*923Intervenors HD Media Company, LLC ("HDM") and The W.P. Company, LLC, d/b/a the Washington Post ("Washington Post") appeal the district court Opinion and Order holding that the data in the Drug Enforcement Administration's Automation of Reports and Consolidated Orders System ("ARCOS") database cannot be disclosed by Plaintiffs pursuant to state public records requests. Intervenors also argue on appeal that the district court erred in permitting pleadings and other documents to be filed under seal or with redactions.
For the reasons set forth below, we VACATE the district court's Protective Order and its orders permitting the filing of court records under seal or with redactions, and we REMAND to permit the district court to consider entering modified orders consistent with this opinion.
BACKGROUND
This interlocutory appeal arises out of a sweeping multidistrict litigation ("MDL"). Plaintiffs in the MDL consist of about 1,300 public entities including cities, counties, and Native American tribes.1 Defendants consist of manufacturers, distributors, and retailers of prescription opiate drugs.2 The United States Department of Justice and Drug Enforcement Administration (collectively, "the DEA") are not parties to the underlying MDL but are involved in this appeal as Interested Parties-Appellees; HDM and the Washington Post are not parties to the MDL but are involved in this appeal as Intervenors-Appellants.
In the underlying MDL, Plaintiffs seek to recover from Defendants the costs of life-threatening health issues caused by the opioid crisis. The district court presiding over this potentially momentous MDL has repeatedly expressed a desire to settle the litigation before it proceeds to trial. (See, e.g. , R. 800, Opinion and Order, Page ID# 18971 (noting that the court's order will assist "in litigating (and hopefully settling) these cases").)3 President Trump has declared the opioid epidemic a national emergency, and as the district court noted, "the circumstances in this case, which affect the health and safety of the entire country, are certainly compelling." (R. 233, Order Regarding ARCOS Data, Page ID# 1119.)
The crux of this appeal is the question of who should receive access to the data in the DEA's ARCOS database, and the related question of how disclosure of the ARCOS data would further the public's interest in understanding the causes, scope, and context of this epidemic. The ARCOS database is "an automated, comprehensive drug reporting system which monitors the flow of DEA controlled substances from their point of manufacture *924through commercial distribution channels to point of sale or distribution at the dispensing/retail level - hospitals, retail pharmacies, practitioners, mid-level practitioners, and teaching institutions." (R. 717-1, Martin Decl., Page ID# 16517.) The data in the database is provided by drug manufacturers and distributors4 and includes "supplier name, registration number, address and business activity; buyer name, registration number and address; as well as drug code, transaction date, total dosage units, and total grams." (R. 717-1, Page ID# 16517.)
In an order, the district court aptly characterized the opioid epidemic that provides the tragic backdrop of this case, observing that "the vast oversupply of opioid drugs in the United States has caused a plague on its citizens and their local and State governments." (R. 233, Page ID# 1124.) Continuing its plague metaphor, the district court concluded that
Plaintiffs' request for [production of] the ARCOS data, which will allow Plaintiffs to discover how and where the virus grew, is a reasonable step toward defeating the disease. See Buckley v. Valeo , 424 U.S. 1, 67 [96 S.Ct. 612, 46 L.Ed.2d 659] [ (1976) ] ("Sunlight is said to be the best of disinfectants.") (quoting Justice Brandeis, Other People's Money 62 (1933)).
(R. 233, Page ID# 1124-25.) Despite its confidence that disclosing the ARCOS data to Plaintiffs constituted such a reasonable step, the court later rejected the argument that a further reasonable step would be to disclose the data to HDM and the Washington Post (and by extension to the public at large, who would learn about the contents of the ARCOS data via reporting by those entities).
The full quote from Justice Brandeis that the district court cited is as follows: "Publicity is justly commended as a remedy for social and industrial diseases. Sunlight is said to be the best of disinfectants; electric light the most efficient policeman." Buckley , 424 U.S. at 67, 96 S.Ct. 612 (quoting L. Brandeis, Other People's Money 62 (1933)). The question before us is whether it was reasonable for the district court to permit only Plaintiffs to examine the data in the otherwise complete darkness created by the Protective Order, or whether the court abused its discretion by denying Intervenors the opportunity to expose the data to the broad daylight of public reporting. For the reasons below, we hold that this denial was an abuse of the district court's discretion.
The events leading up to this appeal were set into motion when, in the course of the MDL, Plaintiffs subpoenaed the DEA to produce transactional data for all 50 States and several Territories from its ARCOS database. Plaintiffs and the DEA stipulated to a protective order concerning the DEA's disclosure of the ARCOS data. (R. 167, Protective Order, Page ID# 937-44.)5 The district court adopted a Protective *925Order "determin[ing] that any [ ] disclosure [of the ARCOS data] shall remain confidential and shall be used only for litigation purposes or in connection with state and local law enforcement efforts." (R. 167, Page ID# 937.)
The Protective Order by its terms covered "ARCOS data" and defined this term to include "any data produced directly from DEA's ARCOS database; any reports generated from DEA's ARCOS database; any information collected and maintained by DEA in its ARCOS database; and any derivative documents that the parties or their employees, agents or experts create using ARCOS data." (R. 167, Page ID# 938.) The Order pertained to documents, as well as electronically stored information. The court restricted the use of the ARCOS data to "mediat[ing], settl[ing], prosecut[ing], or defend[ing] the above-captioned litigation," and "law enforcement purposes," specifically precluding its use "for commercial purposes, in furtherance of business objectives, or to gain a competitive advantage." (R. 167, Page ID# 939.) The Protective Order also authorized the parties to file pleadings, motions, or other documents with the court that would be redacted or sealed to the extent they contained ARCOS data. However, the court noted that if the parties could not agree to a settlement, "[t]he hearing, argument, or trial w[ould] be public in all respects" and there "w[ould] be no restrictions on the use of any document that may be introduced by any party during the trial" absent order of the court. (R. 167, Page ID# 941.) The Protective Order contemplated the return of the ARCOS data to the DEA after dismissal or entry of final judgment. Significantly for purposes of this appeal, the Protective Order stated that if Plaintiffs received requests for any ARCOS data under "applicable Public Records Laws ('Public Records Requests')," Plaintiffs would "immediately notify the DEA and Defendants of the request." (R. 167, Page ID# 942.) After notification, the DEA and Defendants would be able to challenge the Public Records Request by filing their opposition to production of the records with the court.
After entering this Protective Order and over the objections of the DEA, the district court directed the DEA to comply with Plaintiffs' subpoena by producing ARCOS data pertaining to Ohio, West Virginia, Illinois, Alabama, Michigan, and Florida for the period of 2006 through 2014. (R. 233, Order Regarding ARCOS Data, Page ID# 1104.) Specifically, the DEA was ordered to provide Plaintiffs with Excel spreadsheets identifying
the top manufacturers and distributors who sold 95% of the prescription opiates [ ] to each State [ ] during the time period of January 1, 2006 through December 31, 2014 [ ] on a year-by-year and State-by-State basis, along with [ ] the aggregate amount of pills sold and [ ] the market shares of each manufacturer and distributor.
(R. 233, Page ID# 1109.)
In overruling the DEA's objections to disclosure, the district court found that the DEA had not met its burden of showing "good cause" for not disclosing the data. (R. 233, Page ID# 1111 (citing Fed. R. Civ. P. 45 ).) The court's reasoning is highly relevant to this appeal. Regarding the interest in disclosure of the ARCOS data, the court found that "the extent to which each defendant and potential defendant engaged in the allegedly fraudulent marketing of opioids, filling of suspicious orders, and diversion of drugs ... can be revealed only by all of the data." (R. 233, Order, Page ID# 1118.) Regarding the interests in nondisclosure of the data, the court rejected the arguments that "disclosure would reveal investigatory records compiled *926for law enforcement purposes [and] interfere with enforcement proceedings" and that "disclosure would violate DOJ's policy which prohibits the release of information related to ongoing matters." (R. 233, Page ID# 1119, 1120 (quoting 1:17-op-45041-DAP, R. 101, Page ID# 696, 698).) The court rejected these arguments for three reasons:
First, Plaintiffs seek ARCOS data with an end-date of January 1, 2015. Given that the most recent data is over three years old, it is untenable that exposure of the data will actually or meaningfully interfere with any ongoing enforcement proceeding. Second, the ARCOS data are not pure investigatory records compiled for law enforcement purposes. Rather, the data is simply business records of defendants; these "[c]ompanies are legally required to submit the information" to ARCOS, the database does not include any additional DEA analysis or work-product, and the records are used for numerous purposes besides law enforcement. Indeed, Plaintiffs assert that part of the reason for the opioid epidemic is lack of law enforcement. And third, simply saying that disclosure of ARCOS records dating back to 2006 would detrimentally affect law enforcement does not make it so.
