dissenting:
Disregarding settled principles governing the release of government records under the Freedom of Information Act, 5 U.S.C. § 552 et seq., this court holds that the government may keep secret the names of hundreds of persons whom it has detained in connection with its investigation of the September 11, 2001 terrorist attacks without distinguishing between information that can, in FOIA’s words, “reasonably be expected to interfere” with the investigation and information that cannot. 5 U.S.C. § 552(b)(7)(A). While the government’s reasons for withholding some of the information may well be legitimate, the court’s uncritical deference to the government’s vague, poorly explained arguments for withholding broad categories of information about the detainees, as well as its willingness to fill in the factual and logical gaps in the government’s case, eviscerates both FOIA itself and the principles of openness in government that FOIA embodies.
I.
I begin with some preliminary observations about the principles that govern this case. First, no one can doubt that uniquely compelling governmental interests are at stake: the government’s need to respond to the September 11 attacks — unquestionably the worst ever acts of terrorism on American soil — and its ability to defend the nation against future acts of terrorism. But although this court overlooks it, there is another compelling interest at stake in this case: the public’s interest in knowing whether the government, in responding to the attacks, is violating the constitutional rights of the hundreds of persons whom it has detained in connec*353tion with its terrorism investigation — by, as the plaintiffs allege, detaining them mainly because of their religion or ethnicity, holding them in custody for extended periods without charge, or preventing them from seeking or communicating with legal counsel. The government claims that the detainees have access to counsel and freedom to contact whomever they wish, see Op. at 938, but the public has a fundamental interest in being able to examine the veracity of such claims. Just as the government has a compelling interest in ensuring citizens’ safety, so do citizens have a compelling interest in ensuring that their government does not, in discharging its duties, abuse one of its most awesome powers, the power to arrest and jail.
Second, while the governmental interests in this case may be uniquely compelling, the legal principles that govern its resolution are not at all unique. The court’s opinion emphasizes the national-security implications of the September 11 investigation, but as the government conceded at oral argument, this case is not just about September 11. The law that governs this case is the same law that applies whenever the government’s need for confidentiality in a law enforcement investigation runs up against the public’s right to know “what [its] government is up to.” United States Dep’t of Justice v. Reporters Comm. for Freedom of the Press, 489 U.S. 749, 773, 109 S.Ct. 1468, 1481-82, 103 L.Ed.2d 774 (1989) (internal quotation marks omitted). In all such situations, FOIA fully accommodates the government’s concerns about the harms that might arise from the release of information pertaining to its investigations. To be sure, the statute strongly favors openness, since Congress recognized that an informed citizenry is “vital to the functioning of a democratic society, needed to check against corruption and to hold the governors accountable to the governed.” NLRB v. Robbins Tire & Rubber Co., 437 U.S. 214, 242, 98 S.Ct. 2311, 2326-27, 57 L.Ed.2d 159 (1978). But Congress also recognized that “legitimate governmental and private interests could be harmed by release of certain types of information.” John Doe Agency v. John Doe Corp., 493 U.S. 146, 152, 110 S.Ct. 471, 475, 107 L.Ed.2d 462 (1989) (internal quotation marks omitted). It therefore “provided ... specific exemptions under which disclosure could be refused,” id., including the four exemptions relevant to this case: Exemption 7(A), for information that “could reasonably be expected to” interfere with ongoing law enforcement efforts, 5 U.S.C. § 552(b)(7)(A); Exemptions 7(C) and 7(F), for information that “could reasonably be expected to” unjustifiably compromise an individual’s privacy or physical safety, id. § 552(b)(7)(C), (b)(7)(F); and Exemption 3, for information that other statutes exempt from disclosure, id. § 552(b)(3). But ‘“these limited exemptions do not obscure the basic policy that disclosure, not secrecy, is the dominant objective of the Act.’ ” John Doe Agency, 493 U.S. at 152, 110 S.Ct. at 475 (quoting Dep’t of Air Force v. Rose, 425 U.S. 352, 361, 96 S.Ct. 1592, 1599, 48 L.Ed.2d 11 (1976)). Accordingly, courts must “narrowly construe[ ]” the exemptions, and “the burden is on the agency to sustain its action.” Id. (internal quotation marks and citations omitted). The government may in some situations withhold entire categories of records from disclosure, as it seeks to do here by withholding names and other information pertaining to all terrorism-investigation detainees. In order to sustain its burden, however, the government must demonstrate that “the range of circumstances included in the category ‘characteristically supports] an inference’ that the statutory requirements for exemption are satisfied.” Nation Magazine v. United States Customs Serv., 71 F.3d 885, 893 (D.C.Cir.1995) (citing United States v. *354Landano, 508 U.S. 165, 176-80, 113 S.Ct. 2014, 2021-24, 124 L.Ed.2d 84 (1993)).
The third principle relates to the level of deference we owe the government. Invoking the “heightened deference to the judgments of the political branches with respect to matters of national security,” Zadvydas v. Davis, 533 U.S. 678, 696, 121 S.Ct. 2491, 2502, 150 L.Ed.2d 653 (2001), the government refuses to identify the specific categories of information that would actually interfere with its investigation, but rather asks us simply to trust its judgment. This court obeys, declaring that “the judiciary is in an extremely poor position to second-guess the executive’s judgment in this area of national security.” Op. at 928. But requiring agencies to make the detailed showing FOIA requires is not second-guessing their judgment about matters within their expertise. And in any event, this court is also in an extremely poor position to second-guess the legislature’s judgment that the judiciary must play a meaningful role in reviewing FOIA exemption requests. Neither FOIA itself nor this circuit’s interpretation of the statute authorizes the court to invoke the phrase “national security” to reheve the government of its burden of justifying its refusal to release information under FOIA.
