FOR PUBLICATION

UNITED STATES COURT OF APPEALS

FOR THE NINTH CIRCUIT

TWITTER, INC., No. 20-16174 Plaintiff-Appellant, D.C. No. 4:14-cv- v. 04480-YGR

MERRICK B. GARLAND, Attorney General; UNITED STATES OPINION

DEPARTMENT OF JUSTICE;

CHRISTOPHER WRAY, Director of the Federal Bureau of Investigation;

FEDERAL BUREAU OF

INVESTIGATION,

Defendants-Appellees.

Appeal from the United States District Court for the Northern District of California Yvonne Gonzalez Rogers, District Judge, Presiding

Argued and Submitted August 10, 2021 Seattle, Washington

Filed March 6, 2023

Before: Carlos T. Bea, Daniel A. Bress, and Lawrence VanDyke, Circuit Judges. 2 TWITTER, INC. V. GARLAND

Opinion by Judge Bress; Concurrence by Judge VanDyke

SUMMARY *

Civil Rights The panel affirmed the district court’s summary judgment for the United States in an action brought by Twitter alleging First Amendment violations arising from the FBI’s restrictions on Twitter’s publication of a self- described “Transparency Report.” In support of its classified national security investigations, the United States served administrative subpoenas and orders requiring Twitter to provide the government with certain information about Twitter users. In its Transparency Report, Twitter wished publicly to disclose certain information about the aggregate numbers of these governmental requests that it received between July and December 2013. The FBI determined that the number of subpoenas and orders and related information was classified, and that Twitter’s disclosure of this information would harm national security. As a result, the FBI allowed Twitter to release its Transparency Report only in a partially redacted form. The panel held that Twitter’s constitutional challenges failed to persuade. Under this circuit’s case law, strict scrutiny applied because the restriction on Twitter’s speech

* This summary constitutes no part of the opinion of the court. It has been prepared by court staff for the convenience of the reader.

TWITTER, INC. V. GARLAND 3

was content based. The panel acknowledged that Twitter has a First Amendment interest in commenting on matters of public concern involving national security subpoenas. Nevertheless, based on a careful review of classified and unclassified information, the panel held that the government’s redactions of Twitter’s Transparency Report were narrowly tailored in support of the compelling government interest in national security. Against the backdrop of explicit illustrations set forth in the classified materials of the threats that exist and the ways in which the government can best protect its intelligence resources, the panel was able to appreciate why Twitter’s proposed disclosure would risk making foreign adversaries aware of what is being surveilled and what is not being surveilled—if anything at all. Given these concerns and this fuller backdrop, the panel was willing to accept the main conclusions outlined in the unclassified materials, which expressed generally why revealing the information Twitter wished to disclose would significantly harm the government’s national security operations by signaling to adversaries what communication channels to avoid and which to use. The panel concluded that the government’s redactions of Twitter’s Transparency Report did not violate the First Amendment. The panel next held that the statutory scheme governing the permissible disclosure of aggregate data about the receipt of national security legal process allowed for sufficient procedural protections, which Twitter received here. The panel held that the specific procedural requirements of Freedman v. Maryland, 380 U.S. 51 (1965), which were designed to curb traditional censorship regimes, were not required in the context of government restrictions on the disclosure of information transmitted confidentially as part 4 TWITTER, INC. V. GARLAND

of a legitimate government process because such restrictions do not pose the same dangers to speech rights as do traditional censorship regimes. Even though Freedman’s specific procedural framework did not apply, Twitter received considerable process—including some of the process that Freedman envisioned. Finally, the panel held that due process did not require that Twitter’s outside counsel receive classified information by virtue of Twitter filing this lawsuit. Twitter was provided with unclassified versions of the various declarations, which the panel relied upon throughout its opinion. The unclassified declarations provided Twitter with sufficient information by which to advance Twitter’s interests before this Court. And although the panel appreciated Twitter’s concern that it could not respond to what it did not know, Twitter’s interest in the classified information did not rise to the level of constitutional imperative. Concurring in the judgment, Judge VanDyke agreed with the majority’s conclusion in this case, and most aspects of its analysis, with the only significant disagreement being whether the panel needed to rely on classified materials to resolve this case. Judge VanDyke concluded that the unclassified materials were sufficient to meet the government’s burden. Given the “significant weight” a court must afford to the Government’s national security factual findings, Judge VanDyke would hold that the Government’s unclassified declarations—specifically, the unclassified declaration from Jay S. Tabb, Jr., the new Executive Assistant Director of the FBI’s National Security Branch—sufficiently demonstrated that the Government’s restrictions on Twitter’s speech were narrowly tailored to the compelling interest of protecting national security and safeguarding classified information.

TWITTER, INC. V. GARLAND 5

COUNSEL

Lee H. Rubin (argued), Donald M. Falk, and Samantha C. Booth, Mayer Brown LLP, Palo Alto, California; Andrew John Pincus, Mayer Brown LLP, Washington, D.C.; Samantha A. Machock, Wilson Sonsini Goodrich Rosati, San Diego, California; for Plaintiff-Appellant. Lewis S. Yelin (argued) and Scott R. McIntosh, Appellate Staff Attorneys, Civil Division; David L. Anderson, United States Attorney; Jeffrey Bossert Clark, Acting Assistant Attorney General; United States Department of Justice; Washington, D.C.; Stefania M. Porcelli, Assistant General Counsel; Cecilia O. Bessee, Acting Deputy General Counsel; Jason A. Jones, General Counsel; Federal Bureau of Investigation; Washington, D.C.; for Defendants- Appellees. Andrew G. Crocker and Naomi Gilens, Electronic Frontier Foundation, San Francisco, California; Ashley Gorski, Brett Max Kaufman, and Patrick Toomey, American Civil Liberties Union Foundation, New York, New York; Matthew T. Cagle, ACLU Foundation of Northern California, San Francisco, California; Peter J. Eliasberg, ACLU Foundation of Southern California, Los Angeles, California; David Loy, ACLU Foundation of San Diego and Imperial Counties, San Diego, California; for Amici Curiae Electronic Frontier Foundation, American Civil Liberties Union, American Civil Liberties Union of Northern California, American Civil Liberties Union of Southern California, and American Civil Liberties Union of San Diego and Imperial Counties. 6 TWITTER, INC. V. GARLAND

OPINION

BRESS, Circuit Judge:

In support of its classified national security investigations, the United States served administrative subpoenas and orders requiring Twitter to provide the government with certain information about Twitter users. In a self-described “Transparency Report,” Twitter wishes publicly to disclose certain information about the aggregate numbers of these governmental requests that it received between July and December 2013. The FBI determined that the number of subpoenas and orders and related information was classified, and that Twitter’s disclosure of this information would harm national security. As a result, the FBI allowed Twitter to release its Transparency Report only in a partially redacted form. This dispute over what Twitter can and cannot disclose about information it learned as a recipient of national security legal process raises several important questions that lie at the intersection of national security and the freedom of speech: Does the government’s content-based limitation on Twitter’s speech violate the First Amendment? Do the relevant national security statutes provide sufficient procedural protections to Twitter, consistent with the First Amendment? And does due process require that Twitter’s outside counsel be granted access to the classified materials on which the United States relies in objecting to Twitter’s proposed disclosure? We hold that Twitter’s constitutional challenges fail to persuade. Although we acknowledge Twitter’s desire to speak on matters of public concern, after a thorough review of the classified and unclassified record, we conclude that