(R. 233, Page ID# 1119 (citation omitted).)
The court similarly rejected an argument that producing the data would cause Defendants "substantial competitive harm" by revealing "details regarding the scope and breadth of [each manufacturer's and distributor's] market share." (R. 233, Page ID# 1120 (alterations in original) (quoting 1:17-op-45041-DAP, R. 101, Page ID# 697).) The court rejected this objection to disclosure because "the assertion was conclusory and ... market data over three years old carried no risk of competitive harm." (R. 233, Page ID# 1120.)
The DEA complied with the court's order and produced the relevant spreadsheets. Production of the ARCOS data allowed Plaintiffs to identify and add as defendants previously-unknown entities involved in the manufacturing and distribution of opioids and to identify and remove as defendants improperly-named entities. The court noted that other benefits of the ARCOS data included "allowing [the litigation] to proceed based on meaningful, objective data, not conjecture or speculation" and "providing invaluable, highly-specific information regarding historic patterns of opioid sales." (R. 397, Second Order Regarding ARCOS Data, Page ID# 5323.) To expand upon these benefits, the court ordered the DEA to produce further ARCOS data pertaining to "all of the States and Territories" for the same period of 2006 to 2014, with such disclosure being subject to the Protective Order. (R. 397, Page ID# 5323.)
Once the complete production of the ARCOS data occurred, HDM filed a West Virginia Freedom of Information Act request with the Cabell County Commission seeking the ARCOS data that the county received as a Plaintiff in this litigation, and the Washington Post filed similar public records requests with Summit and Cuyahoga counties in Ohio (also Plaintiffs in this litigation). Pursuant to the Protective Order, the three counties notified the district court, Defendants, and the DEA of the requests, and the DEA and Defendants objected to them.
The district court granted HDM and the Washington Post limited Intervenor status "for the limited purpose of addressing their Public Records Requests." (R. 611, Briefing Order, Page ID# 14995.) The arguments in the subsequent briefing as to why the Protective Order should or should not be modified to allow disclosure of the ARCOS data pursuant to Intervenors'
*927requests largely tracked the arguments that had been made on the DEA's earlier objection to disclosing the ARCOS data to Plaintiffs: Defendants argued that the ARCOS data "is sensitive from the perspective of both the pharmacies and distributors because it is confidential business information, and it is sensitive from the perspective of DEA because it is crucial to its law-enforcement efforts." (R. 665, Defendants' Br. Opposing Disclosure, Page ID# 16012.) Intervenors argued that the risk of harm to Defendants and the DEA was speculative and conclusory, and that the public had a compelling interest in receiving "a more complete and accurate story" of a national emergency, which the ARCOS data would allow Intervenors to tell. (R. 718, Wash. Post Br. Supporting Disclosure, Page ID# 16534; see also R. 725, HDM Br. Supporting Disclosure, Page ID# 16601-16.)6 In an Opinion and Order, the district court held that the public records requests must be denied because the requests were barred by the court's Protective Order and Defendants and the DEA had demonstrated "good cause" for the Protective Order's application to such requests, as required under Rule 26(c)(1). (R. 800, Page ID# 18978.) The court specified that its holding extended to all present or future public records requests for the ARCOS data filed with any of the 1,300 public entity Plaintiffs in the underlying litigation.
In its analysis, the district court adopted language from Defendants' briefing, noting that the ARCOS data "is sensitive to pharmacies and distributors because it is confidential business information; and it is sensitive from the DEA's perspective because it is crucial to law enforcement efforts." (R. 800, Page ID# 18979-80.) The court further noted that the Freedom of Information Act ("FOIA") "exempts from public disclosure any confidential commercial information, the disclosure of which is likely to cause substantial competitive harm." (R. 800, Page ID# 18980 (citing 5 U.S.C. § 552(b)(4) and Canadian Commercial Corp. v. Dep't of Air Force , 514 F.3d 37, 39 (D.C. Cir. 2008) ).) It also found relevant that FOIA exempts "records or information compiled for law enforcement purposes, but only to the extent that the production of such law enforcement information could reasonably be expected to interfere with enforcement proceedings and criminal prosecutions." (R. 800, Page ID# 18981 (citing 5 U.S.C. § 552(b)(7) ).) Finally, the court concluded that the ARCOS data "is not a record generated by the Counties" that would be subject to state public records requests. (R. 800, Page ID# 18981.)
Intervenors appealed the Opinion and Order to this Court.
DISCUSSION
Because the DEA challenges this Court's jurisdiction to hear this appeal, we begin with that issue. We will then address whether the district court abused its discretion in finding "good cause" to support its Protective Order forbidding Plaintiffs to disclose the ARCOS data pursuant to state public records requests. Finally, we will address whether the district court erred in allowing court records to be filed under seal or with redactions.
*928I. Jurisdiction
We determine our own jurisdiction de novo. Abu-Khaliel v. Gonzales , 436 F.3d 627, 630 (6th Cir. 2006).
While Defendants concede that Intervenors can appeal the district court order, the DEA disagrees, arguing that this Court lacks jurisdiction over this appeal because it does not concern a final order.
Under 28 U.S.C. § 1291, this Court "ha[s] jurisdiction of appeals from all final decisions of the district courts of the United States." The DEA argues that the district court's Opinion and Order is not a final order under § 1291 because "[t]he district court has not entered judgment in the MDL from which these consolidated appeals arise; the litigation instead remains active." (DEA Br. 27.) For purposes of § 1291, a "final decision" "does not necessarily mean the last order possible to be made in a case." Gillespie v. U.S. Steel Corp. , 379 U.S. 148, 152, 85 S.Ct. 308, 13 L.Ed.2d 199 (1964). Rather, "the requirement of finality is to be given a 'practical rather than a technical construction.' " Id. (quoting Cohen v. Beneficial Indus. Loan Corp. , 337 U.S. 541, 546, 69 S.Ct. 1221, 93 L.Ed. 1528 (1949) ).
The collateral order doctrine first identified in Cohen gives content to the finality requirement. Pursuant to that doctrine, an order that does not terminate a case may be appealed, but the order "(1) must be 'conclusive' on the question it decides, (2) must 'resolve important questions separate from the merits' and (3) must be 'effectively unreviewable' if not addressed through an interlocutory appeal." Swanson v. DeSantis , 606 F.3d 829, 833 (6th Cir. 2010) (quoting Mohawk Indus., Inc. v. Carpenter , 558 U.S. 100, 106, 130 S.Ct. 599, 175 L.Ed.2d 458 (2009) ). Further, "[t]he justification for immediate appeal must ... be sufficiently strong to overcome the usual benefits of deferring appeal until litigation concludes." Id. (quoting Mohawk Indus. , 558 U.S. at 107, 130 S.Ct. 599 ).
The DEA acknowledges that this Court has found "collateral-order jurisdiction over an appeal by a media company that was denied access to sealed court filings and transcripts," (DEA Br. 27 (discussing Nat'l Broad. Co. v. Presser , 828 F.2d 340 (6th Cir. 1987) )), but it suggests that intervening precedent has undermined that decision. In support of that proposition, the DEA cites broad statements in which the Supreme Court "has repeatedly clarified the 'modest scope' of the collateral-order doctrine." (DEA Br. 28 (citing Will v. Hallock , 546 U.S. 345, 350, 126 S.Ct. 952, 163 L.Ed.2d 836 (2006) ).) However, this is not a post- Presser development: from the collateral order doctrine's inception, the Supreme Court has acknowledged that the doctrine only applies to a "small class" of decisions. Cohen , 337 U.S. at 546, 69 S.Ct. 1221.
Presser is on all fours with this case, and the DEA cites no persuasive reason to stray from this binding precedent. In Presser , NBC sought media access to sealed records relating to the federal government's ongoing prosecution of Jackie Presser. 828 F.2d at 341. After the district court denied NBC's application for access to the documents, NBC appealed to this Court the district court's memorandum and order directing that all documents remain under seal. Id. at 341-43. The DEA is correct that this Court did not provide much analysis. Nevertheless, it unequivocally held, "Although all of these orders are interlocutory with respect to the underlying case, we have jurisdiction of this appeal pursuant to 28 U.S.C. § 1291 [because] NBC was permitted to intervene in the district court, and the orders satisfy the 'collateral order doctrine' set forth in *929Cohen [.]" Id. at 343 (citing Cohen , 337 U.S. 541, 69 S.Ct. 1221 ). Moreover, in Presser , this Court cited Application of The Herald Co. , in which the Second Circuit collected cases where federal courts of appeals found appellate jurisdiction to decide whether to grant intervenors access to evidence in pending litigation. Id. ; see Application of The Herald Co. , 734 F.2d 93, 96 (2d Cir. 1984) (collecting cases).