To begin with, I think it not at all obvious that we owe heightened deference to the government in this case. Citing the legislative history of the 1974 amendments to FOIA’s Exemption 1, 5 U.S.C. § 552(b)(1), the exemption for national-security matters, we have held that in evaluating Exemption 1 claims, “ ‘substantial weight’ is to be accorded to detailed agency affidavits setting forth the basis for exemption.” Weissman v. CIA, 565 F.2d 692, 697 n. 10 (D.C.Cir.1977); see also S. Rep. No. 93-1200, at 12 (1974) (‘“[T]he conferees recognize that the Executive departments responsible for national defense and foreign policy matters have unique insights into what adverse effects might occur as a result of public disclosure of a particular classified record. Accordingly, the conferees expect that the federal courts, in making de novó determinations in section 552 (b)(1) cases under the Freedom of Information law, will accord substantial weight to an agency’s affidavit concerning the details of the classified status of the disputed record.’ ”). We have also extended this heightened deference to cases involving Exemption 3 as it incorporates the National Security Act of 1947, which requires the CIA Director to protect “intelligence sources and methods” from unauthorized disclosure, 50 U.S.C § 403-3(c)(7). E.g., Halperin v. CIA, 629 F.2d 144 (D.C.Cir.1980) (National Security Act); Weissman, 565 F.2d 692 (Exemption 1 and National Security Act). The government, however, relies on neither Exemption 1 nor the National Security Act in this case, and contrary to the court’s suggestion, see Op. at 15, we have never held that such heightened deference is also appropriate in Exemption 7 cases. Indeed, in Weissman, which the court cites for the proposition that “we owe the same deference under Exemption 7(A) in appropriate cases” — we found Exemption 7 inapplicable in the case of the CIA’s investigation into the FOIA requester’s background “except under special collateral circumstances,” for instance, to protect the identities of FBI personnel named in requested materials. We instead focused on the deference owed the agency under Exemption 1, as well as Exemption 3 as it incorporates the National Security Act. 565 F.2d at 694-96, 698 & n. 15.
In any event, the government’s case fails even under the heightened deference we have applied in Exemption 1 and National Security Act cases. No matter the level of deference, our review is not “vacuous.” Pratt v. Webster, 673 F.2d 408, 421 (D.C.Cir.1982). Even when reviewing Ex*355emption l’s applicability to materials classified in the interest of national security, we have made clear that no amount of deference can make up for agency allegations that display, for example, a “lack of detail and specificity, bad faith, [or] failure to account for contrary record evidence,” since “deference is not equivalent to acquiescence.” Campbell v. U.S. Dep’t of Justice, 164 F.3d 20, 30 (D.C.Cir.1998). By accepting the government’s vague, poorly explained allegations, and by filling in the gaps in the government’s case with its own assumptions about facts absent from the record, this court has converted deference into acquiescence.
With these principles in mind, I examine each of the government’s arguments for withholding the detainee information. Part II explains why Exemption 7(A), which forms the basis of the court’s holding, cannot justify the government’s refusal to disclose the bulk of the requested information about the detainees. Part III shows why the government’s alternative arguments under Exemptions 7(C), 7(F), and 3 as it incorporates Federal Rule of Criminal Procedure 6(e) likewise fail. Finally, Part IV demonstrates why, on the basis of the record before us, the government has no basis under any exemption for withholding the names of the detainees’ attorneys.
II.
Although FOIA permits agencies to craft rules exempting certain categories of records from disclosure under Exemption 7(A) instead of making a record-by-record showing, see Robbins Tire, 437 U.S. at 236, 98 S.Ct. at 2323-24, an agency’s ability to rely on categorical rales has limits. Specifically, the government must divide information it seeks to withhold into “categories ... [that are] sufficiently distinct to allow a court to grasp ‘how each ... category of documents, if disclosed, would interfere with the investigation.’ ” Crooker v. Bureau of Alcohol, Tobacco & Firearms, 789 F.2d 64, 67 (D.C.Cir.1986) (quoting Campbell v. Dep’t of Health & Human Servs., 682 F.2d 256, 265 (D.C.Cir.1982)). An acceptable category is “functional,” that is, it “allows the court to trace a rational link between the nature of the document and the alleged likely interference.” Id.; see also Nation Magazine, 71 F.3d at 893 (“There are limits ... to when categorical rales may be employed. Only when the range of circumstances included in the category ‘characteristically supports] an inference’ that the statutory requirements for exemption are satisfied is such a rale appropriate.”).
Although I have no doubt that some of the requested information is exempt from FOIA’s mandatory disclosure requirement, the court treats disclosure as an all-or-nothing proposition, repeatedly emphasizing the breadth of the plaintiffs’ request— the fact that they seek the names and other information pertaining to “every single individual detained in the course of the government’s terrorism investigation,” Op. at 929 — as a justification for accepting the government’s own very broad, categorical refusal to release the bulk of the requested information. This all-or-nothing approach runs directly counter to well-established principles governing FOIA requests. Nothing in the statute requires requesters to seek only information not exempt from disclosure. To the contrary, the government bears the burden of reviewing the plaintiffs’ request, identifying functional categories of information that are exempt from disclosure, and disclosing any reasonably segregable, non-exempt portion of the requested materials. 5 U.S.C. § 552(b). The government fails to satisfy that burden in this case, for the range of circumstances included in the government’s exemption request do not “characteristically support” an inference that the information *356would interfere with its terrorism investigation.