TWITTER, INC. V. GARLAND 7

the government’s restriction on Twitter’s speech is narrowly tailored in support of a compelling government interest: our Nation’s security. We further hold that the statutory scheme governing the permissible disclosure of aggregate data about the receipt of national security legal process allows for sufficient procedural protections, which Twitter received here. Due process likewise does not require that Twitter’s outside counsel receive classified information by virtue of Twitter filing this lawsuit. Although the interests on both sides of this case are weighty, under law the government prevails. We affirm the district court’s grant of summary judgment to the United States. I A It is widely recognized that electronic communications are used by persons who seek to harm the United States through terrorist activities or other misdeeds. See, e.g., Clapper v. Amnesty Int’l USA, 568 U.S. 398, 402–06 (2013). To that end, federal law gives the United States the authority to obtain information from electronic communication service providers in support of national security investigations. Two such means of obtaining information are relevant here. First, under 18 U.S.C. § 2709, the FBI is empowered to issue certain requests to any “wire or electronic communication service provider,” such as Twitter. See id. §§ 2510(15), 2711(1) (defining “electronic communication service”). These requests are known as “national security 8 TWITTER, INC. V. GARLAND

letters,” or NSLs. 50 U.S.C. § 1873(e)(3)(A). 1 An NSL directs its recipient to provide the FBI with “subscriber information and toll billing records information, or electronic communication transactional records in [the recipient’s] custody or possession.” 18 U.S.C. § 2709(a). Such information must be “relevant to an authorized investigation to protect against international terrorism or clandestine intelligence activities.” Id. § 2709(b)(1); see also John Doe, Inc. v. Mukasey, 549 F.3d 861, 876 (2d Cir. 2008). NSLs thus allow the government to collect the aforementioned metadata, but not the actual content of electronic communications. See 18 U.S.C. § 2709; David Kris & J. Douglas Wilson, National Security Investigations and Prosecutions § 20:6 (Westlaw, Sept. 2021 Update). To ensure needed secrecy, the FBI may prohibit an NSL recipient from disclosing that it has received an NSL if a sufficiently high-ranking FBI official certifies that the absence of a prohibition on disclosure may result in any one of four enumerated harms. 18 U.S.C. § 2709(c)(1)(A). Those harms consist of: (1) “a danger to the national security of the United States,” (2) “interference with a criminal, counterterrorism, or counterintelligence investigation,” (3) “interference with diplomatic relations,” and (4) “danger to the life or physical safety of any person.” Id. § 2709(c)(1)(B). When such a certification has been made, the NSL recipient may not “disclose to any person that the Federal Bureau of Investigation has sought or obtained access to information or records under this section.” Id. § 2709(c)(1)(A). The prohibition on disclosure is subject to judicial review under 18 U.S.C. § 3511. See id. § 2709(d). We will refer to this prohibition as the “NSL nondisclosure

1 We note that Title 50 has not been enacted as positive law.

TWITTER, INC. V. GARLAND 9

requirement” or the “individual NSL nondisclosure requirement.” Second, the FBI can seek surveillance-related orders under the Foreign Intelligence Surveillance Act (FISA). 50 U.S.C. §§ 1801–1885c. Such orders, commonly known as “FISA orders,” are issued under one of five “titles” of FISA: Title I authorizes electronic surveillance, id. §§ 1801–1813; Title III authorizes physical searches, id. §§ 1821–1829; Title IV authorizes the use of “pen registers” and “trap and trace devices,” id. §§ 1841–46 2; Title V authorizes the compelled production of “tangible things,” such as business records, id. §§ 1861–64; and Title VII authorizes acquisition of foreign intelligence through the targeting of non-U.S. persons located outside the United States, id. §§ 1881– 1881g. While NSLs provide the government with only non- content data, FISA orders may compel the production of either content or non-content data. See Kris & Wilson, supra, § 13:5. With some exceptions, FISA orders relating to domestic surveillance ordinarily require authorization from the Foreign Intelligence Surveillance Court (FISC). Compare 50 U.S.C. §§ 1805(a), 1824(a), 1842(a)–(b), 1862(a)–(b), with id. §§ 1802(a), 1822(a). FISA orders relating to persons reasonably believed to be abroad may be authorized by directives issued by the Attorney General or the Director of National Intelligence. See id. §§ 1881a–1881c. See Kris & Wilson, supra, § 17:17.

2 Pen registers and trap and trace devices, respectively, capture the phone number associated with an outgoing or incoming call (or other communication) on a given communication line. 50 U.S.C. § 1841(2); see also 18 U.S.C. § 3127(3)–(4). 10 TWITTER, INC. V. GARLAND

Recipients of FISA orders generally are required to “protect [the] secrecy” of the government surveillance. 50 U.S.C. § 1805(c)(2)(B) (Title I); see also id. §§ 1824(c)(2)(B) (Title III), 1842(d)(2)(B)(i) (Title IV), 1862(d)(2) (Title V), 1881a(i)(1)(A) (Title VII). Recipients of certain types of FISA orders must also “maintain under security procedures approved by the Attorney General and the Director of Central Intelligence any records concerning the surveillance or the aid furnished that such person wishes to retain.” Id. § 1805(c)(2)(C) (Title I); see also id. §§ 1824(c)(2)(C) (Title III), 1842(d)(2)(B)(ii)(II) (Title IV), 1881a(i)(1)(B) (Title VII). A FISA order recipient may obtain judicial review of a nondisclosure obligation in the FISC. See, e.g., id. § 1881a(i)(4). Further review may be sought in the Foreign Intelligence Surveillance Court of Review. 50 U.S.C. § 1881a(i)(6)(A). B The government closely guards information relating to NSLs and FISA orders. The President has the “authority to classify and control access to information bearing on national security.” Dep’t of Navy v. Egan, 484 U.S. 518, 527 (1988). In the exercise of that authority, the President has “prescribe[d] a uniform system for classifying [and] safeguarding” national security information. Classified National Security Information, Exec. Order No. 13,526 pmbl., 75 Fed. Reg. 707, 707 (Dec. 29, 2009), as corrected by 75 Fed. Reg. 1,013 (Dec. 29, 2009). Information that is classified falls into one of three levels: “Confidential,” “Secret,” and “Top Secret.” Id. § 1.2(a), 75 Fed. Reg. at 707–08. Of these levels, “Top Secret” is the highest, reserved for “information, the unauthorized disclosure of which reasonably could be

TWITTER, INC. V. GARLAND 11

expected to cause exceptionally grave damage to the national security.” Id. § 1.2(a)(1), 75 Fed. Reg. at 707. Certain classified information is further designated “sensitive compartmented information,” or “SCI.” See Office of the Director of National Intelligence, Intelligence Community Directive 703, Protection of Classified National Intelligence, Including Sensitive Compartmented Information (2013) (“ICD 703”). This information “require[s] special controls and handling within the United States Intelligence Community.” Doe v. Cheney, 885 F.2d 898, 902 n.2 (D.C. Cir. 1989); see also ICD 703. Individuals with a security clearance must be granted additional permission to be allowed access to information designated SCI. See, e.g., ICD 703; Romero v. Dep’t of Def., 658 F.3d 1372, 1373–74 (Fed. Circ. 2011). Access to classified information is further restricted to individuals meeting criteria that the President has identified: “A person may have access to classified information provided that: (1) a favorable determination of eligibility for access has been made by an agency head or the agency head’s designee; (2) the person has signed an approved nondisclosure agreement; and (3) the person has a need-to- know the information.” Exec. Order No. 13,526 § 4.1(a), 75 Fed. Reg. at 720. “No information may remain classified indefinitely.” Id. § 1.5(d), 75 Fed. Reg. at 709. The default period during which classified information remains classified is ten years. Id. § 1.5(b), 75 Fed. Reg. at 709. However, the classifying official may specify an earlier date (or the occurrence of a certain event) upon which the information is automatically declassified, or he may extend the duration of classification to up to 25 years where necessary. Id. § 1.5(a), (c), 75 Fed. Reg. at 709. Agencies must also undertake periodic 12 TWITTER, INC. V. GARLAND

