dissenting:
The Freedom of Information Act (“FOIA”) establishes a strong default rule: Government must disclose information in its possession unless it invokes, at its discretion, one of nine narrow exemptions. See, e.g., U.S. Dep’t of Def. v. FLRA, 510 U.S. 487, 494, 114 S.Ct. 1006, 127 L.Ed.2d 325 (1994) (citing U.S. Dep’t of Air Force v. Rose, 425 U.S. 352, 360-61, 96 S.Ct. 1592, 48 L.Ed.2d 11 (1976)); see also 5 U.S.C. § 552(b)(l)-(9) (enumerating exemptions). Here, HUD has invoked only Exemption 6, which authorizes withholding “personnel and medical files and similar files the disclosure of which would constitute a clearly unwarranted invasion of personal privacy.” 5 U.S.C. § 552(b)(6). Applying that exemption requires balancing the personal privacy interests it protects “against the public interest in government openness that would be served by disclosure.” Lahr v. Nat’l Transp. Safety Bd., 569 F.3d 964, 973 (9th Cir.2009).
The majority identifies a personal privacy interest which may not exist, and then weighs the public interest in disclosure too lightly. And it does both in the name of reducing the government’s burden in complying 'with the law, a concern with no application here. I dissent.
I.
Were it up to me, I would do what we did in our-original resolution of this case— remand to the district court for further factual development. See Prudential Locations LLC v. U.S. Dep’t of Housing & Urban Dev., 648 F.3d 768, 778-79 (9th Cir.2011), vacated and rehearing granted, 665 F.3d 1379 (9th Cir.2012). The present record is inadequate to confirm the existence of any personal privacy interest, let alone to balance it against the public interest in naming those informants whose tips our government deems credible enough to merit investigation.
A.
1. Exemption 6 applies only where some “personal privacy” interest is at *436stake. See Rose, 425 U.S. at 371, 375 n. 14, 96 S.Ct. 1592. A “‘nontrivial privacy interest’ ” can be enough to trigger further analysis under the exemption. See Forest Serv. Emps. for Envtl. Ethics v. U.S. Forest Serv., 524 F.3d 1021, 1026-27 (9th Cir.2008) (quoting FLRA, 510 U.S. at 501, 114 S.Ct. 1006). But if no nontrivial privacy interest is in jeopardy, then FOIA requires disclosure. See Multi Ag Media LLC v. Dep’t of Agric., 515 F.3d 1224, 1229-30 (D.C.Cir.2008).
A nontrivial privacy interest does not exist merely because the government asserts it. Rather, FOIA’s “strong presumption in favor of disclosure places the burden on the agency to justify the withholding of any requested documents.” U.S. Dep’t of State v. Ray, 502 U.S. 164, 173, 112 S.Ct. 541, 116 L.Ed.2d 526 (1991); Maricopa Audubon Soc’y v. U.S. Forest Serv., 108 F.3d 1089, 1092 (9th Cir.1997); 5 U.S.C. § 552(a)(4)(B). Like all parties on summary judgment, the government must adduce specific evidence to carry its burden. See Fed.R.Civ.P. 56(c).
The government in this case has submitted no evidence whatever concerning the relevant circumstances or privacy concerns of the individuals who filed the complaints of wrongdoing with HUD. In the face of this vacuum, the majority in effect holds that the government carries its burden of demonstrating a nontrivial privacy interest simply by specifying that the individual whose name is being withheld reported alleged wrongdoing to a government official. FOIA Exemption 6 requires more.
Exemption 6 provides protection for “personal privacy.” 5 U.S.C. § 552(b)(6) (emphasis added). The case law gives effect, as it must,1 to the “personal” modifier, requiring the government to demonstrate either that disclosure of an individual’s name would be linked with personal details about his life or that disclosure would subject an individual to harassment in his personal life.
In United States Department of State v. Ray, for example, the Court invoked both of these circumstances in approving withholding the names of involuntarily repatriated Haitians interviewed pursuant to a State Department investigation. Ray, 502 U.S. at 175-76, 112 S.Ct. 541. It noted that “highly personal information regarding marital and employment status, children, living conditions and attempts to enter the United States, would be linked publicly with particular, named individuals.” Id. In addition, the Court highlighted the danger that Haitians cooperating with the State Department investigation might be subject to retaliatory action for illegally departing Haiti, or that interviewees would be embarrassed “in their social and community relationships.” Id. at 176, 112 S.Ct. 541.