In support of its exemption request, the government offers declarations from two senior officials with responsibility for the terrorism investigation. One of those declarations, by Dale L. Watson, a Federal Bureau of Investigation official charged with supervising the investigation, was prepared not for this case, but for cases involving the closure of deportation hearings. See N. Jersey Media Group, Inc. v. Ashcroft, 308 F.3d 198 (3d Cir.2002), cert. denied, — U.S.-, 123 S.Ct. 2215, 155 L.Ed.2d 1106 (2003); Detroit Free Press v. Ashcroft, 303 F.3d 681 (6th Cir.2002). Watson’s declaration thus speaks not to the harm that would flow from disclosing detainees’ names or other information, but instead to the harm that would flow from publicly airing evidence about particular detainees at such a hearing — i.e., “what evidence led to the detention of each individual,” “[ijnformation about how any given individual entered the country,” and “what evidence the United States has against members of a particular cell.” Watson Deck ¶ ¶ 12-13. Plaintiffs in this case request no such information. The court nevertheless relies on the Watson declaration, as well as North Jersey Media Group, see Op. at 931-32, despite the fact that neither has anything to do with the release of detainee names.
The other declaration, by Department of Justice Terrorism and Violent Crime Section chief James S. Reynolds, does in fact outline the harms that might result from release of some detainee names. But it does not support the government’s request for a 7(A) exemption, since that request treats all detainees the same, even though Reynolds tells us that the only common thread among the detainees is that they were “originally questioned because there were indications that they might have connections with, or possess information pertaining to, terrorist activity against the United States.” Reynolds Deck ¶ 10; see also id. ¶ ¶ 27, 36. As Reynolds himself acknowledges, this group includes some detainees who have turned out to be innocent of any involvement with terrorist activity and have “no information useful to the investigation.” Id. ¶ 36.
Ignoring this' important concession, the court declares that “[t]he clear import of the declarations is that many of the detainees have links to terrorism” — which the court considers “no surprise given that the detainees were apprehended during the course of a terrorism investigation, and given that several detainees have been charged with federal terrorism crimes or held as enemy combatants.” Op. at 931. The court’s approach is unconvincing for two reasons.
To begin with, it rests on what seems to be a faulty assumption about facts not in evidence. As of November 5, 2001, the last time the government released a tally, there were 1,182 detainees. See Dan Eggen & Susan Schmidt, Count of Released Detainees Is Hard to Pin Down, Wash. Post, Nov. 6, 2001, at A10 (quoting Justice Department spokeswoman Mindy Tucker). Nothing in the record tells us how many of those 1,182 detainees have been charged with federal terrorism crimes or held as enemy combatants. What little information the record does contain, however, suggests that the number may be relatively small. A list of federally charged detainees attached to the government’s motion for summary judgment reports that as of the time this suit was filed, only one detainee had been criminally charged in the September 11 attacks and only 108 detainees had been charged with any federal crime — primarily violations of antifraud statutes. Reynolds Deck ¶ 27; Def. Mot. for Summary Judgment, Ex. 8.
*357In any event, the court concedes the point — even if “many” of those “apprehended during the course of a terrorism investigation” have links to terrorism, not all of them do. As the court itself notes, the declarations establish that many of the INS detainees were held because law enforcement agents determined in the course of questioning them that they were in violation of federal laws; only “ ‘in some instances’ ” did agents “also determine[ ] that they had links to other facets of the investigation.” Op. at 931 (quoting Reynolds Decl. ¶ 10). Furthermore, although the court assumes that all those detained on material witness warrants “have relevant knowledge about the terrorism investigation” because a federal judge issues such warrants “based on an affidavit stating that the witness has information relevant to an ongoing criminal investigation,” Op. at 931, that assumption seems unwarranted given the government’s concession that “it may turn out that these individuals have no information useful to the investigation,” Reynolds Deck ¶ 36.
The government gives us no reason to think that releasing the names of these innocent detainees could interfere with its investigation. Indeed, the government never really asks us to believe that disclosure of the names of innocent persons having no knowledge of terrorist activity would in any way impede its ability to gather information from those who do have such knowledge. Instead, it asserts that “a detainee who knows his name will be made public may be deterred from cooperating now or in the future for fear of retaliation by terrorist organizations against him or his family and associates.” Reynolds Deck ¶ 15. Although the court accepts this argument, Op. at 930, it is ultimately not an argument for withholding detainees’ names, but rather for withholding the names of people who have information that might be helpful to law enforcement officials. These are two different categories of people, for as Reynolds acknowledges, many detainees have no information to provide. Reynolds Deck ¶ 36. These two groups thus merit different treatment. In fact, several statutory provisions address precisely the problem the government identifies, but all of them are aimed at protecting the identities of those people who provide information, not people the government questions because it thinks they might have information but who turn out not to. FOIA Exemption 7(A) protects the identities of witnesses where disclosure might pose a risk of interference in the form of witness intimidation or coercion, Robbins Tire, 437 U.S. at 239-40, 98 S.Ct. at 2325-26; FOIA Exemption 7(D) protects the identities of sources who choose to provide information to law enforcement agents on a confidential basis, 5 U.S.C. § 552(b)(7)(D); and the National Security Act protects the identity of intelligence sources in order to prevent those sources from “clos[ing] up like a clam,” CIA v. Sims, 471 U.S. 159, 172, 105 S.Ct. 1881, 1889, 85 L.Ed.2d 173 (1985) (internal quotation marks omitted). The government can and should rely on these provisions to protect the names of detainees who provide information to law enforcement agents or whom the government believes will be able to provide such information in the future. The government may not, however, preemptively withhold the identities of innocent detainees who do not now, and may never, have any information of use to the terrorism investigation.
