymond Randolph–who are said to be playing out-sized roles in the development of the law. This seems to me to miss the larger point. The court has, in the main, spoken with one voice. It has done so repeatedly on complicated and difficult issues that dramatically divided the district court. And it has emerged from these cases with a lot more clarity than existed at the outset of its engagement with these cases. Relatively few commentators will, I expect, agree with the court’s center of gravity on all major points; I certainly don’t, though I am generally sympathetic to the reorientation. But I am deeply impressed at how quickly the court has managed to craft order out of chaos and make the law look predictable and knowable in application to future cases. These two cases are small pieces of that project.

Share on Facebook0Share on Google+0Email this to someoneTweet about this on Twitter4Print this pageShare on Reddit0