Our cases, too, are consistent with this understanding of Exemption 6’s requirement of “personal” exposure or harassment. Forest Service Employees, for example, permitted the application of Exemption 6 to the names of low-level federal employees because they could personally be subject to “embarrassment and stigma” or “harassment” by litigants and media. 524 F.3d at 1026. Similarly, Lahr affirmed the application of Exemptions 6 and 7(C) where disclosing the names of eyewitnesses to an accident could expose them to “undesired contact!]” by litigants and media. 569 F.3d at 975-77.
*437The majority purports to rely on these cases.2 But none of them upheld application of Exemption 6 solely because the identified individuals provided information to the government. In each case, the government revealed more than that about the circumstances of those individuals whose names the government sought to withhold. And it was that additional information that established the danger of exposure of personal information or of harassment in the informants’ personal lives.
Ray, 502 U.S. at 176, 112 S.Ct. 541, for example, concerned the identity and personal information of repatriated Haitian emigrants potentially vulnerable to government persecution in Haiti for having unlawfully fled the country. Forest Service Employees, 524 F.3d at 1026, concerned low-level government employees,' some of whom were associated with serious allegations of misconduct as to themselves, in the context of pending litigation and much public uproar. And Lahr, 569 F.3d at 975-77, concerned bystander eyewitnesses and investigators of an infamous plane crash, whom the plaintiff expressly wished to contact against the backdrop of substantial media attention. Such additional details — not the bald claim that the identified individuals provided information to the government — established the existence of a personal privacy interest. Moreover, in each of these instances, the underlying investigation, about which the plaintiffs sought records, had been triggered by the government. For that reason, the information sought was “personal” in the sense that it concerned individuals who had not voluntarily entered the public sphere with regard to the information collected by the government.
The majority’s opinion thus breaks new ground by finding in the names of voluntary government informants a per se personal privacy interest. Still worse, the majority’s per se rule departs from statute as well as precedent, because it does nothing to verify that the purported privacy *438interest protected by nondisclosure is personal in nature.
The majority’s rule would see a cognizable “personal privacy” interest in any informant’s name, without more. “Personal privacy” under Exemption 6 is not so broad. Disclosure of the names of those providing information to the government is not “inherently and always a significant threat to the privacy of the [named] individuals.” Ray, 502 U.S. at 176 n. 12, 112 S.Ct. 541. To be sure, an informant’s name refers to his person and is in that narrow sense “personal.” But “ ‘[p]ersonal’ in the phrase ‘personal privacy’ conveys more than just ‘of a person.’ It suggests a type of privacy evocative of human concerns.” FCC v. AT & T, Inc., — U.S.-, 131 S.Ct. 1177, 1183, 179 L.Ed.2d 132 (2011).
In some circumstances, revelation of an informant’s name does not raise the “human concerns” evoked by the term “personal privacy.” Consider, for example, the business executive accusing his competitor of lawbreaking. Such an informant acts on behalf of the business that employs him. And business concerns rarely qualify as “human concerns” of the sort encompassed by “personal privacy.” “[W]e often use the word ‘personal’ to mean precisely the opposite of business-related: We speak of personal expenses and business expenses, personal life and work life....” Id. at 1182 (emphasis in original).
For example, Electronic Frontier Foundation v. Office of the Director of National Intelligence, 639 F.3d 876, 888 (9th Cir.2010), which required the government to disclose the names of business lobbyists, minimized almost to the point of non-existence the notion that there was any “personal privacy” interest in the “nature of their employment.” There, we reasoned that “government acknowledgment of a lobbyist’s lobbying activities does not reveal sensitive personal information about the individual rising to a clearly unwarranted invasion of personal privacy.” Id. (internal quotation marks omitted); see also Wash. Post Co. v. U.S. Dep’t of Justice, 863 F.2d 96, 100 (D.C.Cir.1988) (“Information relating to business judgments and relationships does not qualify” as a personal privacy interests under Exemption 7(C), “even if disclosure might tarnish someone’s professional reputation.”). Similarly, if one firm, acting through an officer or other agent, retaliates in the marketplace against another for blowing the whistle, I see no invasion of personal privacy.