The only argument that could conceivably support withholding innocent detainees’ names is the assertion that disclosure of the names “may reveal details about the focus and scope of the investigation and thereby allow terrorists to counteract it.” Reynolds Deck ¶ 16 (emphasis added). That Reynolds believes these harms may result from disclosure is hardly surprising — anything is possible. But before ac*358cepting the government’s argument, this court must insist on knowing whether these harms “could reasonably be expected to” result from disclosure — the standard Congress prescribed for exemption under 7(A). Nothing in Reynolds’s declaration suggests that these harms are in fact reasonably likely to occur.
To begin with, Reynolds never explains how a list of names of persons unknown to terrorist organizations would tell the terrorists anything at all about the investigation, much less allow them to “map [its] progress.” Id. For example, if the government tells us that it detained men named Mohammed Mubeen, Osama Elfar, Ghassan Dahduli, Fathi Mustafa, Nacer Fathi Mustafa, and Hady Omar, Jr., none of whom has any connection to terrorist organizations, see Amy Goldstein, A Deliberate Strategy of Disruption: Massive, Secretive Detention Effort Aimed Mainly at Preventing More Terror, Wash. Post, Nov. 4, 2001, at Al, what could that information possibly tell terrorists about the government’s investigation? Though Reynolds’s declaration provides no answer, the court speculates that the names of these innocent detainees could be valuable to terrorist organizations because “[s]uch detainees could be acquaintances of the September 11 terrorists, or members of the same community groups or mosques.” Op. at 930. That may well be true in some cases, but if it is, Reynolds should tell us so under oath, thus providing a record basis for the government to claim an exemption for those detainees who pose such concerns. But the court’s speculation, supported only by a newspaper article describing a single detainee who attended a mosque that two terrorism suspects also attended, see id. (citing Rachel L. Swarns, Muslims Protest Monthlong Detention Without a Charge, N.Y. Times, April 20, 2003, at A16), falls far short of satisfying the government’s burden under FOIA.
The government’s failure to provide an adequate explanation is all the more glaring given that the detainees represent only a subset — and quite possibly a very small subset — of persons questioned in connection with this investigation. Reynolds Supp. Decl. ¶2. As a result, even if releasing detainee names were to provide some insight into the terrorism investigation, that insight would be limited. Releasing the names of the detainees, but not the names of those questioned in connection with the investigation, can paint only a partial — and possibly misleading — picture of the government’s investigative strategy. For example, if the government detains two people in Detroit but questions a thousand in Chicago, wouldn’t release of the detainee information wrongly lead terrorist organizations to believe that the government was focusing on Detroit, not Chicago?
The second failing in both the government’s request and the court’s analysis is that they treat all detainee information the same, despite the fact that each item of information that plaintiffs seek about the detainees — names, attorneys’ names, dates and locations of arrest, places of detention, and dates of release — is clearly of very different value to terrorists attempting to discern the scope and direction of the government’s investigation. Although the Reynolds declaration tells us that “releasing the names of the detainees who may be associated with terrorism and their place and date of arrest would reveal the direction and progress of the investigations,” Reynolds Decl. ¶ 16, it does not tell us, for example, whether releasing the detainees’ names and dates of arrest, but not their places of arrest — or even releasing the dates of arrest alone — would involve the same danger. The Reynolds declaration, moreover, contains no justification at all for withholding dates of release. Indeed, the government has already disclosed the *359release dates of detainees who had been held on federal criminal charges. Id. ¶ 8. This information may seem unimportant, but from the FOIA requesters’ point of view, it could be highly relevant to the question of how the government is treating the persons it has detained. Taken together, arrest and release dates can tell the public how long persons have been detained, raising concerns about possible constitutional violations. See Appellees’ Br. at 27.
The government’s allegations of harm are also undercut by the fact that it has itself provided several other means by which this information can become public. Not only do detainees remain free to inform whomever they choose of their detention, Reynolds Decl. ¶ 28, but on numerous occasions since September 11, the government itself has disclosed precisely the kind of information it now refuses to provide under FOIA. For example, on April 17, 2002, the government announced the arrest of Issaya Nombo, whom government officials said they suspected of connections to terrorism, although he was arrested on immigration charges. Officials revealed Nombo’s name and the date and place of his arrest. Philip Shenon, African Held After Name Is Left in Cave, N.Y. Times, Apr. 18, 2002, at A15. At a June 10, 2002 press conference, the Attorney General announced the arrest of one Abdulla Al Muhajir, born José Padilla, for suspected terrorism involvement, revealing not only A1 Muhajir’s two names, but also the date and place of his arrest, and the events leading to his capture. Attorney General Ashcroft News Conference, June 10, 2002, available at http:/www.usdoj.gov/ag/speeches/ 2002/061002agtranscripts.htm. And on July 26, 2002, government officials announced they were holding Mohammad Mansur Jabarah on a material witness warrant after his arrest in connection with a terrorist plot in Singapore. William K. Rashbaum, Captured Qaeda Member Gives Details on Group’s Operations, N.Y. Times, July 27, 2002, at A8.
Nothing in the record explains why the government’s concerns about interference with the investigation do not apply with respect to detainees such as Abdulla A1 Muhajir, Issaya Nombo, and Mohammad Mansur Jabarah, but do nevertheless apply with respect to the other detainees. In its reply brief, the government explains that it may have strategic reasons for disclosing certain information, since its disclosures to date “have identified specific individuals in a manner unlikely, in the view of the law enforcement experts, to impede the progress of the investigation.” Appellant’s Reply Br. at 18. While this may well be so, it is an argument of counsel, and though the court accepts it, FOIA requires that the agency — not counsel— explain such judgments under oath. The reason for this requirement is clear: We owe deference to agency expertise, not to lawyers defending the agency in litigation. See, e.g., Church of Scientology v. IRS, 792 F.2d 153, 165-66 (D.C.Cir.1986) (citing SEC v. Chenery Corp., 318 U.S. 80, 63 S.Ct. 454, 87 L.Ed. 626 (1943)). If there are legitimate investigative reasons for releasing the names of some detainees, but not others, then Mr. Reynolds or others responsible for the terrorism investigation should explain those reasons under oath— in an in camera affidavit, if necessary to protect the information — and that explanation would probably warrant judicial deference.