Indeed, little analysis is necessary to demonstrate that Intervenors meet the three Swanson requirements. First, the district court's Opinion and Order was conclusive on the question of public records requests for the ARCOS data, see Swanson , 606 F.3d at 833, in that its decision applied to all present or future public records requests for the ARCOS data filed with any of the 1,300 public entity Plaintiffs in the underlying litigation and no further consideration of this issue will be possible. Further, the broad scope of the order provides "sufficiently strong [justification] to overcome the usual benefits of deferring appeal." Swanson , 606 F.3d at 833 (quoting Mohawk Indus. , 558 U.S. at 107, 130 S.Ct. 599 ).
The order also plainly resolved important questions separate from the merits of the litigation, satisfying the second Swanson requirement. See id. at 833. The final requirement is that the order "be 'effectively unreviewable' if not addressed through an interlocutory appeal." Id. (quoting Mohawk Indus. , 558 U.S. at 106, 130 S.Ct. 599 ). The DEA argues that neither the first nor third element is satisfied because there remains a possibility of trial, at which the ARCOS data may become public. The possibility of trial was certainly also present in Presser , and would seem to be present in virtually every case involving an interlocutory appeal. Thus, contrary to the DEA's assertion, the possibility of trial cannot be a categorical bar to appellate jurisdiction pursuant to the collateral order doctrine. Further, given the district court's strong desire for settlement, disclosure of the ARCOS data at trial in this case is not certain or even necessarily likely.
Because Intervenors' stake in the litigation pertains only to disclosure of the ARCOS data and because the district court's Opinion and Order finally and conclusively decides that issue, we possess jurisdiction over this appeal of the Opinion and Order.
II. "Good Cause" for the Protective Order
This Court reviews the question of whether a district court's protective order was premised upon a showing of good cause for an abuse of discretion. The Courier-Journal v. Marshall , 828 F.2d 361, 364 (6th Cir. 1987).
A protective order shall only be entered upon a showing of "good cause" by the party seeking protection. Fed. R. Civ. P. 26(c)(1). Rule 26(c) contemplates the issuance of protective orders "to protect a party or person from annoyance, embarrassment, oppression, or undue burden or expense." Fed. R. Civ. P. 26(c)(1). To show good cause for a protective order, the moving party is required to make "a particular and specific demonstration of fact, as distinguished from stereotyped and conclusory statements." Nemir v. Mitsubishi Motors Corp. , 381 F.3d 540, 550 (6th Cir. 2004) (quoting Gulf Oil Co. v. Bernard , 452 U.S. 89, 102 n.16, 101 S.Ct. 2193, 68 L.Ed.2d 693 (1981) ). A district court abuses its discretion where it "ma[kes] neither factual findings nor legal arguments supporting the need for" the order. Gulf Oil Co. , 452 U.S. at 102, 101 S.Ct. 2193. Despite these formal requirements, "it is common practice for parties to stipulate to [protective] orders."
*930Procter & Gamble Co. v. Bankers Tr. Co. , 78 F.3d 219, 229 n.1 (6th Cir. 1996) (Brown, J., dissenting). Protective orders "are often blanket in nature, and allow the parties to determine in the first instance whether particular materials fall within the order's protection." Shane Grp., Inc. v. Blue Cross Blue Shield of Mich. , 825 F.3d 299, 305 (6th Cir. 2016).
Because parties may stipulate to a protective order, courts sometimes permit intervenors to challenge protective orders. See, e.g. , Presser , 828 F.2d at 341. If an intervenor challenges a protective order, "the burden of proof will remain with the party seeking protection when the protective order was a stipulated order and no party had made a 'good cause' showing." Phillips ex rel. Estates of Byrd v. Gen. Motors Corp. , 307 F.3d 1206, 1211 n.1 (9th Cir. 2002).
In this case, the parties stipulated to a protective order that would prevent Plaintiffs from disclosing the ARCOS data to the media, and the district court did not make a good cause finding on this issue before entering its Protective Order. The dissent disputes that the parties stipulated to the relevant aspects of the Protective Order, arguing that "the parties energetically fought over the terms of the protective order and never, in fact, fully agreed to all its terms." (Dissent at 31.) We disagree. It is true that during the parties' initial negotiations over disclosure of the ARCOS data (outside the presence of the district court), Plaintiffs "opposed the entry of a broad protective order and recommended that the data be disclosed leaving to the discretion of the Court the ability to share data and/or reports generated therefrom with ... the media." (R. 137, Status Report, Page ID# 742.) However, the scant treatment that this issue receives in the parties' status reports on their disclosure negotiations (compared with issues relating to the scope and content of the data to be disclosed) suggests that this was not a central issue in the parties' discussions. More importantly, at a hearing after these negotiations-which represented the first opportunity Plaintiffs had to raise before the district court the issue of public disclosure of the ARCOS data-Plaintiffs declined to raise this issue. In fact, it does not appear that the district court was even aware that this issue was disputed, stating, "No one is proposing making all this publicly available." (R. 156, Hearing Tr., Page ID# 566.) It is a grave mischaracterization to state that Plaintiffs "energetically fought" over the issue of public disclosure when they neither raised it before the district court nor even objected when the district court stated that the issue was not disputed. Plaintiffs may have suggested the possibility of public disclosure in initial negotiations with Defendant, but they failed to ever raise this issue before the district court and instead stipulated to a protective order that barred public disclosure.
Because the issue of public disclosure of the ARCOS data was never squarely raised before the district court, the court never had occasion to find that Defendants or the DEA had made "a particular and specific demonstration of fact" justifying the Protective Order's permanent blanket ban on such disclosure. Nemir , 381 F.3d at 550. The dissent points to conclusory statements by the district court that "[n]othing is going to be revealed to the media unless there's a trial," as though these statements amounted to a good cause finding. (Dissent at 30 (quoting R. 156, Page ID# 861).) As mentioned, it is unclear that the district court was aware that this issue was disputed at all, so it seems unlikely the court intended these statements to represent a finding of good cause for this aspect of the Protective Order. Moreover, even if the district court intended to make a good *931cause finding, it failed to do so because it "made neither factual findings nor legal arguments supporting the need for" this aspect of the Protective Order, which it must do in order "to provide a[ ] record useful for appellate review." Gulf Oil Co. , 452 U.S. at 102, 101 S.Ct. 2193.
Accordingly, although Intervenors challenge the Protective Order, the burden of demonstrating good cause not to disclose the ARCOS data remains with the DEA and Defendants (as the parties seeking protection). See Phillips , 307 F.3d at 1211 n.1.
Despite the "substantial latitude" afforded to district courts during the discovery process, see Seattle Times Co. v. Rhinehart , 467 U.S. 20, 36, 104 S.Ct. 2199, 81 L.Ed.2d 17 (1984), we hold that the district court abused its discretion in finding that good cause existed to permanently and categorically prevent the ARCOS data from being disclosed pursuant to public records requests. In considering whether good cause for protection exists, we balance the interests in favor of disclosure against the interests in favor of nondisclosure. See The Courier-Journal , 828 F.2d at 367 ; Signature Mgmt. Team, LLC v. Doe , 876 F.3d 831, 838 (6th Cir. 2017). Accordingly, we will balance Intervenors' interest in reporting on the ARCOS data and the public interest in learning what such reporting would reveal against Defendants' and the DEA's interest in keeping the ARCOS data secret. We will also bear in mind that it was the burden of Defendants and the DEA to demonstrate good cause with particularity. See Nemir , 381 F.3d at 550 ; Phillips , 307 F.3d at 1211 n.1.
Ironically, the best evidence that good cause did not exist for the Protective Order comes from the district court's own balancing of the interests in disclosure versus nondisclosure.
In ordering the DEA to disclose the ARCOS data to Plaintiffs, the district court specifically held that the DEA did not meet its burden of showing "good cause" not to comply with Plaintiffs' subpoena for the ARCOS data. (R. 233, Page ID# 1111 (citing Fed. R. Civ. P. 45 ).) The court noted that the data "provid[es] invaluable, highly-specific information regarding historic patterns of opioid sales," (R. 367, Page ID# 5323), and emphasized that the role each Defendant played in the crisis "can be revealed only by all of the data ." (R. 233, Page ID# 1118 (emphasis added).) The district court, comparing the opioid crisis to a plague, even stated that because it is possible to "discover how and where the virus grew" by studying the ARCOS data, disclosure of the ARCOS data "is a reasonable step toward defeating the disease." (R. 233, Page ID# 1124-25.)