Nor is there necessarily a nontrivial personal privacy interest in the name of a senior government official when such an official acts as an informant. True, “ ‘individuals do not waive all privacy interests in information relating to them simply by taking an oath of public office.’” Lahr, 569 F.3d at 977 (quoting Lissner v. U.S. Customs Serv., 241 F.3d 1220, 1223 (9th Cir.2001)). But the personal privacy interests of senior officials are especially attenuated. Lahr, 569 F.3d at 977 (citing Do-bronski v. FCC, 17 F.3d 275, 280 n. 4 (9th Cir.1994)). Where such a senior official voluntarily submits information to the government to trigger a governmental investigation, his affirmative act may further weaken any putative privacy interest in his name. See Elec. Frontier Found., 639 F.3d at 887-89 (noting that voluntary submission of information intended to affect public policy minimizes an informant’s privacy interest). In an appropriate case, a court could well determine that such a senior government official possesses, at best, a trivial personal privacy interest in withholding his name.
Nor, in the particular circumstances of this case, are there any obvious personal privacy interests jeopardized by the disclosure of an informant’s name if the informant is wholly unaffiliated with either Pru*439dential or the real estate industry. Such an unrelated informant is not especially vulnerable to retaliation or harassment from his employer, co-workers, or friends, as might be true if he worked for Prudential or a related firm. For this reason, too, a privacy interest put in danger “more palpable than [a] mere possibility]” would need to be identified. Rose, 425 U.S. at 380 n. 19, 96 S.Ct. 1592.
2. The majority sidesteps these concerns with unwarranted conjecture. “Given the nature of [the informants’] communications,” the majority speculates, “they appear to have inside knowledge of the mortgage industry in Hawai’i.” Maj. Op. at 432 (emphasis added). Inferences of this variety cannot carry the day in the present context. This is an appeal from the grant of summary judgment against Prudential, so “[w]e draw all reasonable inferences in the light most favorable to the non-moving party.” Corns v. Laborers Int’l Union of N. Am., 709 F.3d 901, 907 (9th Cir.2013). “Appearance” seems to be another word for inference, yet other inferences — such as the inference that the information came from a disgruntled former customer — are also possible.
Even were it otherwise proper, the majority’s inference would remain unreasonable. These informants purported to draw their facts from a newspaper article and a conversation with a Prudential employee, respectively — hardly the stuff of “inside knowledge.”
The majority continues by suggesting that Prudential itself may harass, retaliate against, stigmatize, or embarrass the informants. After all, the majority hints, Prudential went to the expense of filing a FOIA request and, when it was dissatisfied with HUD’s response, initiated a lawsuit that it has carried through appeal. Maj. Op. at 432. Once again, the inference of malice that the majority draws from this litigation history is not its to make. This Court may only take judicial notice of “a fact that is not subject to reasonable dispute.” Fed.R.Evid. 201(b). And Prudential’s motivations for pursuing its legal options are reasonably disputable. It may, for example, wish to lobby HUD to improve its complaint screening process to weed out non-credible informants. Or it may wish to improve its own internal confidentiality policies.
My point is not that these alternative inferences are compelled by the record, or even that they are more reasonable than the majority’s inferences. Rather, the reasonable possibility of such motivations renders the majority’s preferred interpretation “subject to reasonable dispute” and hence beyond the scope of its authority in reviewing a summary judgment record as barren as this one.3
3. In sum, there can be no per se role that those who provide information about the alleged wrongdoing of others have a personal privacy interest in their identity subject to a threat “more palpable than [a] mere possibility],” Rose, 425 U.S. at 380 n. 19, 96 S.Ct. 1592. Because the record in this case leaves us speculating about the personal privacy interests of the infor*440mants, I would remand to the district court to develop the record.
Moreover, even if we could infer from the absence of a record — which we cannot — that there is some nontrivial privacy interest on the part of the informants, we have no clue what it is. Once a nontrivial privacy interest is established, the next stage of the analysis is to evaluate whether invasion of that interest is “clearly unwarranted” by the countervailing public interest in disclosure. See Lahr, 569 F.3d at 973. Executing that balancing analysis requires knowing something about the nature and strength of the privacy interest at stake. We cannot balance a personal privacy interest whose measure we have yet to take.
B.
Not only is the majority’s understanding of “personal privacy” interests too broad, its account of the public policy interests protected by disclosure is too narrow. The majority unnecessarily constricts its inquiry by focusing only on a public interest in exposing government impropriety or inefficiency. It fails to acknowledge, let alone reject, the broader public interest that Prudential advances.