It is true, as the court points out, that the Supreme Court in CIA v. Sims acknowledged “the political realities of intelligence operations in which, among other things, our Government may choose to release information deliberately to ‘send a message’ to allies or adversaries” when it upheld the CIA’s right to withhold intelli*360gence information even if the CIA has already released some part of it. 471 U.S. at 180, 105 S.Ct. at 1893-94. Unlike this court, however, the Supreme Court did not simply assume it understood the government’s strategy; it reached its conclusion on the basis of the CIA Director’s affidavit explaining that strategy. Id. at 180 & n. 24, 105 S.Ct. at 1893. The record in this case contains no similar explanation. Moreover, counsel’s argument suggests that the government itself differentiates among detainees on a case-by-case basis for purposes of assessing how disclosure might harm its investigation. If the government itself makes such distinctions in deciding what information to release, then why, particularly in light of FOIA’s exacting standards, doesn’t it make those distinctions in its exemption request before this court?
By asking these questions, the court would not, as it warns, be “second-guessing” the government’s judgments about matters of national security. Op. at 928. It would, rather, be doing the job Congress assigned the judiciary by insisting that the government do the job Congress assigned to it: provide a rational explanation of its reasons for claiming exemption from FOIA’s disclosure requirements.
III.
Because the court concludes that Exemption 7(A) applies to the government’s entire request, it never addresses the government’s alternative arguments under Exemptions 7(C), 7(F), and 3. In my view, none of these provisions supports the government’s refusal to disclose the detainee information either.
Exemption 7(C)
Exemption 7(C) permits the government to withhold law enforcement records where their release “could reasonably be expected to constitute an unwarranted invasion of personal privacy.” 5 U.S.C. § 552(b)(7)(C). Like Exemption 7(A), the application of Exemption 7(C) is subject to a set of wellestablished standards. Because the statute refers not to invasions of privacy generally, but to “unwarranted” invasions of privacy, courts evaluating claims for 7(C) exemption must do more than simply identify a privacy interest that will be compromised by disclosure of information. Instead, they must “balance the public interest in disclosure against the interest Congress intended the Exemption to protect.” Reporters Committee, 489 U.S. at 776,109 S.Ct. at 1483.
Relying on our decision in Nation Magazine, the government argues that the detainees have “ ‘an obvious privacy interest cognizable under Exemption 7(C) in keeping secret the fact that they were subjects of a law enforcement investigation,’ ” and that these privacy concerns are “particularly acute given the nature and magnitude of the September 11 attacks.” Appellant’s Br. at 39-40 (quoting Nation Magazine, 71 F.3d at 894). This argument is unconvincing. For one thing, if the government is so concerned with the detainees’ privacy, why has it released so much information about them? What about Abdulla Al Muhajir’s privacy, or Issaya Nombo’s, or Mohammad Mansur Jabarah’s? Nothing in the Reynolds declaration explains how the government’s press conferences releasing the names of these detainees demonstrate any respect for their privacy.
In any event, we have never held that individuals who have been not only investigated, but also arrested and jailed, have a similar privacy interest in avoiding “unwarranted association with criminal activity or reputational harm.” Nation Magazine, 71 F.3d at 894. Even though being arrested subjects a person suspected of criminal activity to embarrassment and potentially more serious reputational harm, the law is nevertheless clear that no right *361of privacy “is violated by the disclosure of ‘an official act such as an arrest.’” Am. Fed’n of Gov’t Employees, AFL-CIO v. Dep’t of Housing & Urban Dev., 118 F.3d 786, 794 (D.C.Cir.1997) (quoting Paul v. Davis, 424 U.S. 693, 713, 96 S.Ct. 1155, 1166, 47 L.Ed.2d 405 (1976)).
To be sure, detainees may have a unique interest in avoiding association with the crimes of September 11. Even so, that interest is clearly outweighed by the public interest in knowing whether the government, in investigating those heinous crimes, is violating the rights of persons it has detained. And while FOIA asks only whether the public interest in disclosure outweighs the private interest in secrecy, it bears noting that the private interests in this case weigh on both sides of the balance: Plaintiffs’ request for disclosure of the detainees’ names seeks to vindicate not only the public’s right to know what its government is up to, but also the detainees’ own rights, including the right to counsel and to speedy trial.
Nothing in SafeCard Services, Inc. v. SEC, 926 F.2d 1197 (D.C.Cir.1991), requires a different result. SafeCard establishes that names appearing in law enforcement files will often fall within the scope of Exemption 7(C), since records containing such information are generally far less probative of an agency’s behavior or performance than of the behavior of the persons whose names appear in the records. Id. at 1205; see also Nation Magazine, 71 F.3d at 895 (“In some, perhaps many, instances where a third party asks if an agency has information regarding a named individual in law enforcement files, the cognizable public interest in that information will be negligible; the requester will be seeking records about a private citizen, not agency conduct.”). The Safe-Card court therefore formulated a categorical rule exempting disclosure of such information unless the requester can show (1) compelling evidence that the agency is engaged in illegal activity, and (2) that the information is necessary to confirm or refute that evidence. SafeCard, 926 F.2d at 1205-06. Plaintiffs’ FOIA request satisfies both elements of this rule.