reassessments of classified designations upon request. Id. § 3.5(a), (d), 75 Fed. Reg. at 717–18. Unauthorized disclosure of classified materials is subject to punishment. See, e.g., 18 U.S.C. § 798(a)(3). Under this classification system, and prior to 2014, all information about the aggregate number and types of national security legal process received by any recipient was considered classified and therefore barred from public disclosure. But following Edward Snowden’s unauthorized disclosure of classified documents in 2013, and in response to requests from electronic service providers, the government made a change in policy to achieve greater transparency. In early 2014, then-Director of National Intelligence James Clapper issued the “DNI Memorandum,” which declassified “certain data related to requests by the United States to communication providers for customer information” made through FISA orders and NSLs. That same day, then-Deputy Attorney General James M. Cole issued a letter (“the DAG Letter”) addressed to Facebook, Google, LinkedIn, Microsoft, and Yahoo!, which permitted the same disclosures as the DNI Memorandum. The following year, Congress enacted the USA FREEDOM Act of 2015, Pub. L. No. 114-23, 129 Stat. 268 (found in scattered sections of Titles 18 and 50 U.S. Code), which enacted into law, and expanded, the categories of information that the DNI Memorandum and DAG Letter allowed to be disclosed. Id. §§ 603–604, 129 Stat. 295–97. The relevant provision, now codified at 50 U.S.C. § 1874, allows any “person subject to a nondisclosure requirement accompanying” a FISA order or an NSL publicly to disclose certain limited information about his receipt of national

TWITTER, INC. V. GARLAND 13

security process using one of four enumerated pathways. Id. § 1874(a). A recipient may choose to release information under one of the following four options:

(1) a semiannual report on the number of NSLs, FISA content orders and FISA non- content orders in bands of 1000, with some breakdowns by authority for non-content information; (2) a semiannual report on the number of NSLs, FISA content orders and FISA non-content orders in bands of 500; (3) a semiannual report on the total national security process received in bands of 250; or (4) an annual report on the total national security process received in bands of 100.

H.R. Rep. No. 114-109, pt. 1, at 27 (2015) (summarizing the statutory provisions); see also 50 U.S.C. § 1874(a); Kris & Wilson, supra, § 13:5. These bands are notable not only for their breadth—the tightest band is 100—but for the fact that each of the bands begins with, and includes, zero. For instance, an entity reporting under option two that received one FISA content order and three FISA non-content orders, but no NSLs, would indicate that it received between 0 and 499 FISA content orders, between 0 and 499 FISA non-content orders, and between 0 and 499 NSLs. Under the statute, such an entity could not indicate that it received no NSLs at all. We will refer to this as the “aggregate nondisclosure requirement,” so as to distinguish this system from the NSL nondisclosure requirement in 18 U.S.C. § 2709(c), which pertains to disclosing the receipt of individual NSLs. 14 TWITTER, INC. V. GARLAND

C On April 1, 2014, Twitter transmitted to the FBI a two- page draft document that it referred to as a “Transparency Report.” The Report was entitled “Empowering users with more #transparency on national security surveillance.” In its Transparency Report, and as described in its operative complaint in this lawsuit, Twitter sought to publish the following information regarding the NSLs and FISA orders that it had received from July 1, 2013 to December 31, 2013:

a. The number of NSLs and FISA orders Twitter received, if any, in actual aggregate numbers (including “zero,” to the extent that that number was applicable to an aggregate number of NSLs or FISA orders, or to specific kinds of FISA orders that Twitter may have received); b. The number of NSLs and FISA orders received, if any, reported separately, in ranges of one hundred, beginning with 1– 99; c. The combined number of NSLs and FISA orders received, if any, in ranges of twenty-five, beginning with 1–24; d. A comparison of Twitter’s proposed (i.e., smaller) ranges with those authorized by the DAG Letter; e. A comparison of the aggregate numbers of NSLs and FISA orders received, if any, by Twitter and the five providers to

TWITTER, INC. V. GARLAND 15

whom the DAG Letter was addressed [Facebook, Google, LinkedIn, Microsoft, and Yahoo!]; and f. A descriptive statement about Twitter’s exposure to national security surveillance, if any, to express the overall degree of government surveillance it is or may be subject to.

In a letter to the FBI accompanying the Report, Twitter requested “a determination as to exactly which, if any, parts of its Transparency Report are classified or, in the [Department of Justice’s] view, otherwise may not lawfully be published online.” The Transparency Report sought to educate Twitter users about the extent of the federal government’s surveillance requests of Twitter and the degree to which Twitter’s platform was safe from secret governmental prying. The draft Transparency Report further expressed Twitter’s “inten[t] to make this kind of report on a regular basis.” 3 In a September 9, 2014 response to Twitter, the FBI set forth its “conclu[sion] that information contained in the report is classified and cannot be publicly released.” The letter indicated that the Transparency Report was “inconsistent with the [DAG Letter] framework and discloses properly classified information.” In particular, the FBI explained that information in the Report “would reveal classified details about the surveillance . . . that go beyond what the government has permitted other companies to

3 Nothing in our opinion quotes or discloses materials that have been deemed classified. Any quotations of Twitter’s draft Transparency Report are to unredacted portions only. 16 TWITTER, INC. V. GARLAND

report.” The FBI specifically objected that the Transparency Report “would disclose specific numbers of orders received, including characterizing the numbers in fractions or percentages, and would break out particular types of process received.” This information, the FBI explained, was classified, and its release would harm national security. But the FBI nonetheless “believe[d] there [was] significant room for Twitter to place the numbers in context” by informing its users that, for instance, “only an infinitesimally small percentage of its total number of active users was affected” by government surveillance requests. Twitter would thus be “permitted to qualify its description of the total number of accounts affected by all national security legal process it has received but it cannot quantify that description with the specific detail” that Twitter desired. D After receiving the FBI’s letter, Twitter filed this lawsuit in October 2014, challenging the government’s suppression of the full Transparency Report and seeking injunctive and declaratory relief. Subsequently, after the USA FREEDOM Act was passed in June 2015, Twitter filed the operative Second Amended Complaint (SAC). In its SAC, Twitter claimed that the redacted information in the Report “was improperly classified” and that the government’s prohibition on Twitter’s publishing that information violated the First Amendment. Twitter also sought “to disclose that it received ‘zero’ FISA orders, or ‘zero’ of a specific kind of FISA order, for [the] period [covered by the Report], if either of those circumstances is true.” On November 22, 2016, the government moved for summary judgment. The government’s motion relied on the unclassified and classified declarations of Michael B.

TWITTER, INC. V. GARLAND 17

Steinbach, the then-Executive Assistant Director (EAD) of the FBI’s National Security Branch. The EAD is responsible for “overseeing the national security operations of the FBI’s Counterintelligence Division, Counterterrorism Division, High-Value Detainee Interrogation Group, Terrorist Screening Center, and Weapons of Mass Destruction Directorate.” The EAD has “also been delegated original classification authority by the Director of the FBI.” The classified Steinbach declaration, and all future classified declarations that the government would submit, were filed ex parte with the district court for the court’s in camera review. The district court denied the government’s motion for summary judgment “without prejudice to a renewed motion upon a more fulsome record.” The district court held that strict scrutiny applied to the government’s attempt to restrict the Transparency Report’s full publication and that the Steinbach declarations were insufficient to show that the government’s required redactions were narrowly tailored in support of the government’s compelling interest in national security. Twitter had also argued that under Freedman v. Maryland, 380 U.S. 51 (1965), neither the government’s classification decision nor the USA FREEDOM Act contained sufficient procedural safeguards to ensure the protection of Twitter’s First Amendment rights. For applicable prior restraints, Freedman requires that “(1) any restraint prior to judicial review can be imposed only for a specified brief period during which the status quo must be maintained,” “(2) expeditious judicial review of that decision must be available,” and “(3) the censor must bear the burden of going to court to suppress the speech and must bear the burden of proof once in court.” Thomas v. Chicago 18 TWITTER, INC. V. GARLAND