As the majority correctly notes, only a disclosure that “would ‘she[d] light on an agency’s performance of its statutory duties’ or otherwise let citizens know ‘what their government is up to,’ ” qualifies as a cognizable public interest under FOIA. FLRA, 510 U.S. at 497, 114 S.Ct. 1006 (quoting U.S. Dep’t of Justice v. Reporters Comm, for Freedom of the Press, 489 U.S. 749, 773, 109 S.Ct. 1468, 103 L.Ed.2d 774 (1989)). But that does not mean, as the majority’s analysis at some points suggests, that the plaintiff must have demonstrated that the government’s behavior may have been improper or inefficient. “[P]ublic understanding of the operations or activities of the government,” Reporters Comm., 489 U.S. at 775, 109 S.Ct. 1468, includes not only exposure of government misfeasance, but also of governmental effectiveness and other aspects of internal agency functioning.4
Consider, for example, Reporters Committee, 489 U.S. 749, 109 S.Ct. 1468. There, the Supreme Court held out Rose, 425 U.S. 352, 96 S.Ct. 1592, as an example of disclosure that advanced the public interest in understanding the workings of government. That case, Reporters Committee explained, required the disclosure of Air Force Academy disciplinary-hearing summaries because they “obviously contained information that would explain how the disciplinary procedure actually functioned and therefore were an appropriate subject of a FOIA request.” Reporters Comm., 489 U.S. at 773, 109 S.Ct. 1468. The interest was purely informational; suspected government misfeasance had nothing to do with it. Because the public interest in disclosure is broader than the exposure of official wrongdoing, the majority’s preoccupation with Prudential’s failure to allege that “HUD performed either of the two investigations improperly or inefficiently” is beside the point. Maj. Op. at 432-33.
On this more expansive view of the public interest in knowing “what the govern*441ment is up to,” Prudential did explain how disclosure would advance the public knowledge of HUD’s performance of its statutory duties: Prudential argued that HUD’s operations “cannot be fully understood without knowing who initiates or influences investigations.” It articulated concern that HUD had initiated investigations “based on accusations from potentially biased, self-interested, or otherwise unreliable sources.”
Although the majority’s does not so state, this concern is legitimate. Government investigations carry costs, burden the public fisc, and divert resources from other potentially meritorious activities. Meanwhile, targeted individuals and entities are saddled with disruptive inquiries, the costs of compliance, and, above all, tarnished reputations. See, e.g., Richard A. Bierschbach & Alex Stein, Overenforcement, 93 Geo. L.J. 1743, 1771-72 (2005) (collecting sources concerning the “significant extralegal sanctions for the defendants and their employees” associated with investigation of corporate misfeasance or conviction). Agencies, and the firms and individuals they investigate, will bear those costs even when investigation does not support further prosecution. Given the costs and collateral damage, the public is entitled to know what tips its government deems sufficiently reliable to merit investigation.5
Fully evaluating reliability, in turn, usually requires knowing an informant’s name. Reliability is generated by features both external and internal to a statement. External indicia of reliability relate to the identity of the informant. Knowledge of an informant’s identity enables assessment credibility, through what Florida v. J.L., 529 U.S. 266, 270, 120 S.Ct. 1375, 146 L.Ed.2d 254 (2000), termed “reputation.” See Illinois v. Gates, 462 U.S. 213, 233-34, 103 S.Ct. 2317, 76 L.Ed.2d 527 (1983) (noting that “an unquestionably honest citizen” or an informant with a history of “unusual reliability” would be especially credible). Where an informant is “known ... personally and had provided ... information in the past,” for example, his statement is more reliable than “an anonymous telephone tip.” Adams v. Williams, 407 U.S. 143, 146, 92 S.Ct. 1921, 32 L.Ed.2d 612 (1972). By contrast, internal indicia relate to the contents of an informant’s statement — its measure of detail or the evidentiary base on which it purports to rest. See Gates, 462 U.S. at 234, 103 S.Ct. 2317 (noting that an “explicit and detailed description of alleged wrongdoing, along with a statement that the event was observed firsthand,” enhances a tip’s credibility). Without more, however, “ ‘an anonymous tip alone seldom demonstrates the informant’s basis of knowledge or veracity.’ ” J.L., 529 U.S. at 270, 120 S.Ct. 1375 (quoting Alabama v. White, 496 U.S. 325, 329, 110 S.Ct. 2412, 110 L.Ed.2d 301 (1990)). To be deemed reliable, the statement must include independently verifiable details, such as accurate prediction of a suspect’s future behavior. See id. at 270-71, 120 S.Ct. 1375; United States v. Morales, 252 F.3d 1070, 1076 (9th Cir.2001) (reviewing cases concerning anonymous tips). In *442other words, some external validation is always needed; a statement cannot be considered reliable without it. And the informant’s identity is almost always the best way of providing such external validation. Thus, the “marginal additional usefulness,” Forest Serv. Emps., 524 F.3d at 1027 (internal quotation marks omitted), of knowing an informant’s name — that is, the degree to which the informant’s name adds to one’s ability to assess reliability— is substantial. And such a significant aid to the evaluation of reliability must be taken into account in determining whether an invasion of personal privacy is “clearly unwarranted.” 5 U.S.C. § 552(b)(6) (emphasis added).