To begin with, this case does not implicate SafeCard’s concern that disclosure of names in law enforcement files will generally shed less light on the government’s behavior than it does on the behavior of private citizens. In SafeCard, the FOIA requester sought information relating to organizations and individuals whom the SEC had suspected of manipulating the requester’s stock and who might be witnesses or litigants in the SEC’s investigations. 926 F.2d at 1200, 1205. Similarly, in many other Exemption 7(C) eases, FOIA requesters seek names in law enforcement files primarily in order to attack their convictions or otherwise exculpate themselves — a “personal stake” in disclosure that “does not count in the calculation of the public interest.” Oguaju v. United States, 288 F.3d 448, 450 (D.C.Cir.2002); see also Billington v. United States Dep’t of Justice, 233 F.3d 581, 582 (D.C.Cir.2000). Here, in contrast, plaintiffs have little if any personal stake in their FOIA request, which aims solely to glean information relating to the government’s conduct of its terrorism investigation and its treatment of the detainees. Designed to “shed[] light on an agency’s performance of its statutory duties,” this request implicates precisely the kind of public interest lying at the heart of Exemption 7(C)’s balancing test. Reporters Committee, 489 U.S. at 773,109 S.Ct. at 1481-82.
Moreover, plaintiffs offer ample evidence of agency wrongdoing. The record includes hundreds of pages of newspaper articles, human rights reports, and congressional testimony reporting alleged governmental abuses such as holding de*362tainees for long periods without allowing them to seek or communicate with counsel and without charging them. See, e.g., Alison Leigh Cowen, Detainees’ Lawyers Complain of Unfair Treatment, N.Y. Times, Oct. 21, 2001, at Bl; Richard A. Serrano, Many Held in Terror Probe Report Rights Being Abused, L.A. Times, Oct. 15, 2001, at Al; Amnesty International, Amnesty International’s Concerns Regarding Post September 11 Detentions in the USA, available at http://web.amnesty.org/aidoc/aidoc_pdf.nsf; Human Rights Watch, Presumption of Guilt: Human Rights Abuses of Post-September 11 Detainees, available at http://www.hrw. org/reports/2002/us911/USA0802.pdf; Department of Justice Oversight: Preserving Our Freedoms While Defending Against Terrorism: Hearing Before the Senate Judiciary Comm., 107th Cong. (2001) (statement of Gerald H. Goldstein, Attorney, National Ass’n of Criminal Defense Lawyers), available at http://judiciary.senate.gov/hearing.cfm?=28; id. (statement of Michael Boyle, Attorney, American Immigration Lawyers Ass’n). To be sure, none of this evidence has been tested and proved in a court of law. But SafeCard requires only “compelling” evidence — not tested evidence, and not even evidence that would be admissible at trial. If hundreds of pages of first-hand reports of governmental abuses do not qualify as “compelling” evidence sufficient to justify an investigation into the government’s conduct, then I cannot imagine what would. After all, FOIA’s purpose, as SafeCard recognizes, is to allow the public access to records necessary to ascertain whether the government has acted illegally. If requesters already had tried and tested proof of such illegal activity, then resort to FOIA would be unnecessary. History, moreover, is full of examples of situations in which just these sorts of allegations led to the discovery of serious government wrongdoing — from Teapot Dome in the 1920s to the FBI’s COINTELPRO counterintelligence program in the 1960s to Watergate in the 1970s.
In short, by interpreting SafeCard to require anything more than compelling “allegations of illegal agency activity,” Nation Magazine, 71 F.3d at 896, the government would transform the SafeCard test into a categorical ban on the disclosure of names contained in law enforcement records. That result finds justification in neither FOIA nor our cases interpreting Exemption 7(C). See Nation Magazine, 71 F.3d at 896 (holding that a blanket exemption for all names in law enforcement records “would be contrary to FOIA’s overall purpose of disclosure, and thus is not a permissible reading of Exemption 7(C)”); Stern v. FBI, 737 F.2d 84 (D.C.Cir.1984) (ordering the disclosure of the name of a high-ranking FBI official in internal reports concerning the agency’s investigation of a cover-up).
Finally, plaintiffs need the information they request to confirm or refute the compelling evidence of agency wrongdoing— the SafeCard test’s second requirement. While it is true that a list of names alone would shed no light on whether the government has respected detainees’ constitutional rights, plaintiffs need the names in order to gather information about the government’s treatment of the detainees. Appellees’ Br. at 30. In this respect, plaintiffs’ request differs from the vast majority of FOIA requests for information concerning named individuals in law enforcement files, where the only plausible public interest is knowing to what extent an agency believed the named individuals were involved in illegal activity. Cf. Rosenfeld v. United States Dep’t of Justice, 57 F.3d 803, 812 (9th Cir.1995) (holding that Exemption 7(C) does not justify withholding the identities of persons investigated for *363subversive activities in FBI files, where the names would make it possible to determine whether the FBI had investigated student activists for participating in political protests by comparing the FBI’s investigations to a roster of a student activist group’s leadership).
Amici Washington Legal Foundation and the Jewish Institute for National Security Affairs contend that release of the information is not necessary to evaluate whether the government is operating within the bounds of the law in detaining persons in connection with its terrorism investigation, since the public has other means of obtaining the information: Individual detainees can bring individual lawsuits, the Department’s Inspector General has investigated allegations of misconduct, and media reports and congressional investigations all tell the public what its “government is up to.” Washington Legal Found. Br. in Support of Appellant at 17. But Amici’ s argument has no basis in FOIA. If Congress had intended for individual lawsuits, internal investigations, or newspaper reports to relieve the government of its obligations under FOIA, then it would have expressed that intent in the law.