In the same order concerning disclosure to Plaintiffs, the district court rejected Defendants' and the DEA's arguments that there was "good cause" for nondisclosure. The court specifically rejected the DEA's arguments that disclosing the data would interfere with law enforcement interests. Emphasizing the speculative nature of the harm given the age of the data, the court concluded that "it is untenable that exposure of the data will actually or meaningfully interfere with any ongoing enforcement proceeding." (R. 233, Page ID# 1119.)7 In sum, the district court found the DEA's stated law enforcement interests to be vague and attenuated. (See R. 233, Page ID# 1119 ("[S]imply saying that disclosure of ARCOS records dating back to 2006 would detrimentally affect law enforcement does not make it so.").)
*932The court likewise rejected the argument that producing the data would cause Defendants competitive harm, explaining that "the assertion was conclusory and ... market data over three years old carried no risk of competitive harm." (R. 233, Page ID# 1120 (emphasis added).)
Between the time it ordered the DEA to produce the ARCOS data to Plaintiffs and the time it denied Intervenors' requests for the data, the district court seems to have done a complete about-face concerning the relevant interests at stake. It is true that this about-face might be explained in part by the different interests at stake when disclosure is made only to parties to a case pursuant to a protective order, as compared to third parties that intend to publicly report on the disclosed information. Cf. Shane Grp. , 825 F.3d at 305 (recognizing that there is a lower requirement for protective orders relating to discovery, during which secrecy is permitted, than for orders to seal court records, which carry a strong presumption of openness).8 In other words, the fact that the district court ordered the DEA to disclose the ARCOS data to Plaintiffs pursuant to the Protective Order does not necessarily imply that the same considerations would require disclosing that data to Intervenors and, by extension, the public.
However, it is readily apparent from the record that the district court's analysis in its first order did take into account the public's interest in obtaining the ARCOS data and the interests of Defendants and the DEA in keeping this data from the public.9 If the district court ordered the DEA to disclose the ARCOS data with the understanding that it would only be seen by Plaintiffs and only used for litigation purposes, there would have been no reason to write that "market data over three years old carried no risk of competitive *933harm." (R. 233, Page ID# 1120 (emphasis added).) Nor would it have been necessary to state that "[g]iven that the most recent data is over three years old, it is untenable that exposure of the data will actually or meaningfully interfere with any ongoing enforcement proceeding." (R. 233, Page ID# 1119.) These statements speak to the interests that Defendants and the DEA had in keeping the ARCOS data away from public eyes-not just the eyes of Plaintiffs. The dissent argues that we take these quotes out of context; however, the totality of the district court's balancing analysis supports our position and nothing quoted in the dissent suggests otherwise.
Given the balancing of interests in its order compelling the DEA to disclose the ARCOS data to Plaintiffs, it is bizarre that the district court could later hold that the ARCOS data at issue "is sensitive to pharmacies and distributors because it is confidential business information; and it is sensitive from the DEA's perspective because it is crucial to law enforcement efforts." (R. 800, Page ID# 18979-80.) The district court repeatedly expressed its desire that the underlying litigation settle before proceeding to trial. The court also warned the parties when it was considering a protective order that if the case went to trial, the ARCOS data would likely become public. (See R. 156, Page ID# 861 ("Nothing is going to be revealed to the media unless there's a trial. If there's a trial, obviously trials in our country are public. Hopefully there will be no trials.").) These statements suggest that at least part of the reason for the district court's about-face on what interests Defendants and the DEA have in nondisclosure of the ARCOS data might have been a desire to use the threat of publicly disclosing the data as a bargaining chip in settlement discussions. If this was a motivation for its holding, then the district court abused its discretion by considering an improper factor. See Just Film, Inc. v. Buono , 847 F.3d 1108, 1115 (9th Cir. 2017) ("An abuse of discretion occurs when the district court, in making a discretionary ruling, relies upon an improper factor[.]" (quoting Parra v. Bashas', Inc. , 536 F.3d 975, 977-78 (9th Cir. 2008) )). And even if this was not part of the district court's motivation, it appears that the court abused its discretion by acting irrationally. See United States v. Swift , 809 F.2d 320, 323 (6th Cir. 1987) (noting that a court of appeals should "uphold the trial judge's exercise of discretion unless he acts arbitrarily or irrationally" (quoting United States v. Robinson , 560 F.2d 507, 515 (2d Cir. 1977) )).
Further, the district court was largely correct in its initial analysis of the relevant interests in this case: Intervenors, as representatives of the public, have a substantial interest in disclosure of the ARCOS data, while the DEA and Defendants have only a lesser interest in avoiding potential harms that can be avoided by narrower, less categorical means. The district court correctly observed that the ARCOS data "provid[es] invaluable, highly-specific information regarding historic patterns of opioid sales." (R. 397, Page ID# 5323.) The ARCOS data will aid us in understanding the full enormity of the opioid epidemic and might thereby aid us in ending it.
Intervenors' reporting bears out these conclusions. HDM was able to receive some ARCOS data from West Virginia's Attorney General in a previous, unrelated litigation. This data included "hundreds of printed pages of ARCOS data spreadsheets that revealed the number of hydrocodone and oxycodone dosage units sold to every retail pharmacy in West Virginia from 2007 to 2012." (HDM Br. 9.) That data was used in HDM's extensive reporting on the opioid crisis, which was awarded a Pulitzer Prize for exposing the causes, *934context, and scope of the epidemic. Reporting by Intervenors also prompted a committee of the House of Representatives to investigate and issue a report on the opioid epidemic. See Energy and Commerce Committee, Red Flags and Warning Signs Ignored: Opioid Distribution and Enforcement Concerns in West Virginia (2018), available at https://www.ruralhealthinfo.org/assets/2616-9819/OpioidDistribution-Report-FinalREV.pdf.
The DEA and Defendants attempt to undermine the importance of the ARCOS data in educating the public about and drawing attention to the opioid crisis. Defendants argue that Intervenors "cannot explain why they need transaction-level data ... to educate the public about the depth and magnitude of the prescription drug crisis" when "publicly-available reports [provide] the volume of opioids distributed per quarter in any three-digit zip code prefix." (Defendants Br. 34, 35 (citation omitted).) Intervenors respond:
The aggregate data [ ] identifies narcotics only by weight and the number of grams that were shipped to a generalized geographic area; it does not identify the number of pills that were shipped, the type of pills that were shipped, the dosage units of the pills, the pharmacy that ordered the pills, or the manufacturers and the distributors that shipped them. This is all extraordinarily relevant information, essential to learn how, in little more than a decade, routine drug abuse escalated into the worst drug epidemic in American history.
(Wash. Post Reply Br. 3.) Intervenors convincingly argue that "[t]he dosage of the pill is of immense public interest, as people want to know whether their neighborhood was supplied with 5 mg oxycodone pills, such as Percocet, which are generally prescribed for minor dental procedures and routine injuries, or 30 mg oxycodone tablets, which have been shown to be the most abused and diverted pills[.]" (Wash. Post Reply Br. 2.)
Defendants' argument that aggregate data is sufficient might be more availing if there were no direct, tangible evidence of the compelling nature of specific transactional data. But, as Intervenors point out, specific transactional data has proved extremely effective and consequential in calling attention to the horrors of the opioid crisis. For example, in a report on the opioid crisis in West Virginia, the Energy and Commerce Committee of the United States House of Representatives noted that it became interested in the crisis after reading reporting in the Charleston Gazette-Mail (part of HDM) and the Washington Post. Energy and Commerce Committee Report, supra at 4. Not only did the Committee specifically reference reporting by Intervenors, it called out for special attention details from their reporting, like one instance in which "distributors sent more than 20.82 million doses of hydrocodone and oxycodone to two pharmacies located four blocks apart in a town of approximately 3,000 people" and another in which "a single pharmacy in a town of 406 people received nearly 13 million doses of hydrocodone and oxycodone from all distributors between 2006 and 2012." Energy and Commerce Committee Report, supra at 100. The available aggregate data does not provide such granular detail as the number of doses sent to individual pharmacies, meaning that this reporting would have been impossible without the ARCOS data. Thus, Intervenors have presented substantial evidence of the significant public interest in transactional-level data.
By contrast, as the district court recognized, most of Defendants' and the DEA's asserted interests pertain only to the potential for future harm. In its order requiring *935disclosure of the ARCOS data to Plaintiffs, the district court concluded that these harms were vague and speculative, even suggesting that there was no law enforcement interest in the data due to its age.10
We need not find, as the district court seemed to, that Defendants' and the DEA's interests carry no weight in order to hold that there was not "good cause" to protect the ARCOS data from disclosure pursuant to state public records requests. It is true that some of the identified harms are not sufficiently particularized to carry much weight, like the DEA's vague assertion that disclosing the ARCOS data "would undermine DEA's mission of investigating and prosecuting misconduct involving controlled substances." (DEA Br. 44.) How this interest would be impeded by the public release of the data is not made clear.