In particular, the possibility of what Prudential calls “biased” or “self-interested” tips enhances the public interest in identifying informants. As noted, an investigation injures its targets, even if it ultimately exonerates them. For this reason, what children call “snitching” can be an effective way to settle a personal score or gain a competitive advantage in the marketplace.6 If an informant successfully triggers a baseless investigation for personal gain, he will have put government power to private purpose. That danger is preeminently a public concern. Cf. Exxon Shipping Co. v. U.S. Dep’t of Interior, 34 F.3d 774, 779 (9th Cir.1994) (acknowledging, in the context of discovery requests arising out of litigation to which the government was not party, “the government’s serious and legitimate concern that its employee resources not be commandeered into service by private litigants to the detriment of the smooth functioning of government operations”). Such a danger, moreover, is precisely the sort of potential corruption that the FOIA sought to forestall by limiting secrecy.7 And knowing who alleges illegality can be helpful in assessing the scope of such possible corruption. Without names, the public is hampered in evaluating the prevalence of self-interested informants or the magnitude of their manipulation.
I do not understand the majority affirmatively to reject these considerations, but instead to disregard them, evidently as insufficiently raised. I disagree on the latter point. And I expect that a future panel, adjudicating a case where such concerns are fully articulated, will be attentive to the full scope of the public interest in *443learning “what the government is up to” in the processing of complaints by outside informants.
C.
Recognizing the public interest in disclosing informants’ names does not, by any means, preclude the application of Exemption 6. Had HUD produced evidence {in camera, if necessary) concerning its informants’ vulnerability to retaliation, for example, that personal privacy interest might well outweigh the public interest in disclosure, rendering disclosure of the informants’ identities “clearly unwarranted.”
That is not, however, the balance struck by the majority. To the contrary, it arrives at a balance unsupported by the record: It assumes a personal privacy interest where there may well be none, and it ignores countervailing interests that go beyond governmental misconduct. Again, I would remand for further development of the record.
II.
The majority’s treatment of this case is all the more frustrating because it is unnecessary. The history of this case establishes the background for understanding why: In our initial resolution of the matter, we remanded for further factual development as to the personal privacy interests of the informants on the grounds articulated above, supra Part I.A. Prudential Locations, 648 F.3d at 778-79. In a petition for rehearing, however, the government complained that demanding some evidence of a personal privacy interest would burden its administration of the statute, and sought a more deferential rule. In reversing course, today’s decision would seem to have answered the government’s call. The Supreme Court, however, has commanded us to refuse such invitations to enervate FOIA. In any case, Congress has already provided less burdensome means for the government to protect its informants.
A.
The premise of the government’s objection is sound enough: The FOIA is a burden. See, e.g., Antonin Scalia, The Freedom of Information Act Has No Clothes, Regulation, Mar./Apr.1982, at 15 (quipping that FOIA is “the Sistine Chapel of cost-benefit analysis ignored”). In 2012 alone, HUD fielded some 2,544 FOIA requests, which it processed with the equivalent of over thirty staff members at an aggregate cost of over $3 million (not counting litigation expenses). U.S. Dep’t of Housing and Urban Dev., 2012 Annual FOIA Report, at 6,18 (2012).