Park Dist., 534 U.S. 316, 321 (2002) (quoting FW/PBS, Inc. v. Dallas, 493 U.S. 215, 227 (1990) (principal opinion of O’Connor, J.)). In denying summary judgment without prejudice, the district court suggested that the government’s classification decision and the governing statutory scheme violated Freedman’s commands. In that same order, the district court also directed the government to proceed with granting two of Twitter’s outside lawyers, including Lee H. Rubin, “security clearances that would permit review of relevant classified materials in this matter.” Although Rubin’s background investigation was completed and favorably adjudicated, the government refused to provide Rubin access to the classified materials on the ground that Twitter’s outside counsel lacked a need to know this information. In support of its refusal to allow Rubin access to classified materials, the government submitted an unclassified declaration from Carl Ghattas, then-EAD of the FBI’s National Security Branch. Mr. Ghattas indicated that the classified Steinbach declaration and its exhibits were “classified at the TOP SECRET level and contain Sensitive Compartmented Information.” Mr. Ghattas then explained that Twitter’s counsel “do[es] not have a need for access to or a need-to-know the classified FBI information at issue in this case.” Specifically, Mr. Ghattas concluded, “it does not serve a governmental function . . . to allow plaintiff’s counsel access to the classified FBI information at issue in this case to assist in representing the interests of a private plaintiff who has filed this civil suit against the government.” On December 5, 2018, Twitter filed a request that Rubin be given access to the classified Steinbach declaration. In response, the government filed a declaration by then- Attorney General William P. Barr asserting the state secrets

TWITTER, INC. V. GARLAND 19

privilege over the information contained in the classified Steinbach declaration. Attorney General Barr’s declaration relied on a pair of unclassified and classified declarations by Michael C. McGarrity, then-Acting EAD of the FBI’s National Security Branch. The government submitted the McGarrity declarations for the district court’s review. On June 21, 2019, and in response to the new declarations, the district court issued an order to show cause why it should not reconsider its denial of summary judgment. The district court indicated that the classified McGarrity declaration “provides an explanation of the Government’s basis for restricting the information that can be published in the Draft Transparency Report, and the grave and imminent harm that could reasonably be expected to arise from its disclosure, in far greater detail than the Government provided previously.” The court thus indicated its likely view that the government’s restrictions on Twitter’s speech were narrowly tailored and that Rubin should not receive the classified materials because of “national security concerns.” The parties then filed cross-motions for summary judgment. In support of its motion, the government relied on newly submitted classified and unclassified declarations from Jay S. Tabb, Jr., the new EAD of the FBI’s National Security Branch. This time, the district court granted the government’s motion for summary judgment. The district court did not revise its earlier conclusion that “the restrictions on Twitter’s speech are subject to strict scrutiny as a content-based restriction and a prior restraint.” But it found that based on “the totality of the evidence provided in this case,” including all three of the classified declarations from EADs Steinbach, McGarrity, and Tabb, 20 TWITTER, INC. V. GARLAND

that the government had satisfied strict scrutiny. Citing “the specific reasons identified in the classified declarations,” the district court found that those declarations “explain the gravity of the risks inherent in disclosure of the information” at issue by providing “a sufficiently specific explanation of the reasons disclosure of mere aggregate numbers, even years after the relevant time period in the Draft Transparency Report, could be expected to give rise to grave or imminent harm to the national security.” The district court determined that the government’s supporting declarations sufficiently justified its classification decision, and that “no more narrow tailoring of the restrictions can be made.” The district court also denied Twitter relief under the procedural requirements of Freedman, but only on the basis that “Twitter’s SAC d[id] not allege a challenge, facial or otherwise, based upon the principles in Freedman.” The district court reasoned that “nothing in the SAC challenges a ‘system of prior restraints’ as in Freedman.” Accordingly, the court did not “reach the question of whether the Government’s decision here satisfied those procedural safeguards.” In a footnote, however, the district court noted that “[t]he sort of pre-disclosure review and approval process that restricts speech about metadata compiled by a recipient closely resembles the censorship systems raised in Freedman and its progeny.” The district court further opined that the government had “offered no applicable procedural protections similar to those cited with approval in” In re National Security Letter, 863 F.3d 1110, 1128 (9th Cir. 2017), and John Doe, Inc. v. Mukasey, 549 F.3d 861, 875 (2d Cir. 2008), two cases that we discuss further below. Finally, the district court denied Twitter’s request that its counsel receive access to the classified Tabb declaration. That declaration could not “be disclosed to counsel for Twitter

TWITTER, INC. V. GARLAND 21

based upon the national security concerns it raises, despite counsel’s clearance approval.” Twitter timely appealed. We have jurisdiction under 28 U.S.C. § 1291 and review the grant of summary judgment de novo. Butcher v. Knudsen, 38 F.4th 1163, 1168 (9th Cir. 2022). On appeal, we were provided with classified information, which was made available for our review using specialized procedures that ensured its protection. See generally Robert Reagan, Keeping Government Secrets: A Pocket Guide on the State-Secrets Privilege, the Classified Information Procedures Act, and Classified Information Security Officers (Federal Judicial Center, 2d ed. 2013) [hereinafter Reagan, Keeping Government Secrets]. II We turn first to the question of whether the government’s restriction on Twitter’s speech violates the First Amendment. We hold that under our case law, strict scrutiny applies to that inquiry. We acknowledge that Twitter has a First Amendment interest in commenting on matters of public concern involving national security subpoenas. Nevertheless, based on our careful review of classified and unclassified information, we hold that the government’s redactions of Twitter’s Transparency Report were narrowly tailored in support of the compelling government interest in national security. A The First Amendment provides that “Congress shall make no law . . . abridging the freedom of speech.” U.S. Const. amend. I. “When enforcing this prohibition, [courts] distinguish between content-based and content-neutral regulations of speech.” Nat’l Inst. of Family & Life 22 TWITTER, INC. V. GARLAND

Advocates v. Becerra, 138 S. Ct. 2361, 2371 (2018). Content-based restrictions “are presumptively unconstitutional and may be justified only if the government proves that they are narrowly tailored to serve compelling state interests.” Reed v. Town of Gilbert, 576 U.S. 155, 163 (2015). “A regulation of speech is facially content based under the First Amendment if it targets speech based on its communicative content—that is, if it applies to particular speech because of the topic discussed or the idea or message expressed.” City of Austin v. Reagan Nat’l Advert. of Austin, LLC, 142 S. Ct. 1464, 1471 (2022) (quotations and alteration omitted). “Regulations draw such a distinction if they ‘target speech based on its communicative content,’ prohibit ‘public discussion of an entire topic,’ or ‘single[] out specific subject matter for differential treatment.’” In re Nat’l Sec. Letter (“NSL”), 33 F.4th 1058, 1072 (9th Cir. 2022) (amended opinion) (alteration in original) (quoting Reed, 576 U.S. at 163, 169). In NSL, we considered a First Amendment challenge to 18 U.S.C. § 2709(c), which, as we discussed above, generally prohibits the recipient of a national security letter from disclosing the fact of its receipt. Id. at 1063. NSL recognized that § 2709(c) “prohibits speech about one specific issue: the recipient may not ‘disclose to any person that the Federal Bureau of Investigation has sought or obtained access to information or records’ by means of an NSL.” Id. at 1072 (quoting 18 U.S.C. § 2709(c)). We therefore recognized that the restriction § 2709(c) imposes was content based because it “target[ed] speech based on its communicative content,’ and restricts speech based on its ‘function or purpose.’” Id. (quoting Reed, 576 U.S. at 163). And “[w]hen the government restricts speech based on its