The law enforcement interests in the ARCOS data identified in the declaration of DEA Assistant Administrator John J. Martin are somewhat more concrete. Martin notes, "Frequently, DEA investigations remain open for multiple years .... Therefore, it is not unusual for ARCOS data first generated a decade ago to continue to have relevance in ongoing investigations and enforcement actions." (R. 717-1, Page ID# 16519.) But insufficient explanation is given as to how law enforcement interests are furthered by permanently and categorically keeping confidential data that is at least four years old. Even accepting Martin's statements, it is undeniable that data becomes less valuable as it ages-particularly in the case of ARCOS data, because there is a five-year statute of limitations on controlled substance offenses. 18 U.S.C. § 3282.
Moreover, the interests set forth in Martin's declaration also suffer from a lack of particularity. In a redacted portion of his declaration, Martin notes an example of one ongoing case that the disclosure of the ARCOS data could impede: "Public release of ARCOS data that is the subject of this pending action would be detrimental to DEA's prosecution of [an administrative action involving DEA's efforts to revoke a distributor's DEA Certificate of Registration] because DEA intends to provide testimony regarding ARCOS data in this action." (R.662-1, SEALED Martin Decl., Page ID# 15973.)11 It is not clear what this statement means or what *936we are supposed to take from it. Martin does not attempt to explain what the ARCOS data in the action will evidence or the nature of the testimony about the data. If the testimony will simply establish how the ARCOS database operates, for example, no law enforcement interest would be compromised by disclosing the ARCOS data to Intervenors. Martin's declaration is simply too vague in its evaluation of the law enforcement interests at issue to demonstrate "good cause" for a blanket, permanent Protective Order. Similarly, the one-page report included with Martin's declaration that provides the number of "open cases from 2006 to 2014 involving opioids" without explaining the nature or status of any of those cases, (R. 663-1, Page ID# 16001, 16005), fails to establish "good cause" for the Protective Order.
It is important to emphasize that the ARCOS data "are not pure investigatory records compiled for law enforcement purposes, [but] simply business records of defendants; ... the database does not include any additional DEA analysis or work-product[.]" (R. 233, Page ID# 1119 (emphasis added).) At oral argument, the DEA was questioned about why a permanent blanket ban on disclosure was needed rather than a narrower protective order that would permit the DEA to object to disclosure of specific pieces of ARCOS data as they relate to specific investigations. The DEA responded that "if we delete [data relating to a specific investigation] from [the ARCOS] database and give [Intervenors] the data without those things, I suspect that [a] manufacturer [whose data was not included in the disclosure] will say, 'Huh, the Washington Post published this dataset that removed everything that was related to an ongoing investigation and I see that I'm not on there, so maybe I'm the subject of an ongoing investigation.' " (May 2, 2018, Oral Arg. 43:40-44:00.) However, it is difficult to understand this response given the nature of the ARCOS data. This response seems to assume that the DEA is unable to disclose data about a manufacturer under investigation-but it is unclear why this should be the case. If data about a manufacturer is included in the ARCOS data, it is only because the manufacturer (or an entity with which it transacted) kept the data as a business record and submitted it to the DEA. There would therefore be no compelling need for the DEA to hide the information from the very manufacturer that likely provided the information to the DEA. This is not to say that there could never be a law enforcement interest in keeping ARCOS data secret, but the DEA has not adequately explained why the data should be subject to a permanent blanket ban on disclosure, rather than a narrower protective order that would allow the DEA to object to disclosure as specific investigations may require.
Further, the DEA's argument as to the risk to law enforcement interests if the data is disclosed is undermined to some degree by the DEA's failure to point to any harm caused by HDM's reporting on the ARCOS data it received from the West Virginia Attorney General in 2016. Instead of doing so, the DEA asserts, without further explanation, that "HD[M] is in no position to assess the harm that publication of sensitive federal law-enforcement data may have done to DEA's law-enforcement activities." (DEA Br. 45.) The DEA argues that this prior disclosure "says nothing about the jeopardy that a much broader disclosure would create for the federal government's law-enforcement activities." (DEA Br. 45.) We disagree. At the very least, the fact that this disclosure occurred and the DEA cannot point to any resulting harm demonstrates that there is little chance of imminent harm from disclosure of the ARCOS data. In sum, the *937DEA's stated law enforcement interests do not seem very weighty, given that they primarily pertain to potential future harms that could be avoided by limited redactions to those particular portions of the ARCOS data that correspond to specific ongoing investigations.12
Last, but importantly, the DEA has never explained why it could not simply redact the portions of the ARCOS data that relate to this and other ongoing investigations. Cf. Madel v. U.S. Dep't of Justice , 784 F.3d 448, 453 (8th Cir. 2015) (holding that the DEA could "not automatically withhold an entire document when some information is exempt" from production). Our "good cause" inquiry takes into account "[t]he scope of the protective order" as it relates to the relevant interests. In re Ohio Execution Protocol Litig. , 845 F.3d 231, 238 (6th Cir. 2016) ; see also The Courier-Journal , 828 F.2d at 366. Because the Protective Order in this case prevented any disclosure of any ARCOS data by any Plaintiff, and because this ban on disclosure would remain in effect in perpetuity, the DEA and Defendants faced a high hurdle in demonstrating "good cause" for these extreme restrictions.
With respect to Defendants' interests, the district court correctly noted the great "public interest in solving the opioid crisis" and held that these and other interests "outweigh[ed] any slight risk of anticompetitive harm." (R. 199, Order, Page ID# 1008-09.) The ARCOS data does not contain sensitive information like trade secrets, and the age of the data makes the risk of anticompetitive harm slight and speculative. See United States v. Int'l Bus. Machines Corp. , 67 F.R.D. 40, 49 (S.D.N.Y. 1975) (rejecting a business's request for continued protection of its commercial data, all of which was at least two years old, because "it reveals directly little, if anything at all, about [the business's] current operations" and because "the value of this data to [ ] competitors is speculative."). Defendants have not alleged any harm resulting from the publication of the ARCOS data HDM received from the West Virginia Attorney General in 2016. Defendants underscore the speculative nature of the harm they assert in stating that "[i]t likely is too soon in any event to draw firm conclusions about the competitive harm caused by those earlier disclosures." (Defendants' Br. 31.) Defendants have offered no new reasons on appeal to question the district court's analysis of their interest in nondisclosure,13 and we conclude *938that Defendants' interests are far outweighed by the specific, concrete interest Intervenors and the public have in disclosure of the ARCOS data.
The reporting on the ARCOS data that HDM received from the West Virginia Attorney General resulted in no demonstrated commercial harm to Defendants and no demonstrated interference with law enforcement interests; but this reporting did result in a Pulitzer Prize, a Congressional Committee report, and a broader public understanding of the scope, context, and causes of the opioid epidemic. Further disclosure of the ARCOS data is warranted because the DEA and Defendants have failed to demonstrate "good cause" not to disclose the data to Intervenors. As the district court acknowledged, "[s]unlight is said to be the best of disinfectants," and the ARCOS data and the insight it will provide into the opioid epidemic should be brought to light. Buckley , 424 U.S. at 67, 96 S.Ct. 612 (quoting L. Brandeis, Other People's Money 62 (1933)).
For the foregoing reasons, we hold that the district court abused its discretion in finding "good cause" not to permit disclosure of the ARCOS data pursuant to state public records requests. We vacate the district court's Protective Order and remand to permit the district court to consider entering a new protective order consistent with the proper legal standards as set forth in this opinion. On remand the district court may entertain arguments by the DEA as to why particular pieces of ARCOS data that relate to specific ongoing investigations should not be disclosed; however, the district court shall not enter a blanket, wholesale ban on disclosure pursuant to state public records requests. Nor shall any modified protective order specify that the ARCOS data be destroyed or returned to the DEA at the conclusion of this litigation.
III. Sealing and Redaction of Pleadings
Intervenors argue that the district court erred in allowing Defendants and the DEA to file pleadings and other court documents under seal and with redactions. We review a court's decision to seal its records for an abuse of discretion, but we note that " '[i]n light of the important rights involved, the district court's decision is not accorded' the deference that standard normally brings." Shane Grp. , 825 F.3d at 306 (quoting In re Knoxville News-Sentinel Co. , 723 F.2d 470, 476 (6th Cir. 1983) ).
As an initial matter, because the district court allowed Intervenors to intervene "for the limited purpose of addressing their Public Records Requests," (R. 611, Briefing Order, Page ID# 14995), Defendants and the DEA argue that this issue is beyond the scope of this appeal. However, we have in past cases " 'reach[ed] the question' of the district court's seal 'on our own motion,' " without any party having raised the issue. Shane Grp. , 825 F.3d at 305 (quoting Brown & Williamson Tobacco Corp. v. F.T.C. , 710 F.2d 1165, 1176 (6th Cir. 1983) ). We therefore need not concern ourselves with whether this issue is within the scope of Intervenors' intervention; rather, the issue is within our authority to decide regardless of whether or not the district court conferred intervenor status upon HDM or the Washington Post to make arguments about the issue. See id.