Exemption 7(F)
The government next invokes Exemption 7(F), which permits withholding law enforcement records where their release “could reasonably be expected to endanger the life or physical safety of any individual.” 5 U.S.C. § 552(b)(7)(F). Here again, the government’s evidence fails to establish that the entire range of records encompassed in the plaintiffs’ FOIA request “could reasonably be expected” to endanger the detainees.
The government’s declarations tell us only that (1) “[d]etainees who are, in fact[,] affiliated with a terrorist group may be perceived by such groups as informants for the United States and be killed to preclude their future cooperation,” Reynolds Decl. ¶ 37, and (2) “[i]f prisoners learn that an individual who was detained as a result of the investigation emanating from the September 11 attacks is in their own prison facility, some may try to retaliate against this individual,” id. ¶ 29. The government tells us nothing about what threat, if any, disclosure would pose to detainees who are neither affiliated with a terrorist group nor currently imprisoned. And the government’s own disclosures again undermine its assertions about detainees’ safety. Plaintiffs point out that the Justice Department Inspector General himself named two of the detention centers used to house the terrorism investigation detainees, a fact that the government neither denies nor explains. Appellees’ Br. at 27. Again, the government may have had reasons for disclosing the names of only these two detention centers, but nothing in the Reynolds declaration tells us what those reasons might be.
Exemption 3
Finally, the government invokes Exemption 3, which exempts from disclosure matters that are “specifically exempted from disclosure by statute ..., provided that such statute ... requires that the matters be withheld from the public in such a manner as to leave no discretion on the issue.” 5 U.S.C. § 552(b)(3). According to the government, Exemption 3, which encompasses Federal Rule of Criminal Procedure 6(e)’s prohibition on the disclosure of “matters occurring before the grand jury,” see Fund for Constitutional Gov’t v. Nat’l Archives & Records Serv., 656 F.2d 856, 867-68 (D.C.Cir.1981), excuses it from disclosing the names of detainees held on material witness warrants, since “each of these warrants was issued to procure a witness’s testimony before a grand jury,” Reynolds Second Supp. Deck *364¶4. As such, the government contends that Exemption 3 provides a ground for nondisclosure independent of Exemption 7.
Rule 6(e) forbids disclosure of “not only what has occurred and what is occurring, but also what is likely to occur” before a grand jury, including disclosure of witnesses’ identities. In re Motions of Dow Jones & Co., 142 F.3d 496, 500 (D.C.Cir.1998). Therefore, the names of persons detained on material witness warrants who have actually testified before grand juries are unquestionably exempt from disclosure. The government, however, insists that Exemption 3 also covers the names of material witness detainees who have neither testified before grand juries nor are scheduled to do so, as well as the names of detainees who were released without ever having testified, because all of these detainees were originally detained in order to “procure [their] testimony before a grand jury.” Reynolds Second Supp. Deck ¶ 4.
Saying that the material witness detainees were held in order to secure their testimony is quite different from saying that their testimony is “likely to occur” before a grand jury. Indeed, the record indicates that at least seven material witnesses have been released without testifying before a grand jury, so in their cases, it seems more accurate to say that their testimony is quite unlikely to occur before a grand jury. See Indiana Men Ordered to Testify to Return to Evansville, Assoc. Press, Oct. 25, 2001. Furthermore, although current detainees may be, on the whole, somewhat more likely to testify before grand juries, their testimony is not necessarily “likely to occur” for purposes of Rule 6(e). We have said that the “likely to occur” language must be read sensibly: It does not authorize the government to draw “a veil of secrecy ... over all matters occurring in the world that happen to be investigated by a grand jury.” In re Sealed Case, 192 F.3d 995, 1001-02 (D.C.Cir.1999) (internal quotation marks and citation omitted). Accordingly, we have made clear that Rule 6(e) covers “testimony about to be presented to a grand jury” (emphasis added) — hence the “likely to occur” language — but does not cover government investigations that merely parallel grand jury investigations. Id. at 1002-03. Because the government fails to show that all material witness detainees are likely to testify before grand juries, it may not, on this record, withhold their names under Rule 6(e). To hold otherwise would convert this circuit’s carefully crafted standard into an absolute rule that would permit the government to keep secret the name of any witness whom it ever thought might testify at a grand jury proceeding, or who might testify at some indefinite point in the future. Neither Rule 6(e) nor the law of this circuit justifies that result.
IV.
No part of the government’s exemption request better illustrates its infirmities than its refusal to disclose the names of the detainees’ attorneys. Essentially rehashing its arguments for withholding the names of the attorneys’ clients, the government argues — and the court agrees, see Op. at 932 — that releasing attorneys’ names would interfere with the terrorism investigation and would compromise the detainees’ privacy interests, since releasing a list of attorneys “may facilitate the identification of the detainees themselves.” Reynolds Deck ¶ 18. The government also claims to be withholding the attorneys’ names for their own good, warning that attorneys identified as representing individuals detained in connection with the terrorism investigation “run the risk that they will be subjected to harassment or retaliation in their personal as well as professional lives.” Reynolds Deck ¶ 26.