But these costs cannot be dismissed as useless expenditures. Congress imposed them for a reason — to “ensure an informed citizenry, vital to the functioning of a democratic society, needed to check against corruption and to hold the governors accountable to the governed.” John Doe Agency v. John Doe Corp., 493 U.S. 146, 152, 110 S.Ct. 471, 107 L.Ed.2d 462 (1989) (citation and quotation marks omitted). Consistent with the objective of facilitating disclosure, the Supreme Court has repeatedly commanded that FOIA’s exemptions are exclusive and “ ‘narrowly construed.’ ” Milner v. U.S. Dep’t of Navy, - U.S. -, 131 S.Ct. 1259, 1262, 179 L.Ed.2d 268 (2011) (quoting FBI v. Abramson, 456 U.S. 615, 630, 102 S.Ct. 2054, 72 L.Ed.2d 376 (1982)); see also id. at 1265-66 (collecting additional Supreme Court cases). Indeed, recent jurisprudence has taken a sharply textual approach to interpreting FOIA exemptions. See Peter L. Strauss et ah, Gellhom and Byse’s Administrative Law: Cases and Comments 471 (11th ed.2011) (citing, as *444examples, Milner, 131 S.Ct. 1259, and AT & T, 131 S.Ct. 1177).
This Court cannot relieve the government of the burden that Congress expressly placed upon it. Although a per se rule protecting the identify of government informants would be simple to administer, “we are not free to engraft that policy choice onto the statute that Congress passed.” U.S. Dep’t of Justice v. Landano, 508 U.S. 165, 181, 113 S.Ct. 2014, 124 L.Ed.2d 84 (1993). To the contrary, Congress “created a scheme of categorical exclusion; it did not invite a judicial weighing of the benefits and evils of disclosure on a case-by-case basis.” Abramson, 456 U.S. at 631, 102 S.Ct. 2054.
This Court should, of course, seek to develop “workable rules” for the agencies to follow, Reporters Comm., 489 U.S. at 779, 109 S.Ct. 1468, but only within the confines of the exemptions Congress created. Milner, for instance, rejected on textual grounds a long-settled interpretation of Exemption 2, despite the danger that disclosure might “wreak havoc and make catastrophe more likely.” 131 S.Ct. at 1271 (internal quotation marks and brackets omitted). The Court reasoned that, if “Congress has not enacted the FOIA exemption the Government desires,” or if the exemption that Congress enacted proves unworkable, then the government — here, HUD — may take its complaint to Congress. Id.
In any case, demanding some evidence with which a reviewing court can substantiate and evaluate the nature and magnitude of a putative personal privacy interest is not especially burdensome. Take Lan-dano: That case required an agency invoking FOIA exemption 7(D) to protect the identify of informants to “point to more narrowly defined circumstances that will support” the application of that exemption. Landano, 508 U.S. at 179, 113 S.Ct. 2014. Such a “particularized approach,” the Court held, “is .consistent with Congress’ intent to provide workable rules of FOIA disclosure.” Id. at 180, 113 S.Ct. 2014 (internal quotation marks omitted).
I would require no more. Such evidence need not be adduced via elaborate, costly procedures; affidavits that “contain reasonably detailed descriptions of the documents and allege facts sufficient to establish an exemption” suffice. Bowen v. U.S. Food & Drug Admin., 925 F.2d 1225, 1227 (9th Cir.1991). In those rare circumstances that such affidavits would themselves compromise the identity of informants, they may be reviewed in camera. Landano, 508 U.S. at 180, 113 S.Ct. 2014.
B.
In addition, as the majority recognizes, Maj. Op. at 434-35, much of the difficulty the government has putatively encountered in protecting the identity of voluntary government informers under Exemption 6 is of its own creation. FOIA includes a separate exemption expressly designed to facilitate the protection of informants, but the government did not invoke it here.
Exemption 7(D) authorizes withholding “records or information compiled for law enforcement purposes, but only to the extent” that disclosure “could reasonably be expected to disclose the identity of a confidential source.... ” 5 U.S.C. § 552(b)(7)(D) (emphasis added). “[A] source is ‘confidential,’ ” under this exemption, “if it ‘provided information under an express assurance of confidentiality or in circumstances from which such an assurance could be reasonably inferred.’ ” Ro-senfeld v. U.S. Dep’t of Justice, 57 F.3d 803, 814 (9th Cir.1995) (quoting Landano, 508 U.S. at 172, 113 S.Ct. 2014). The exemption applies equally to the enforce*445ment of civil and criminal law. Church of Scientology Int’l v. IRS, 995 F.2d 916, 919 (9th Cir.1993). Its application is not contingent on the existence of a personal privacy interest in nondisclosure, and does not require balancing that interest against the interest in disclosure. To the contrary, Exemption 7(D) represents Congress’ own assessment of the appropriate balance when it comes to informants.