Concerning nondisclosure in litigation, this Court has distinguished between secrecy in the context of discovery, which as discussed above is permissible with a showing of "good cause," and secrecy *939in the context of adjudication, which is generally impermissible due to the "strong presumption in favor of openness" of court records. Shane Grp. , 825 F.3d at 305 (quoting Brown & Williamson , 710 F.2d at 1179 ). We have stated that "[t]he line between these two stages, discovery and adjudicative, is crossed when the parties place material in the court record." Id. The presumption in favor of openness of court records is justified because "[t]he public has an interest in ascertaining what evidence and records the District Court and this Court have relied upon in reaching our decisions." Id. (quoting Brown & Williamson , 710 F.2d at 1181 ). This strong presumption in favor of openness is only overcome if a party "can show a compelling reason why certain documents or portions thereof should be sealed, [and] the seal itself [is] narrowly tailored to serve that reason." Id. Further, "the greater the public interest in the litigation's subject matter, the greater the showing necessary to overcome the presumption of access." Id.
In this case, the sealed or redacted pleadings, briefs, or other documents that the parties have filed with the court, as well as any reports or exhibits that accompanied those filings,14 are the sort of records that would help the public "assess for itself the merits of judicial decisions." Id. ; see id. at 304-05 (treating as court records entitled to the presumption of openness the following: pleadings, motions for class certification, evidentiary motions, and exhibits accompanying the parties' filings). These documents are therefore subject to the strong presumption in favor of openness, which applies here with extra strength given the paramount importance of the litigation's subject matter.
The district court abused its discretion in permitting Defendants and the DEA to file their pleadings under seal. "[A] district court that chooses to seal court records must set forth specific findings and conclusions 'which justify nondisclosure to the public,' " even if no party objects to their sealing. Shane Grp. , 825 F.3d at 306 (quoting Brown & Williamson , 710 F.2d at 1176 ). We have made clear that "a court's failure to set forth those reasons-as to why the interests in support of nondisclosure are compelling, why the interests supporting access are less so, and why the seal itself is no broader than necessary-is itself grounds to vacate" an order allowing court documents to be filed under seal or with redactions. Id. at 306. No such findings or conclusions were made in this case,15 and the district court ipso facto abused its discretion. Id.
We therefore vacate any district court orders to the extent they permit sealing or redacting of court records. We remand for the district court to reconsider each pleading filed under seal or with redactions and to make a specific determination as to the necessity of nondisclosure in *940each instance. The court is advised to bear in mind that the party seeking to file under seal must provide a "compelling reason" to do so and demonstrate that the seal is "narrowly tailored to serve that reason." Shane Grp. , 825 F.3d at 305. On remand, if the district court permits a pleading to be filed under seal or with redactions, it shall be incumbent upon the court to adequately explain "why the interests in support of nondisclosure are compelling, why the interests supporting access are less so, and why the seal itself is no broader than necessary." Id. at 306. In doing so, the district court is to pay special attention to this Court's statement that "[o]nly the most compelling reasons can justify non-disclosure of judicial records." Id. at 305 (internal quotation marks omitted). The district court's findings and conclusions must also be consistent with the proper balancing of interests with respect to the ARCOS data, as discussed in the previous section.
CONCLUSION
For the reasons stated above, we VACATE the district court's Protective Order and any orders permitting the parties to file pleadings under seal or with redactions, and REMAND to permit the district court to consider entering new orders consistent with this opinion.
CONCURRING IN PART AND DISSENTING IN PART
RALPH B. GUY, JR., Circuit Judge, concurring in part and dissenting in part.
I agree with the majority on the narrow matter of the sealed records. Although I would affirm all of the decisions of the district court that were directly appealed from, I agree that the district court has not established adequate, on-the-record reasons for permitting many other docket entries to be filed under seal. I thus agree with the decision to proceed on our own motion and direct the district court to reevaluate those decisions on remand. I dissent, however, from the balance of the opinion.
Prescription opiates are being abused on a large scale throughout the country. In the face of this problem, different solutions are emerging. The United States is attempting to curb the trend through investigations and prosecutions. Cities, counties, and tribes are using private lawsuits to remedy the damage done. And newspapers and other media outlets are attempting to educate the public on the matter. Those different approaches converged in this case and required the district court to balance the various interests at play.
Plaintiffs face a daunting task in their civil suit. At trial, they must ultimately prove which manufacturers, distributors, and pharmacies handled the pills that ended up in their specific jurisdictions. The task is daunting because drugs take a circuitous path from the manufacturer to the consumer and there are around 1,200 companies potentially involved. Fortunately for plaintiffs, the DEA has spent years compiling the ARCOS database, which gives plaintiffs precisely the information they need.
Also fortunately for plaintiffs, discovery is quite broad under the Federal Rules of Civil Procedure. See Lewis v. ACB Bus. Servs., Inc. , 135 F.3d 389, 402 (6th Cir. 1998). Rule 26(b) draws the general scope of discovery:
Unless otherwise limited by court order, the scope of discovery is as follows: Parties may obtain discovery regarding any nonprivileged matter that is relevant to any party's claim or defense and proportional to the needs of the case, considering the importance of the issues at stake in the action, the amount in controversy, *941the parties' relative access to relevant information, the parties' resources, the importance of the discovery in resolving the issues, and whether the burden or expense of the proposed discovery outweighs its likely benefit. Information within this scope of discovery need not be admissible in evidence to be discoverable.
Fed. R. Civ. P. 26(b)(1). Such a permissive standard allows "extensive intrusion into the affairs of both litigants and third parties." Seattle Times Co. v. Rhinehart , 467 U.S. 20, 30, 104 S.Ct. 2199, 81 L.Ed.2d 17 (1984). Given that reality, it is not surprising that some persons and entities would prefer not to turn over certain documents or would prefer to do so only with certain protections in place. This is why Rule 26 also allows courts to enter protective orders. Under Rule 26(c), a party seeking information is obligated to first try and reach a compromise with the person or entity from which it seeks the information. Failing a compromise, the "court may, for good cause, issue an order to protect a party or person from annoyance, embarrassment, oppression, or undue burden or expense[.]" Fed. R. Civ. P. 26(c)(1).
That is what happened here. The DEA is not a party to this lawsuit but was pulled in by plaintiffs' subpoena. The DEA owns the ARCOS data and it initially resisted giving plaintiffs the data altogether, thus necessitating the court's involvement. Footnote five of the majority's opinion suggests that a stipulated protective order resolved the dispute. I disagree with that characterization, as I explain below. But it is true that the court entered a protective order, thus leading the DEA to give plaintiffs the data they sought. With this treasure trove of information, plaintiffs could continue their preparations for trial and engage in more meaningful settlement talks with defendants.
Then another group entered the fray. Some newspapers realized that the ARCOS data could be helpful in their own quest to educate the public about the cause and extent of the opioid crisis. But the newspapers are not parties to this lawsuit, so they do not have Rule 26 on their side; they could not subpoena the DEA. They are, of course, free to file a federal Freedom of Information Act (FOIA) request, which the DEA recommended they do. Although the federal FOIA exempts certain types of records from disclosure, the Washington Post wrote in its briefing that it believes no FOIA exemption would be applicable. Yet for reasons of their own, the newspapers have not chosen to pursue that route for obtaining the records.
Instead, the newspapers took a roundabout way. Realizing that plaintiffs now possessed the information they sought, the newspapers filed the state-law equivalents of FOIA requests with three of the plaintiffs-one county in West Virginia and two counties in Ohio. The newspapers reasoned that the ARCOS data was now a "public record" of those counties and must be turned over upon request.
But the protective order stood in their way. Under the terms of the protective order, if a plaintiff governmental entity received a public record request for the ARCOS data, the plaintiff had to "immediately notify the DEA and Defendants of the request." (R. 167, PageID 942.) A defendant or the DEA could then "file an appropriate action in [the district court] opposing production of the records." (Id. at 943.) The plaintiff would then be forbidden to release the information absent an order from the district court. (Id. ) This process played out and ultimately the district court decided not to allow plaintiffs to pass along the ARCOS data to the newspapers.
The majority concludes that this decision was an abuse of discretion, but it *942begins with a false premise. According to the majority, the parties stipulated to a protective order. (Maj. Op. at 929-30.) The majority contends that although the parties submitted "rival orders" that differed in some respects, their orders were "identical with respect to every aspect of the Protective Order relevant to this appeal." (Id. at 924 n.5.) This conclusion skips over the relevant lead-up.