TWITTER, INC. V. GARLAND 23

content, a court will subject the restriction to strict scrutiny.” Id. at 1070. NSL requires strict scrutiny here because the restriction on Twitter’s speech is content based. Twitter is subject to a series of statutory nondisclosure obligations based on its receipt of NSLs and FISA orders. See 18 U.S.C. § 2709(c); 50 U.S.C. §§ 1805(c)(2)(B), 1824(c)(2)(B), 1842(d)(2)(B)(i), 1862(d)(2), 1881a(i)(1). The USA FREEDOM Act effectively created an exception to these prohibitions for certain disclosures about the aggregate receipt of national security process, within the predefined numerical bands explained above. See 50 U.S.C. § 1874(a); H.R. Rep. No. 114-109, pt. 1, at 27; Kris & Wilson, supra, § 13:5. But Twitter’s Transparency Report seeks to provide more detail than the USA FREEDOM Act allows to be disclosed. The nature of the government’s restriction on Twitter therefore necessarily arises from the content of Twitter’s proposed disclosure. Indeed, the government’s entire basis for seeking to limit Twitter’s disclosure is that public release of the classified content will harm national security. Thus, we are confronted with a content-based restriction, just as we were in NSL. Under circuit precedent, we review the government’s restriction on Twitter’s speech under the traditional First Amendment strict scrutiny framework. NSL was clear on this point, holding that strict scrutiny applied to the nondisclosure requirement in 18 U.S.C. § 2709(c) applicable to the receipt of individual NSLs. NSL, 33 F.4th at 1071–73. NSL governs us; there is no basis in law or logic to apply a different tier of scrutiny to the speech restriction now before us. The nondisclosure requirements imposed on recipients of national security legal process at issue here are effectively identical to those we considered in NSL, just aggregated— 24 TWITTER, INC. V. GARLAND

instead of being prohibited from disclosing the receipt of one letter, the recipient is prohibited from disclosing the receipt of a certain number of letters or orders. Both sides in this case ask for something other than strict scrutiny, but their arguments are foreclosed by our holding in NSL. The government suggests that some lesser form of scrutiny should apply, but NSL is directly contrary on this point. See id. at 1071–73. And while Twitter maintains that an even higher standard of “extraordinarily exacting” scrutiny should apply, NSL specifically rejected that argument. See id. at 1076 n.21 (holding that a request to apply “a higher standard than strict scrutiny” was “meritless,” and that New York Times Co. v. United States (Pentagon Papers), 403 U.S. 713 (1971) (per curiam), did not require otherwise). Thus, under circuit precedent, the restriction on Twitter’s speech is a content-based limitation that we review under the strict scrutiny framework. B To satisfy strict scrutiny, a restriction on speech is justified only if the government demonstrates that it is narrowly tailored to serve a compelling state interest. Reed, 576 U.S. at 163; NSL, 33 F.4th at 1070. There is no dispute about the government’s compelling interest here. “It is ‘obvious and unarguable’ that no governmental interest is more compelling than the security of the Nation.” Haig v. Agee, 453 U.S. 280, 307 (1981) (quoting Aptheker v. Sec’y of State, 378 U.S. 500, 509 (1964)); see also Holder v. Humanitarian Law Project, 561 U.S. 1, 28 (2010). It follows that “keeping sensitive information confidential in order to protect national security is a compelling government interest,” too. NSL, 33 F.4th at 1072 (citing Egan, 484 U.S.

TWITTER, INC. V. GARLAND 25

at 527; Snepp v. United States, 444 U.S. 507, 509 n.3 (1980)). This case thus turns on the narrow-tailoring prong of the strict scrutiny framework. To be narrowly drawn, a “curtailment of free speech must be actually necessary to the solution.” Brown v. Ent. Merchs. Ass’n, 564 U.S. 786, 799 (2011). “If a less restrictive alternative would serve the Government’s purpose, the legislature must use that alternative.” United States v. Playboy Ent. Grp., Inc., 529 U.S. 803, 813 (2000). But while a “restriction is not narrowly tailored if less restrictive alternatives would be at least as effective in achieving the legitimate purpose that the statute was enacted to serve,” we have previously observed in this same general context that strict scrutiny does not require the content-based restriction to be “perfectly tailored.” NSL, 33 F.4th at 1073 (first quoting Reno v. Am. Civil Liberties Union, 521 U.S. 844, 874 (1997); and then quoting Williams-Yulee v. Fla. Bar, 575 U.S. 433, 454 (2015)). The government is entitled deference when it comes to factual judgments bearing on national security. See Humanitarian Law Project, 561 U.S. at 33–34 (explaining that in the area of national security and foreign affairs, the “evaluation of the facts by the Executive, like Congress’s assessment, is entitled to deference”). But at the same time, “[w]e do not defer to the Government’s reading of the First Amendment, even when” national security interests are at stake. Id. at 34. Our decision in NSL, which is the most closely analogous precedent, demonstrates the type of careful review that strict scrutiny requires in this context. In NSL, we considered and rejected an as-applied challenge to the statutory provisions governing the non-disclosure requirements attached to individual NSLs. See NSL, 33 F.4th at 1073–76; 18 U.S.C. 26 TWITTER, INC. V. GARLAND

§ 2709(c). “Analyzing the statute as a whole,” we held that the statutory scheme was narrowly tailored to the government’s compelling national security interest in protecting the details of its intelligence investigations. NSL, 33 F.4th at 1074. In particular, we emphasized the statutory requirement that the government must make an “individualized analysis of each [NSL] recipient” when imposing nondisclosure restrictions—an analysis that “may include consideration of the size of the recipient’s customer base.” Id. This mandatory, focused inquiry ensured that the government would not exercise unfettered discretion but rather guaranteed that it would have to substantiate each nondisclosure requirement based on the individual circumstances. Id. The required means-end connection between the restriction imposed and the government’s national security interest was also established through the “narrow, objective, and definite” statutory standards that defined the contours of the government’s authority to impose nondisclosure restrictions. Id. The statute confined the imposition of individual NSL nondisclosure obligations to particular situations, such as when disclosure threatened ongoing counterintelligence operations or would endanger the lives of others. Id. (citing 18 U.S.C. § 2709(c)). This supported a sufficiently close fit between the government’s speech restriction and its national security goals. Id. We also noted that the statute ameliorated concerns that a change in circumstances could render the continued imposition of the nondisclosure obligations unnecessary. Id. at 1075. We highlighted the ready availability of judicial review in which the government had the burden of demonstrating the “continued necessity” of the restriction, which ensured that the limitation on speech was not “in place

TWITTER, INC. V. GARLAND 27

longer than [wa]s necessary to serve the government’s compelling interest.” Id. at 1075–76. Considered as a whole, NSL instructs that under the narrow tailoring prong of the strict scrutiny analysis, we must guarantee that the means by which the government is limiting Twitter’s speech—here, redacting portions of the Transparency Report—is sufficiently calibrated toward protecting the government’s proffered national security interest. And to guarantee that fit, we must satisfy ourselves that the government made a sufficiently particularized inquiry that substantiates the need for the redactions in the specific context in which Twitter operates. See id. at 1073– 76. Against this legal backdrop, we now turn to whether the government’s restriction on Twitter’s speech is narrowly tailored. To meet its burden under strict scrutiny, the government in the district court submitted three rounds of classified and unclassified declarations to support its position that information in Twitter’s Transparency Report was classified and could not be publicly disclosed without endangering national security. These declarations culminated in the classified and unclassified declarations of Jay S. Tabb, Jr., the EAD of the FBI’s National Security Branch. After an extended review of these materials, the district court found that based on “the totality of the evidence provided in this case, including the classified declarations,” the government had satisfied strict scrutiny and “no more narrow tailoring of the restrictions can be made.” In a case such as this involving information that the government contends is classified, we may review the classified materials ex parte and in camera. See, e.g., Kashem v. Barr, 941 F.3d 358, 385 (9th Cir. 2019); 28 TWITTER, INC. V. GARLAND