*365Both parts of this argument not only are profoundly wrong, but also reflect a complete misunderstanding of the nature of this country’s legal profession. In the first place, attorneys’ names are quite clearly not a proxy for the names of their clients. Indeed, recognizing that knowledge of the lawyer’s identity does not inexorably lead to identifying the client, ethical rules forbid lawyers from identifying their clients without their consent, except in extraordinary circumstances. Rule 1.6 of the Model Rules of Professional Conduct provides that absent extraordinary circumstances, “[a] lawyer shall not reveal information relating to the representation of a client unless the client gives consent after consultation” — a prohibition that generally includes disclosure of a client’s identity. See Ctr. for Prof’l Responsibility, Am. Bar Ass’n, Annotated Model Rules of Professional Conduct 88 (4th ed.1999). Because the decision ultimately belongs to the detainees and not their lawyers, providing a list of the lawyers’ names would do little more to facilitate identification of the detainees than the government’s current policy of allowing the detainees to identify themselves to the media and to whomever else they choose.
Even assuming that releasing attorneys’ names will somehow facilitate identification of the detainees, the court’s all-or-nothing approach again impermissibly shifts the burden of identifying exempt information from the government to plaintiffs. The government’s Exemption 7(A) argument for withholding lawyers’ names thus fails for the same reason as its 7(A) argument for withholding the names of all detainees. How would releasing the names of attorneys representing innocent clients with no connection to terrorist activities interfere with the government’s terrorism investigation? Neither the court nor the government provides an explanation.
The government’s second argument fares no better. The notion that the government must withhold the attorneys’ names for their own good is flatly inconsistent with lawyers’ roles as advocates and officers of the court in our fundamentally open legal system. Having voluntarily assumed this public role, lawyers have little expectation of anonymity. I, for one, know of no lawyer who has ever asked for the kind of protection the government now asserts on behalf of the detainees’ lawyers. Moreover, I have no doubt that lawyers will represent individuals associated with the terrorism investigation even without the protection the government urges. As Judge Kessler noted in her opinion in this case, the history of our profession is full of examples of brave men and women who have taken on unpopular causes in the interest of justice. Ctr. for Nat’l Sec. Studies v. Dep’t of Justice, 215 F.Supp.2d 94, 109 (D.D.C.2002). Not only do lawyers regularly represent persons accused of terrible and highly publicized crimes against individuals, but many of this country’s most prominent and well-respected lawyers have defended persons accused of heinous crimes against the state, from Aaron Burr to Nazi saboteurs to Soviet spies, as well as persons merely suspected of a propensity to commit such crimes, such as Japanese internees in World War II.
In addition, the government completely fails to substantiate its concerns about releasing attorneys’ names. The government insists that “[i]n light of the brutality of the acts committed against the United States, even the mere possibility of retaliation against these lawyers justifies withholding their identities.” Reynolds Decl. ¶ 38. FOIA, of course, does not allow the government to withhold information based on “mere possibilities.” And the Reynolds declaration fails to establish that retaliation is reasonably likely. Indeed, if the *366government is so worried about retaliation against lawyers, why did it release a comprehensive list of attorneys representing federally charged detainees? See Def. Mot. for Summary Judgment, Ex. 8. Reynolds provides no answer. And if the risk of retaliation has materialized in the case of these attorneys, Reynolds certainly does not tell us so.
If the government has legitimate concerns about the safety of one or more of the lawyers, FOIA requires it to describe those concerns with reasonable specificity — again in an in camera affidavit, if necessary — and explain why it believes the harms it fears “could reasonably be expected” to occur. Giving appropriate deference to law enforcement expertise, the district court would then be in a position to evaluate the government’s concerns and determine whether withholding the attorneys’ names under Exemption 7(F) — or some other, less extraordinary measures— are needed to protect the detainees’ attorneys. Absent such evidence, however, the government has no basis for its flat refusal to release lawyers’ names.
V.
Although I think it unreasonable to infer that all of the information plaintiffs seek in their FOIA request qualifies for exemption, the government may be able to point to more narrowly defined categories of information that might justify the inference. For example, while nothing in the record supports the government’s contention that releasing the names of innocent detainees would harm the investigation, perhaps the government could justify withholding the places of arrest on the ground that such information might provide terrorist organizations with some insight into the government’s investigative methods and strategy. I would therefore remand to allow the government to describe, for each detainee or reasonably defined category of detainees, on what basis it may withhold their names and other information.
This “more particularized approach” comports with both “Congress’ intent to provide workable rules of FOIA disclosure,” United States v. Landano, 508 U.S. 165, 180, 113 S.Ct. 2014, 2023-24, 124 L.Ed.2d 84 (1993) (internal quotation marks and citations omitted), and FOIA’s ultimate aim: to give the public “access to information about how Government is exercising its trust,” at a time when “Government is becoming involved in more and more aspects of every citizen’s personal and business life.” 112 Cong. Rec. 13654 (1966) (statement of Rep. Rumsfeld). It would also ensure that this court treat FOIA as “a disclosure statute and not as an excuse to withhold information from the public.” Id.
Rather than hold the government to clearly established standards governing FOIA exemptions, the court sustains the government’s vague, ill-explained decision to withhold information, invoking principles of deference and engaging in its own speculation to fill in the gaps in the government’s showing. In my view, the court’s approach drastically diminishes, if not eliminates, the judiciary’s role in FOIA cases that implicate national-security interests. Congress certainly could have written FOIA that way, but chose instead to require meaningful judicial review of all government exemption claims. If the government thinks that a new, broader FOIA exemption is need.ed for terrorism cases, it should ask Congress to create one, just as in the wake of September 11 it asked Congress to authorize roving wiretaps of suspected terrorists and to permit detention of non-U.S. citizens suspected of terrorism without specific charges. See USA PATRIOT Act, Pub. L. No. 107-56, 115 Stat. 272 (2001). But this court may not change the law in Congress’s stead. For *367all its concern about the separation-of-powers principles at issue in this case, the court violates those principles by essentially abdicating its responsibility to apply the law as Congress wrote it. I dissent.