dissenting from the denial of rehearing en banc:
I appreciate the panel’s efforts to clarify the Government’s production burden in these CSRT reviews. The panel assumes the phrase “reasonably available” adequately defines the scope of the record because that phrase comes from the CSRT regulations. However, because the record so defined does not arise naturally from the proceedings, the panel may have left much to litigate. The Government is clearly uncertain about what information is “reasonably available,” and is searching laboriously through “all relevant federal agencies” to make sure it gathers at least that much information. Pet. at 10. The panel has, naturally, refused to opine on whether the results of such an exhaustive *1307search are reasonably available, Bismullah v. Gates, 503 F.3d 137, 141 n. 3 (D.C.Cir.2007) (denial of panel rehearing) (Bismullah II), but it seems to think that too intensive a search would be unreasonable, see id. at 142. The panel avers that it did not require “[a] search for information without regard to whether it is ‘reasonably available.’ ” Id. at 141. But reliance on this sort of verbal formulation may confuse rather than clarify the obligation. Using the phrase “reasonably available” provides not a process — based definition, but an abstract legal standard. If the Government must populate the record based on this standard, it will have to conduct a new search for materials that satisfy it. Under the panel’s order, the record may be congruent with the universe of information identified by the regulations, but it bears no direct relationship to the CSRT process-or any process at all. Although the panel might have been right to reject the Government’s offer of only the record that a CSRT considered, that version of the record is at least the definite product of a process that actually happened.1 The likely result of relying on a theoretical record will be continued litigation over the inclusion or exclusion of various pieces of information, so that any review of the merits of these cases will be substantially delayed. This would be fair to neither the Government nor the detainees.
The denial of rehearing has generated five separate opinions disputing the proper scope of production; this continuing debate suggests the court has not yet found the right paradigm. Although we strain for familiar analogies to guide us, none of them is apt, because they all miss a central point: CSRTs are not adversarial proceedings. Detainees are not represented by advocates, but only by Personal Representatives whose sole duty is to assist, not defend, them. Conversely, the Recorders and the CSRTs have an obligation, under the procedures, to find and examine exculpatory evidence. That being so, it seems improbable that the Government need turn over only the Record of Proceedings compiled after the CSRT, as it originally urged, Bismullah v. Gates, 501 F.3d 178, 185 (D.C.Cir.2007) (Bismullah I). On the other hand, to demand everything means engaging this court in de novo review of the CSRTs, as the panel acknowledges. See Bismullah II, 503 F.3d at 139-40. Is such review what Congress intended when it passed the Detainee Treatment Act?
Congress mandated this court to review the CSRTs. An adversarial appeal from a nonadversarial hearing is an unfamiliar process in this country, but it is common in other parts of the world. Indeed, since the military’s prisoner-of-war procedures were developed to implement international law, Army Reg. 190-8 §§ 1 — 1(b)(3), l-6(a) (citing Geneva Convention Relative to the Treatment of Prisoners of War art. 5, Aug. 12, 1949, 6 U.S.T. 3316), it is conceivable that they were intentionally modeled on traditional inquisitorial procedures. Many aspects seem similar, including the role of the Recorder as both judge and investigator. Not only does he prepare the “official record of the Tribunal’s decision,” Memo, from the Sec’y of the Navy on Implementation of Combatant Status Review Tribunal Procedures Enel. 2 § C(10) (July 29, 2004); he also gathers the Government Information, which includes all “reasonably *1308available information ... bearing on ... whether the detainee” is an enemy combatant, id. Enel. 1 § E(3), including evidence both for and against that determination. Cf. Jacqueline Hodgson, French Criminal Justioe 30 (2005) (investigating magistrate must “gather[ ] evidence which might exculpate as well as incriminate the suspect”). Most important for this case, a civil-law inquisition prepares a well-defined record for review, consisting of the material that the magistrate actually gathered. Bron McKillop, Anatomy of a French Murder Case, 45 AM. J. Comp. L. 527, 544-46 (1997). Naturally, this record contains significantly less information than what the magistrate could have gathered because it was available.
My point is not to hold out continental criminal procedure as the perfect model for CSRT review, although it may be the closest (and may actually have been the original) model for the military’s prisoner-of-war tribunals. Nor, of course, is it a source of law, although it can be a useful source of ideas given that the military’s prisoner-of-war regulations expressly advert to international law. Nevertheless, this court could define the record in other ways than the “all” required by the panel or the “nothing” offered by the Government, and this definition is one of a set of decisions this court should make about how we are to conduct this novel form of review. I am now convinced we should have begun by discussing the problems much more thoroughly en banc. Accordingly, I dissent from the denial of rehearing.
. As a corollary, reconvening a CSRT, as the panel proposes, Bismullah II, 503 F.3d at 141, will only postpone the issue, because the abstract set of Government Information will have no relation to that proceeding either. The court will still review whether the Recorder for the new panel gathered all reasonably available information. Bismullah I, 501 F.3d at 185; Stmt, of Ginsburg, C.J., at 1295-96.