Mohamed v. Jeppesen Dataplan, Inc., 614 F.3d 1070, 1086 (9th Cir. 2010) (en banc); Kasza v. Whitman, 325 F.3d 1178, 1180 (9th Cir. 2003); Kasza v. Browner, 133 F.3d 1159, 1168–69 (9th Cir. 1998). Having intently studied the classified and unclassified materials in the record, we agree with the district court’s considered assessment. While we are not at liberty to disclose the contents of the classified materials that we reviewed, our analysis under the narrow tailoring prong depends principally on the knowledge we gleaned from our review of that material. The classified materials provided granular details regarding the threat landscape and national security concerns that animated the higher-level conclusions presented in the unclassified declarations. The classified declarations spell out in greater detail the importance of maintaining confidentiality regarding the type of matters as to which intelligence requests are made, as well as the frequency of these requests. Against the fuller backdrop of these explicit illustrations of the threats that exist and the ways in which the government can best protect its intelligence resources, we are able to appreciate why Twitter’s proposed disclosure would risk making our foreign adversaries aware of what is being surveilled and what is not being surveilled—if anything at all. Given these concerns and this fuller backdrop, we are willing to accept the main conclusions outlined in the unclassified materials, which express generally why revealing the information Twitter wishes to disclose would significantly harm the government’s national security operations by signaling to our adversaries what communication channels to avoid and which to use. Viewed in light of the classified declarations, Mr. Tabb’s Unclassified Declaration thus compellingly explains how the redactions on Twitter’s Transparency Report in the

TWITTER, INC. V. GARLAND 29

specific context of Twitter’s operations are well-calibrated to achieving the government’s national security goals. Taken as a whole, the government’s declarations specifically and persuasively explain why Twitter’s proposed Transparency Report may not be released in fully unredacted form. Mr. Tabb’s Unclassified Declaration explains that “the information about Twitter’s receipt of national security process that was redacted from Twitter’s draft Transparency Report is properly classified.” Unclassified Tabb Decl. ¶ 5. Mr. Tabb also well describes how any “unauthorized disclosure reasonably could be expected to result in serious damage to the national security.” Id. In particular, “disclosure of the information at issue here would provide international terrorists” and other bad actors with “a roadmap to the existence or extent of Government surveillance and capabilities associated with Twitter.” Id. More generally, “[d]isclosure of the information Twitter seeks to publish would provide highly valuable insights into where and how the United States is or is not deploying its investigative and intelligence resources.” Id. at ¶ 7. This “would tend to reveal which communications services may or may not be secure, which types of information may or may not have been collected, and thus whether or to what extent the United States is or is not aware of the activities of these adversaries.” Id. Mr. Tabb further explained why the granular nature of the information that Twitter seeks to publish would pose particular problems. Specifically, Mr. Tabb cautioned, “[d]isclosure of the kind of granular data regarding the national security legal process received by Twitter, as set forth in its draft Transparency Report, would reveal such information as: 30 TWITTER, INC. V. GARLAND

(i) incremental increases or decreases in collection over time, which would show whether the Government has a significant presence or investigative focus on a particular platform; (ii) the collection of content or non-content information, which would show whether and to what extent the Government is collecting certain types of information on that platform; and (iii)the fact of whether or when the recipient received a particular type of process at all, which may reflect different collection capabilities and focus on that platform, different types of information collected, and locations of FBI targets.

Unclassified Tabb Decl. ¶ 17. “[B]y detailing the amount, if any, of each particular type of process Twitter had received during a particular period, and over time, this data would reveal the extent to which Twitter was or was not a safe channel of communication for our adversaries.” Id. ¶ 18. As Mr. Tabb concluded, “[t]he granularity of the data that Twitter seeks to publish would reveal or tend to reveal information about the extent, scope, and reach of the Government’s national security collection capabilities and investigative interests—including its limitations and vulnerabilities.” Id. ¶ 21. Mr. Tabb also explained that if Twitter were allowed to make its granular disclosures, other recipients of national security process would seek to do the same. And the result would be an even greater exposure of U.S. intelligence

TWITTER, INC. V. GARLAND 31

capabilities and strategies. As Mr. Tabb wrote, “[i]f the Court were to grant Twitter the relief it seeks in this case, other providers would almost certainly seek to make the same types of disaggregated, granular disclosures regarding their receipt of national security process.” Id. ¶ 20. If that were allowed, it would “provide a comprehensive picture of the Government’s use of national security process that adversaries would use to evaluate the Government’s collection capabilities and vulnerabilities, as well as its investigative practices.” Id. Throughout his Unclassified Declaration, Mr. Tabb notes that greater detail and further explanation is provided in his Classified Declaration. Mr. Tabb further incorporates the classified declarations from the other government officials who preceded him in his role. As noted above, we have carefully reviewed the classified declarations in camera. And, as we have explained, those declarations provide more particularized reasons why the specific information Twitter seeks to publish would harm national security, reflecting the government’s individualized analysis of Twitter’s proposed disclosure. Mr. Tabb’s Classified Declaration, and the additional classified materials on which it relies, are compelling. His Classified Declaration, in combination with the other classified and unclassified materials, has convinced us that the government’s restriction on Twitter’s speech is narrowly tailored and survives strict scrutiny. Twitter’s arguments to the contrary are unpersuasive. Twitter argues that the government, prior to preventing the Transparency Report’s full disclosure, should have conducted an “individualized” inquiry into whether the publication should be prevented, and that the government failed to do so. Twitter bases this asserted requirement on 32 TWITTER, INC. V. GARLAND

Florida Star v. B.J.F., 491 U.S. 524, 540 (1989), in which the Supreme Court held that some “individualized adjudication” was required before a state could impose tort liability upon a newspaper that published the name of a rape victim. But here the record indicates that the FBI did conduct an individualized analysis of the harms that would be caused by Twitter’s disclosure of the unredacted Report. Our review of the record, including the classified materials, confirms that Twitter’s allegation is not correct. Twitter also argues that the government’s consideration of how other companies might disclose similar information violates the “individualized” inquiry requirement. But Twitter conflates two separate issues. The government could conduct an individualized inquiry into the harm that Twitter’s disclosure would make, including the harm that would be caused if an adversary considered the information Twitter disclosed alongside similar information from other companies. The government’s inquiry is no less “individualized” simply because it took into account the fact that if Twitter were allowed to publish the information in question, many other companies would do the same, leading to serious national security consequences. Twitter points to no contrary authority. And again, we conclude that the government’s review was sufficiently individualized, particularly in view of the Classified Tabb Declaration. In sum, the classified and unclassified materials in this case confirm that the government’s restrictions on Twitter’s speech survive strict scrutiny. We hold that the government’s redactions of Twitter’s Transparency Report do not violate the First Amendment.