The entry of the protective order happened in stages. First, the district court directed the parties to "meet to see if they can reach agreement on what part of the [ARCOS] database the DEA will produce." (R. 112, PageID 686.) The district court affirmed that there "is a legitimate need for Plaintiffs to obtain this data," but "believe[d] that production must be tailored - perhaps through a protective order - in a way to address the DEA's concerns regarding breadth, years in question, potential interference in investigations and enforcement actions, and divulging the location of warehouses where opioids are stored." (Id. ) The parties met, but failed to reach an agreement, and a summary of their competing status reports reveals why:
- The DEA was willing to share with plaintiffs only two years of data rather than the nine years of data requested;
- The DEA insisted on providing the data in Excel spreadsheets, rather than allow plaintiffs to access it in its native format;
- The DEA refused to disclose the names of the manufacturers, distributors, and pharmacies but instead suggested replacing them with numeric identifiers;
- The DEA insisted on replacing the city or county with the first three digits of the relevant postal code;
- Plaintiffs were unwilling to agree to a protective order that prevented them from disclosing the ARCOS data to the media.
(See R. 137, PageID 741-43; R. 139.) Thus, one of the sticking points was the parties' disagreement over whether the media should have access to the ARCOS data. And that dispute kept them from reaching an agreement.
The parties' failure to agree required the court to hold a hearing which lasted one hour and 45 minutes. (R. 112, PageID 686; R. 156.) At the hearing, the court rejected some of the restrictions that the DEA had been demanding,1 but the court was clear that a protective order of some kind was going to be put into place and that the order would, at a minimum, limit the use of the information to "two purposes; litigation, law enforcement. That's it." (R. 156, PageID 861.) When counsel for defendants voiced concern about plaintiffs' insistence that they be allowed to disclose ARCOS data to the media, the district court interjected to flatly put that concern to rest. "Nothing is going to be revealed to the media unless there's a trial. If there's a trial, obviously trials in our country are public. Hopefully there will be no trials." (Id. ) Counsel responded, "We appreciate that and appreciate the opportunity to work on a protective order, as DEA has suggested."2 (Id. ) The court concluded the *943hearing by ordering the parties to jointly propose an order by the end of that week.
The DEA submitted a proposed order soon after but confessed the parties' inability to resolve two matters.3 One matter concerned the tiers of confidentiality while the other concerned the extent to which ARCOS data could be shared with law enforcement agencies. (R. 162, PageID 915-16.) A few days later, the court entered a protective order that adopted the plaintiff's approach to the tiers of confidentiality and adopted the DEA's language defining when law enforcement could be given the ARCOS data. (R. 167, PageID 938, 939.)
So while it is true that the parties agreed to the relevant terms of the protective order, they only did so after the court told them at the hearing what type of order it was willing to enter. In my view, the parties energetically fought over the terms of the protective order and never, in fact, fully agreed to all its terms. Rather, they argued their positions at the hearing, accepted the court's decisions on those positions, and then did their best to come up with a document that complied with the court's requirements and timeline. The good cause for the protective order can thus be found throughout the transcript of the hearing. There, the court explains why a protective order was necessary to protect investigations and trade secrets but also explains why the order need not be as restrictive as what the DEA and defendants were requesting. The court reemphasized this good-cause finding in its order sustaining the objections to releasing ARCOS data to the newspapers. (R. 800, PageID 18978.)
That leads to my second disagreement. According to the majority, the district court did "a complete about-face concerning the relevant interests at stake" when it declined to allow the newspapers to access the ARCOS data. (Maj. Op. at 932.) The majority concedes that "this about-face might be explained in part by the different interests at stake when disclosure is made only to parties to a case pursuant to a protective order, as compared to third parties that intend to publicly report on the disclosed information." (Id .) But the majority rejects this possibility outright because in its view, "it is readily apparent from the record that the district court's analysis in its first order did contemplate the public's interest in obtaining the ARCOS data" and the DEA's and defendants' interests in keeping it private. (Id. ) The majority is correct that the court was balancing different interests, but misdescribes how the court balanced them.
Again, the context matters. The majority selectively quotes docket entry 233, which is the court's order directing the DEA to give plaintiffs the raw ARCOS data. The DEA had already given plaintiffs aggregate data from ARCOS, but with a trial date now firmly set, the court needed to consider whether plaintiffs were entitled to the more specific data direct from ARCOS. (See R. 182; R. 232; R. 233, PageID 1110-11.) In the 233 order, the court considered and rejected each of the DEA's objections to giving plaintiffs the raw ARCOS data. (See generally R. 233.) In my view, the majority misdescribes this order.
Here is the order's full discussion on the DEA's concern about competitive harm to drug companies:
DEA "objects to the production of the requested information under DOJ's Touhy regulations ( *94428 C.F.R. § 16.26(b)(6) ) because disclosure would improperly reveal trade secrets without the owners' consent." Docket no. 101 at 6. Specifically, DEA asserts the data would reveal "details regarding the scope and breadth of [each manufacturer's and distributor's] market share, which is likely to cause [them] substantial competitive harm." Id. The Madel court explicitly rejected this argument, noting the assertion was conclusory and also that market data over three years old carried no risk of competitive harm. This objection is overruled.
(R. 233, PageID 1120.) The reference to Madel refers to Madel v. United States Dep't of Justice , 784 F.3d 448 (8th Cir. 2015) and the district court's decision on remand, Madel v. United States , No. 13-cv-2832, 2017 WL 111302 (D. Minn. Jan. 11, 2017). That case arose when a man named Madel submitted a FOIA request for ARCOS data and the DEA largely refused to provide it on the grounds that it would cause competitive harm. 784 F.3d at 452. Madel sued and the district court granted summary judgment to the DEA. The court of appeals partially affirmed because the DEA had provided affidavits that supported the claim that substantial competitive harm was likely. Id. at 453. The court of appeals remanded the case, however, so that the district court could determine whether portions of some of the ARCOS data were segregable from the portions likely to cause harm. Id. at 453-54.
In my reading of the 233 order, the district court was simply pointing out that the DEA could not avoid giving plaintiffs the relevant discovery materials based on a blanket assertion about competitive harm, particularly when the data was several years old. The court was not making a finding that the release of "market data over three years old" never carries the risk of competitive harm, regardless of who receives it. Rather, the risk was simply not so great that plaintiffs should be deprived of relevant discovery material to which they are entitled under Rule 26, absent a stronger showing. "Plaintiffs," of course, is the operative word because the court was simply ordering the DEA to give only plaintiffs the ARCOS data-and even then, still subject to the protective order. That is why it ended the order with this paragraph:
Having overruled DEA's objections, DEA shall produce the requested information for the States of Ohio, West Virginia, Illinois, Alabama, Michigan, and Florida. Use of the ARCOS database shall be limited to this litigation and for State and local law enforcement purposes only. No person shall disclose the data or allow use of the data except as necessary for a submission to the court or at trial. The ARCOS data produced pursuant to this Order shall be governed by the Protective Order previously entered at docket no. 167.
(R. 233, PageID1125.)
The same goes for the 233 order's discussion of ongoing criminal investigations. The DEA objected to providing plaintiffs with any raw ARCOS data because it "would interfere with enforcement proceedings" and other "ongoing matters." (Id. at 1119-20.) The court found it "untenable that exposure of the data will actually or meaningfully interfere with any ongoing enforcement proceeding" but invited the DEA to provide evidence that "specific data would interfere with a particular, pre-existing and ongoing enforcement proceeding," and to make any necessary showing about "non-segregability." (Id .) Here too, the court was speaking with the assumption that only plaintiffs (and some law enforcement agencies) would receive the ARCOS data and, even then, the data *945would still be subject to the protective order.
When properly viewed, I do not believe, as the majority does, that the court's subsequent decision to prevent the press from obtaining the ARCOS data was "bizarre." (Maj. Op. at 932-33.) The district court balanced different interests, under different circumstances, and at different times. The court's statements in the 233 order came with the caveat that the ARCOS data was not going to be made public unless and until the case went to trial and it became a public record. The court ordered the DEA to turn over the data because Rule 26 entitled plaintiffs to acquire it, even if it carried a risk of harm to the DEA or defendants. The newspapers are not parties and thus are not entitled to the material.4 The bar for keeping discovery materials out of their hands is thus considerably lower. In my view, the court did not abuse its discretion in striking those balances.
The proper balance may change over time. If we were to affirm the district court's decision, as I believe we should, the newspapers would still be free to intervene in the future-perhaps after trial-and ask the district court to modify the protective order. Given that the DEA did establish good cause at the time the protective order was entered, the burden would be on the newspapers to show good cause for modifying it. See Pansy v. Borough of Stroudsburg , 23 F.3d 772, 790 (3d Cir. 1994) (noting that the party seeking to modify a protective order "must come forward with a reason to modify the order"); see also Phillips v. Gen. Motors Corp. , 307 F.3d 1206, 1211 n.1 (9th Cir. 2002) (placing the burden on the party seeking the protection only because the protective order had been stipulated to and no party had made a "good cause" showing). The district court could then determine whether the DEA's concerns were still valid and compare the newspapers' need for the data with the parties' reliance interests. See Meyer Goldberg, Inc., of Lorain v. Fisher Foods, Inc ., 823 F.2d 159, 163 (6th Cir. 1987).