HUD has not invoked that exemption here. Whether it could sustain HUD’s nondisclosure of these informants’ names is thus a question for another day and another court.
The very existence of Exemption 7(D), however, strongly implies that Exemption 6 applies quite narrowly, if at all, to information concerning government informants. “ ‘[I]t is a cqmmonplace of statutory construction that .the specific governs the general.’ That is particularly time where, as [here], ‘Congress has enacted a comprehensive scheme and has deliberately targeted specific problems with specific solutions.’” RadLAX Gateway Hotel, LLC v. Amalgamated Bank, — U.S. -, 132 S.Ct. 2065, 2071, 182 L.Ed.2d 967 (2012) (quoting Morales v. Trans World Airlines, Inc., 504 U.S. 374, 384, 112 S.Ct. 2031, 119 L.Ed.2d 157 (1992), and Yarity Corp. v. Howe, 516 U.S. 489, 519, 116 S.Ct. 1065, 134 L.Ed.2d 130 (1996) (Thomas, J., dissenting)). Exemption 7(D) specifically applies to the withholding of information provided by informants. And it, unlike Exemption 6,., sets a condition, requiring that any such informant be assured that his identity will remain confidential. Given this statutory scheme, Exemption 7(D) — not Exemption 6 — ought to control the government’s request to withhold information concerning its informants.8
Moreover, HUD’s legitimate desire to withhold informants’ identities can be readily accommodated by implementing procedures to document HUD’s assurance that informants’ names would be kept confidential. Where “the Government has other tools at hand to shield” the information it seeks to withhold, Milner, 131 S.Ct. at 1271, judicial distortion of other of FOIA’s narrow exemptions, to avoid the requirements of the most directly applicable one, is unnecessary as well as illegitimate.
HUD, and other federal agencies, would be well advised to restrict their invocation of FOIA exemptions for voluntary informants to Exemption 7(D). Here, Prudential failed to contest whether HUD’s correspondence with the informants qualified as “similar files” within the meaning of Exemption 6. Future plaintiffs will be less reticent. Meanwhile, the majority, while not rejecting their pertinence, does not weigh in its analysis the full range of the public interests in disclosure, seemingly on the ground of inadequate invocation. Future litigants, I suspect, will be more specific in cases like this one in articulating the public interests in disclosure, so future panels will conduct a more complete balancing analysis.
CONCLUSION
For all these reasons, I respectfully dissent.
. " 'We have long followed the principle that ‘[s]tatutes should not be construed to make surplusage of any provision.' " Local Joint Exec. Bd. of Las Vegas v. NLRB, 657 F.3d 865, 875-76 (9th Cir.2011) (quoting Nw. Forest Res. Council v. Glickman, 82 F.3d 825, 834 (9th Cir.1996)).
. I agree with the majority that treating files such as those in this case, Forest Employees, and Lahr as “similar files" for the purpose of Exemption 6 is problematic, and add some additional reasons why that is so. Maj. Op. at 429-30.
First, the grammatical “rule of the last antecedent” compels a narrow construction of "similar files.” Under that rule, a relative pronoun refers to the nearest reasonable antecedent. See, e.g., Barnhart v. Thomas, 540 U.S. 20, 26-27, 124 S.Ct. 376, 157 L.Ed.2d 333 (2003); Antonin Scalia & Bryan A. Garner, Reading Law: The Interpretation of Legal Texts 144-46 (2012). Here, the relative pronoun “which” attaches only to "similar files” — not "personnel and medical files”— and describes how the relevant files must be "similar” to personnel or medical materials. That description thus indicates that disclosure of the files as a whole (although not every record in the files) must, like the disclosure of personnel and medical files, entail a "clearly unwarranted invasion of personal privacy.”