TWITTER, INC. V. GARLAND 33

III

We turn next to Twitter’s claim that the procedures associated with the government’s restriction on Twitter’s speech failed to comport with Freedman v. Maryland, 380 U.S. 51 (1965). We hold that the specific procedural requirements of Freedman do not apply here. And we further conclude that the procedures that were followed— which were robust and which resembled the Freedman requirements in key respects—were sufficient to withstand any broader procedural challenge that Twitter has raised. 4 A In First Amendment law, a prior restraint is an order “forbidding certain communications when issued in advance of the time that such communications are to occur.” Alexander v. United States, 509 U.S. 544, 550 (1993) (emphasis and quotations omitted). In Freedman and its progeny, the Supreme Court developed a set of procedural safeguards for censorship regimes involving content-based prior restraints: “(1) any restraint prior to judicial review can be imposed only for a specified brief period during which the status quo must be maintained; (2) expeditious judicial

4 In its decision granting summary judgment to the government, the district court concluded that Twitter had not raised a challenge under Freedman. The parties agree this determination was mistaken. The record reflects that Twitter raised Freedman at various points in the litigation. The government thus concedes that it “had adequate notice of Twitter’s claims concerning [the Freedman] procedural safeguards.” The Freedman issue has been fully briefed on appeal, and the district court has already offered its tentative conclusions that Freedman may be implicated here. As Twitter itself argues, judicial economy counsels in favor of resolving the Freedman issue rather than remanding for further consideration. 34 TWITTER, INC. V. GARLAND

review of that decision must be available; and (3) the censor must bear the burden of going to court to suppress the speech and must bear the burden of proof once in court.” Thomas, 534 U.S. at 321 (quoting FW/PBS, 493 U.S. at 227 (principal opinion of O’Connor, J.)). Some background on the Freedman line of cases helps explicate these procedural requirements and demonstrates why the law imposes them in some situations. In Freedman, a Maryland “motion picture censorship statute” made it “unlawful to sell, lease, lend, exhibit or use” any film unless it was submitted to and approved by the state’s “Board of Censors.” 380 U.S. at 52 & n.1. That board had the authority to “license such films . . . which are moral and proper,” and to refuse to permit films that, “in the judgment of the Board,” are “obscene,” or that tend “to debase or corrupt morals or incite to crimes.” Id. at 52 & n.2. Recognizing that “any system of prior restraints of expression comes to this Court bearing a heavy presumption against its constitutional validity,” the Supreme Court held that the Maryland scheme was an unconstitutional prior restraint. Id. at 57–60 (quoting Bantam Books, Inc. v. Sullivan, 372 U.S. 58, 70 (1963)). In particular, the Court concluded that “a noncriminal process which requires the prior submission of a film to a censor avoids constitutional infirmity only if it takes place under procedural safeguards designed to obviate the dangers of a censorship system.” Id. at 58. Although the Supreme Court would later crystallize the Freedman procedures into the three-part formulation that we set forth above, see Thomas, 534 U.S. at 321, Freedman outlined those same basic procedural features and explained why they were constitutionally mandated in the censorship context.

TWITTER, INC. V. GARLAND 35

Freedman first explained that because of the “transcendent value of speech,” as a matter of “due process,” the “burden of proving that the film is unprotected expression must rest on the censor.” 380 U.S. at 58 (quoting Speiser v. Randall, 357 U.S. 513, 526 (1958)). Second, it was constitutionally necessary to have ready access to judicial review “because only a judicial determination in an adversary proceeding ensures the necessary sensitivity to freedom of expression,” and “only a procedure requiring a judicial determination suffices to impose a valid final restraint.” Id. Because “the censor’s business is to censor,” the Court concluded that a censor “may well be less responsive than a court . . . to the constitutionally protected interests in free expression.” Id. at 57–58. And “[i]f it is made unduly onerous, by reason of delay or otherwise, to seek judicial review, the censor’s determination may in practice be final.” Id. at 58. Thus, “[a]ny restraint imposed in advance of a final judicial determination on the merits must similarly be limited to preservation of the status quo for the shortest fixed period compatible with sound judicial resolution.” Id. at 59. Finally, Freedman explained, “the procedure must also assure a prompt final judicial decision” because of the potential that the temporary and “possibly erroneous” denial of a license could have a “deterrent effect” against speech. Id. That said, the Court made clear that it did “not mean to lay down rigid time limits or procedures, but to suggest considerations” in devising a legislative scheme that would “avoid the potentially chilling effect of the Maryland statute on protected expression.” Id. at 61. Although it eschewed imposing rigid formalities on these types of schemes, Freedman concluded that the basic procedural safeguards it set forth were constitutionally necessary because without them, “it may prove too 36 TWITTER, INC. V. GARLAND

burdensome to seek review of the censor’s determination.” Id. at 59. The Court pointed specifically to the nature of the film industry and to the incentives that film exhibitors and distributors would have (or would lack) in this context. As to the exhibitor, its “stake in any one picture may be insufficient to warrant a protracted and onerous course of litigation.” Id. And the film’s distributor might also forgo a costly challenge if the distributor could show the film freely in most other places. Id. Beyond Freedman, the Supreme Court has imposed these procedural protections in other cases as well, but it “has generally focused on two types of government schemes requiring safeguards: censorship schemes and licensing schemes.” NSL, 33 F.4th at 1076–77. Thus, the Court has applied Freedman to customs officials’ seizing “obscene or immoral” articles, United States v. Thirty-Seven (37) Photographs, 402 U.S. 363, 365 & n.1, 366–68, 373–75 (1971), the postmaster’s halting mail that contains “allegedly obscene materials,” Blount v. Rizzi, 400 U.S. 410, 411–14, 417–19 (1971), a board’s requiring permission before showing an allegedly obscene play at a municipal theater, Southeastern Promotions, Ltd. v. Conrad, 420 U.S. 546, 547–48, 559–62 (1975), and a court’s ex parte restraining order that prevented a planned rally of offensive “political” speech, Carroll v. President & Comm’rs of Princess Anne, 393 U.S. 175, 176–77, 181–82 (1968); see also, e.g., Teitel Film Corp. v. Cusack, 390 U.S. 139, 140– 42 (1968) (per curiam) (invalidating a city’s “Motion Picture Censorship Ordinance”); Bantam Books, 372 U.S. at 59–62, 71 (invalidating, pre-Freedman, a scheme by which a state “Commission to Encourage Morality in Youth” would declare books or magazines objectionable for distribution to young people). In these cases, the Court recognized that “a

TWITTER, INC. V. GARLAND 37

scheme conditioning expression on a licensing body’s prior approval of content ‘presents peculiar dangers to constitutionally protected speech.’” Thomas, 534 U.S. at 321 (quoting Freedman, 380 U.S. at 57). At the same time, the Supreme Court has emphasized that in some licensing contexts, the required safeguards “are less extensive than those required in Freedman because they do ‘not present the grave dangers of a censorship system.’” NSL, 33 F.4th at 1077 (quoting City of Littleton v. Z.J. Gifts D-4, L.L.C., 541 U.S. 774, 783 (2004)). Most relevant here is City of Littleton, in which the Court rejected a facial challenge to a municipal licensing scheme for adult businesses. 541 U.S. at 776. The city ordinance at issue there established certain circumstances that required the city to deny a license to operate an adult business, such as if the applicant were underage or had not timely paid taxes. Id. at 783. The Court concluded that specially expedited time frames for judicial review were not required in that context. Id. at 782–84. As the Court explained, where “the regulation simply conditions the operation of an adult business on compliance with neutral and nondiscretionary criteria, and does not seek to censor content, an adult business is not entitled to an unusually speedy judicial decision of the Freedman type.” Id. at 784 (citations omitted). In such cases, the state’s “ordinary judicial review procedures suffice as long as the courts remain sensitive to the need to prevent First Amendment harms.” Id. at 781. And “whether the courts do so is a matter normally fit for case-by-case determination rather than a facial challenge.” Id. at 782; see also FW/PBS, 493 U.S. at 228 (principal opinion of O’Connor, J.) (“Because the licensing scheme at issue in these cases does not present the grave ‘dangers of a 38 TWITTER, INC. V. GARLAND

censorship system,’ we conclude that the full procedural protections set forth in Freedman are not required.” (quoting Freedman, 380 U.S. at 58)). In addition to not insisting on compliance with inflexible procedures even within the contexts in which Freedman might otherwise apply, the Supreme Court has not held that compliance with Freedman’s safeguards is required in every instance in which expression is restrained in advance because of its content. In particular, the Court “has not held that . . . [certain] government confidentiality restrictions must have the sorts of procedural safeguards required for censorship and licensing schemes.” NSL, 33 F.4th at 1078. Two precedents are most relevant in this regard. In Seattle Times Company v. Rhinehart, 467 U.S. 20, 27–28 (1984), a newspaper company challenged a protective order preventing it from disseminating information it acquired through pretrial discovery. Although the order restricted the newspaper’s ability to share information of significant public interest, the Court concluded that the order was “not the kind of classic prior restraint that requires exacting First Amendment scrutiny.” Id. at 33. The newspaper company had received the information it sought to publish only as “a matter of legislative grace” through the mechanisms of civil discovery. Id. at 32. Information obtained through discovery requests in litigation does not come from “a traditionally public source of information.” Id. at 33. Limitations on the disclosure of such information thus “do[] not raise the same specter of government censorship that such control might suggest in other situations.” Id. at 32. The Court ultimately affirmed that the protective order satisfied the First Amendment without discussing Freedman. Id. at 37.