For now, the parties are attempting to negotiate a settlement while simultaneously preparing for a trial that is only a few months away. The district court is marshalling those efforts and did not abuse its discretion in keeping the protective order in place for the time being. The majority sees it otherwise, so I respectfully dissent.
Plaintiffs are not involved in this appeal.
Defendants are involved in this appeal as Appellees.
Unless otherwise stated, all citations to the record refer to Case No. 1:17-md-02804-DAP.
The district court noted that the ARCOS data "are not pure investigatory records compiled for law enforcement purposes, [but] simply business records of defendants; ... the database does not include any additional DEA analysis or work-product[.]" (R. 233, Page ID# 1119.)
The DEA and Defendants argue that this Protective Order was not stipulated because Plaintiffs and the DEA proposed rival protective orders; however, these rival orders were identical with respect to every aspect of the Protective Order relevant to this appeal. (See R. 167, Page ID# 937-38 (discussing the differences between the DEA's and Plaintiffs' proposed protective orders).) Because no party demonstrated "good cause" for these aspects of the Protective Order and no party challenged these aspects, we treat the Protective Order as one to which the parties stipulated.
The DEA initially filed its brief in support of objections with "heav[y] redact[ions]," and the Washington Post moved to access the unredacted brief. (R. 800, Page ID# 18972.) Before the district court ruled on this motion, the DEA filed an amended brief with fewer redactions. The district court ultimately dismissed the Washington Post's motion as moot, holding that the DEA's amended brief had "removed all but necessary redactions." (R. 800, Page ID# 18973.)
The district court even noted as relevant that "Plaintiffs assert that part of the reason for the opioid epidemic is lack of law enforcement." (R. 233, Page ID# 1119.)
Intervenors argue that this Court's line of cases emphasizing the "strong presumption in favor of openness in the courtroom" supports their position. Brown & Williamson Tobacco Corp. v. F.T.C. , 710 F.2d 1165, 1179 (6th Cir. 1983) ; see Signature Mgmt. Team, LLC v. Doe , 876 F.3d 831, 836 (6th Cir. 2017) ; In re Knoxville News-Sentinel Co. , 723 F.2d 470, 476 (6th Cir. 1983). However, the strong presumption of openness in the courtroom and for court records does not apply to the discovery process, which occurs before the parties get to the courtroom. See Shane Grp. , 825 F.3d at 305. These cases are thus inapplicable to this issue-except to the extent that they demonstrate a more generalized, but less intense, public interest in the disclosure of documents related to litigation. Nevertheless, while there may not be a strong presumption in favor of disclosure in the discovery context, the party seeking nondisclosure still must demonstrate "good cause" for a protective order "specifying terms ... for the disclosure or discovery." Fed. R. Civ. P. 26(c)(1).
Nor was this the first time that the district court had raised the risk of public disclosure of the ARCOS data notwithstanding the Protective Order. At the same hearing where the district court stated that any protective order it would enter would limit the use of the ARCOS data to "two purposes; litigation, law enforcement," it is clear that the court was also concerned with the potential for harm if the data leaked. (R. 156, Page ID# 861.) For example, before it was informed that the location of warehouses in which large quantities of drugs were stored was already publicly available, the district court was greatly concerned that this information would be part of the ARCOS data being disclosed to Plaintiffs. (R. 156, Page ID# 836-38.) The hearing transcript makes clear that the district court's concerns stemmed from the possibility that a criminal could steal drugs from these warehouses if he knew their locations. (See R. 156, Page ID# 836-38, 865, 888.) Obviously, if the Protective Order guaranteed that no one other than the parties would access the data, such concerns would be completely unfounded. The fact that the district court expressed concerns about the risk of public disclosure well before its order that the DEA disclose the ARCOS data to Plaintiffs provides strong evidence that these concerns were on the district court's mind when it considered that order as well.
The DEA asserts that the district court reconsidered its position on the law enforcement interests at issue after reading the declaration of DEA Assistant Administrator John J. Martin. (DEA Br. 41) (citing R. 717-1, Martin Decl. Page ID# 16519.) There is no record evidence for the DEA's assertion, as the district court did not cite or refer to the Martin declaration in its Opinion and Order. Further, it would be surprising if this was the case, since the declaration asserted substantially the same points that the DEA had made in all of its previous briefing-points which the district court had rejected.
We quote this portion of the declaration even though it was redacted in the unsealed court filing because, for reasons discussed in the following section, it was error for the district court to allow portions like this to be filed under seal. In brief, there is a "strong presumption in favor of openness" of court records, which include court filings (like the DEA Brief in Support of Objections to Disclosure of ARCOS Data) and exhibits thereto (like the Martin declaration). See Shane Grp. , 825 F.3d at 305. "Only the most compelling reasons can justify nondisclosure of judicial records." Id. at 305 (quoting Knoxville News-Sentinel Co. , 723 F.2d at 476 ). The quoted sentence from Martin's declaration contains only very general information about an ongoing administrative action into an unidentified distributor. We reject the notion that compelling reasons justified redacting this sentence and therefore quote it without redaction. See id. at 303-04 (quoting from a sealed report in an opinion vacating the district court's orders to seal court records).
The DEA also argues that allowing Intervenors to obtain the ARCOS data would be to allow them to get around the requirements of the Freedom of Information Act. However, for the reasons stated above, we do not believe that disclosing the ARCOS data, particularly with the option of partial redaction, "could reasonably be expected to interfere with enforcement proceedings," which means that the data would be available under FOIA. 5 U.S.C. § 552(b)(7).
Instead, Defendants argue that the district court was correct in holding that "the ARCOS data is not a record generated by the Counties that are, or may be, subject to state public records requests." (R. 800, Page ID# 18981.) It is not clear why the district court found this relevant to its inquiry; there is no reason for this Court or any other federal court (rather than the courts of Ohio and West Virginia) to decide the scope of those state laws. Further, the text of both state statutes strongly suggests that the data would be subject to the public records request. See Ohio Rev. Code § 149.43(A)(1) (" 'Public record' means records kept by any public office, including, but not limited to, state, county, city, village, township, and school district units[.]"); W. Va. Code § 29B-1-2(5) (" 'Public record' includes any writing containing information prepared or received by a public body, the content or context of which, judged either by content or context, relates to the conduct of the public's business."). The fact that the Attorney General of West Virginia previously provided ARCOS data to HDM is further evidence that the West Virginia public records law covers the data.
These documents include, but are not limited to, the DEA's Amended Brief in Support of Objections (R. 717), John J. Martin's Declaration (R. 663-1), and any pleadings filed under seal or with redactions. On remand, the district court shall conduct a full review of court documents filed under seal or with redactions, and it shall in each instance reevaluate whether redaction or seal is necessary in light of the proper legal standards as set forth in this opinion.
The DEA argues that the district court's statement, in a footnote, that the DEA's Amended Brief in Support of Objections had "removed all but necessary redactions" was sufficient analysis. (DEA Br. 64 (quoting R. 800, Page ID# 18973).) This statement does not explain "why the interests in support of nondisclosure are compelling, why the interests supporting access are less so, and why the seal itself is no broader than necessary." Shane Grp. , 825 F.3d at 306. It is therefore insufficient to justify the redactions. Id.
For instance, the DEA opposed disclosing where warehouses were located even though, as counsel for plaintiffs pointed out, the information was already publicly available. (See R. 156, PageID 838-41.)
Contrary to the majority's suggestion, neither this exchange nor the others cited, when read in context, can be fairly described as "us[ing] the threat of publicly disclosing the data as a bargaining chip in settlement discussions." (Maj. Op. at 933.)
The DEA's motion and the court's subsequent protective order suggest that plaintiffs independently submitted their own proposed order, but that proposed order is not part of the record. (R. 162-1, PageID 916; R. 167, PageID 938.)
The newspapers attempted to get the ARCOS data directly from plaintiffs shortly after they received it from the DEA. At that point, the data was "the raw fruit[ ] of discovery" not yet in the possession of the court. See Wilk v. Am. Med. Ass'n , 635 F.2d 1295, 1299 n.7 (7th Cir. 1980) ; see also Bond v. Utreras , 585 F.3d 1061, 1066 (7th Cir. 2009) ; Rushford v. New Yorker Magazine, Inc. , 846 F.2d 249, 252 (4th Cir. 1988).