Second, application of the familiar ejusdem generis canon also suggests that the catch— all term' — "similar files the disclosure of which would constitute a clearly unwarranted invasion of personal privacy” — should be " ‘construed to embrace only objects similar in nature to those objects enumerated by the preceding specific words.’ ” Circuit City Stores, Inc. v. Adams, 532 U.S. 105, 114-15, 121 S.Ct. 1302, 149 L.Ed.2d 234 (2001) (quoting 2A N. Singer, Sutherland on Statutes and Statutory Construction § 47.17 (1991)). “[Pjersonnel and medical files” are, by definition, about individuals and their personal lives — their bodies, their families, their financial circumstances, their criminal and disciplinary records, if any. "Similar files” ought to carry a parallel meaning, and so not reach files that are about the wrongdoing of other people and contain no such personal data about the informant.
. The majority’s attempts to shore up its inference by observing that the district court opinion noted that Prudential had "mention[ed] the possibility of a civil lawsuit against the unidentified individuals for their 'sham' complaints.” Maj. Op. at 432. Any such "mention," however, appears nowhere in the record. Rather, the district court misinterpreted one of Prudential’s legal arguments — that public policy does not protect those who file unsupported complaints with the government — as a statement of Prudential's intentions. In any case, HUD does not rely on the district court's unsupported finding on appeal. The majority should not do so, either.
. See Cooper Cameron Corp. v. U.S. Dep’t of Labor, 280 F.3d 539, 548 (5th Cir.2002) ("[TJhere is a cognizable public interest in monitoring agencies’ enforcement of the law....”); Nation Magazine, Wash. Bureau v. U.S. Customs Serv., 71 F.3d 885, 895 (D.C.Cir. 1995) (recognizing a public interest in understanding the " 'privatization of government functions,’ ” particularly law enforcement investigations); cf. Brock v. Pierce Cnty., 476 U.S. 253, 262, 106 S.Ct. 1834, 90 L.Ed.2d 248 (1986) ("[T]lie protection of the public fisc is a matter that is of interest to every citizen. ...”).
. I note that although Prudential is a business entity, government agencies investigate many individuals for wrongdoing-for example, taxpayers, social security and other benefit recipients, and alleged undocumented immigrants. Undoubtedly, many of these investigations are triggered by information voluntarily provided by members of the public, for motives good and ill. The majority’s approach to this case would presumably apply to such instances as well. That approach could preclude, in the absence of allegations of misfeasance, the revelation of information that might bring to light the criteria used to determine which such informant-provided information is investigated and to what degree.
. These are hardly far-fetched possibilities. See, e.g., Portomene v. United States, 221 F.2d 582, 583 (5th Cir.1955) (describing an allegation that "there was bad blood between [an informant] and the defendant, and that this furnished the motive for [the informant’s] desiring to involve [the defendant] by false charges that he had dealt in narcotics”); Medifast, Inc. v. Minicow, No. 10-CV-382 JLS (BGS), 2011 WL 1157625, at *1 (S.D.Cal. Mar. 29, 2011) (describing an alleged "series of coordinated attacks” on a company, including accusations of regulatory violations, designed 'to increase the value of a short position’ ”).
. FOIA ensures citizen oversight over the government, “a structural necessity in a real democracy.” Nat’l Archives & Records Admin. v. Favish, 541 U.S. 157, 172, 124 S.Ct. 1570, 158 L.Ed.2d 319 (2004). In the debate over the FOIA amendments, legislators sometimes contrasted that commitment to democratic oversight with "the unhappy consequences of the other alternative,” playing out "behind the Iron Curtain.” 1 Freedom of Information Act Source Book: Legislative Materials, Cases, Articles 78 (1974); see also id. at 83 (calling a "policy of secrecy ... the cornerstone of a totalitarian bureaucracy”). And behind the Iron Curtain, where informing was widespread, informants often "snitched” on their rivals for personal gain. See, e.g., Robert Gellately, Denunciations in Twentieth-Century Germany: Aspects of Self-Policing in the Third Reich and the German Democratic Republic, 68 J. Modern Hist. 931, 944-50, 956 (1996) (documenting the prevalence of informing for personal gain in Nazi and communist Germany)..
. The same point applies with even greater force to Exemption 7(C), which the majority urges the government to invoke in the future. Maj. Op. at 434-35. Exemption 7(C) is part of the very same subsection of FOIA as Exemption 7(D), so the general/specific canon should have particular force.