TWITTER, INC. V. GARLAND 39

Nor did the Court mention Freedman in Butterworth v. Smith, 494 U.S. 624 (1990). In Butterworth, the Court considered a state law preventing grand jury witnesses from disclosing the testimony that they gave before the grand jury. Id. at 626–27. The Court held the restriction unconstitutional “insofar as [it] prohibits a grand jury witness from disclosing his own testimony after the term of the grand jury has ended.” Id. at 626. In support of its conclusion, the Court emphasized that the statute’s effect was “dramatic.” Id. at 635. Before the witness was called to testify, he “possessed [] information on matters of admitted public concern about which he was free to speak at will.” Id. But after testifying, the statute restrained his speech. Id. The state’s interest in preserving the secrecy of grand jury proceedings did not overcome the witness’s “First Amendment right to make a truthful statement of information he acquired on his own.” Id. at 636 (emphasis added). Critically, however, Butterworth left intact “that part of the Florida statute which prohibit[ed] the witness from disclosing the testimony of another witness.” Id. at 633; see also id. at 632 (distinguishing Seattle Times because “[h]ere, by contrast, we deal only with respondent’s right to divulge information of which he was in possession before he testified before the grand jury, and not information which he may have obtained as a result of his participation in the proceedings of the grand jury”); id. at 636 (Scalia, J., concurring) (expressing “considerable doubt” over the state’s ability to restrain a witness from disclosing information that he already knew before he entered the grand jury room, but noting that it would present “[q]uite a different question” to restrict the witness from disclosing 40 TWITTER, INC. V. GARLAND

what he learned from others, “which is in a way information of the State’s own creation”); NSL, 33 F.4th at 1078. As it stands, therefore, Freedman applies to some speech restrictions, but the Supreme Court has not held that Freedman’s specific procedures apply to every limitation that restricts speech in advance of its disclosure. B With the Freedman doctrine set forth, we now turn to the question of whether the government was required to comply with Freedman’s exact procedures in restricting Twitter’s publication of its Transparency Report. We also consider whether, as a general matter, Freedman applies when the government prohibits the publication of information that exceeds the limited aggregate disclosures that the USA FREEDOM Act allows. These are largely issues of first impression, although they bear similarities to the First Amendment challenge to the individual NSL nondisclosure requirement that we considered in NSL. That case likewise involved the post- USA FREEDOM Act version of the statute. NSL, 33 F.4th at 1068–69. As we explained above, the relevant provision at issue in NSL, 18 U.S.C. § 2709(c), generally prohibits the recipient of a national security letter from disclosing the fact of its receipt. NSL, 33 F.4th at 1063. Among the issues posed in NSL was whether Freedman applied to § 2709(c)’s speech restriction. We concluded in NSL that we did not need to answer that question because even if the procedural safeguards of Freedman were required, the statute “in fact provides all of them.” Id. at 1079. But although it was unnecessary to reach the question, we provided several reasons in NSL why we

TWITTER, INC. V. GARLAND 41

were skeptical that Freedman applied to the individual NSL nondisclosure obligations at issue. Id. at 1076–78. We noted that the NSL nondisclosure requirement “does not resemble the[] government censorship and licensing schemes” to which Freedman traditionally applies because the NSL law “neither requires a speaker to submit proposed speech for review and approval, nor does it require a speaker to obtain a license before engaging in business.” Id. at 1077. “Rather,” we continued, the statute “prohibits the disclosure of a single, specific piece of information that was generated by the government: the fact that the government has requested information to assist in an investigation addressing sensitive national security concerns.” Id. Citing Seattle Times and Butterworth, we opined that a restriction on the dissemination of this type of information was “more similar to government confidentiality requirements that have been upheld by courts”—requirements to which Freedman has not been extended. Id. at 1078. That is, Seattle Times and Butterworth demonstrated that the Supreme Court “ha[d] not held that these sorts of confidentiality restrictions must have the sorts of procedural safeguards required for censorship and licensing schemes.” Id. As we have noted, however, these comments in NSL were dicta and thus do not bind us here. Unlike NSL, this case does require us to pass upon whether Freedman applies to the government’s restriction on Twitter’s dissemination of classified information. Having undertaken our own independent review of the issue, we conclude that Freedman’s specific procedures do not apply in this case. Freedman established constitutionally mandated “procedural safeguards designed to obviate the dangers of a censorship system.” 380 U.S. at 58. These procedures were founded on the recognition that “a scheme 42 TWITTER, INC. V. GARLAND

conditioning expression on a licensing body’s prior approval of content ‘presents peculiar dangers to constitutionally protected speech.’” Thomas, 534 U.S. at 321 (quoting Freedman, 380 U.S. at 57). But as the Supreme Court explained in City of Littleton, the specific procedural requirements of Freedman do not come into play in the case of statutory schemes that “do not present the grave dangers of a censorship system.” 541 U.S. at 783 (quoting FW/PBS, 493 U.S. at 228 (principal opinion of O’Connor, J.)); see also FW/PBS, 493 U.S. at 228 (principal opinion of O’Connor, J.) (“Because the licensing scheme at issue in these cases does not present the grave ‘dangers of a censorship system,’ we conclude that the full procedural protections set forth in Freedman are not required.” (quoting Freedman, 380 U.S. at 58)). Although the licensing scheme at issue in City of Littleton was a different type of regime than what we have here, City of Littleton confirms that Freedman has not been extended to every regime that may be characterized as an advance restriction on speech. In this case, a restriction on the disclosure of classified information is not akin to the censorship schemes to which Freedman has been applied. As in the context of information obtained in civil discovery subject to a protective order, see Seattle Times, 467 U.S. at 32–33, or learned in grand jury proceedings, see Butterworth, 494 U.S. at 632–33, 635–36, the recipient of the classified information at issue here is restrained only in speaking about information it received from the government. And that restriction is taking place in an area in which courts have regarded government confidentiality restrictions not as censorship, but as legitimate means of protecting certain government- provided confidential information. See, e.g., Egan, 484 U.S.

TWITTER, INC. V. GARLAND 43

at 527; Snepp, 444 U.S. at 509 n.3. As we recognized in NSL, courts have upheld certain government confidentiality requirements—regardless of the type of information being quelled—without discussing or considering Freedman’s application. 33 F.4th at 1078. Freedman’s procedures, which were designed to curb traditional censorship regimes, are not required in the context of government restrictions on the disclosure of information transmitted confidentially as part of a legitimate government process, because such restrictions do not pose the same dangers to speech rights as do traditional censorship regimes. See NSL, 33 F.4th at 1078 (citing Seattle Times and Butterworth). This does not mean, of course, that Twitter is entitled to no procedural protections. As we explain below, the process afforded here was both substantial and sufficient. But the specific procedural framework of Freedman is not constitutionally required. What we have here is not “a classic prior restraint,” Seattle Times, 467 U.S. at 33, and for the reasons we have explained, Freedman’s particular procedural framework does not govern. Twitter’s arguments to the contrary are unavailing. Twitter is correct that, as noted above, the Supreme Court has required compliance with Freedman in some cases beyond the quintessential film censorship scheme. See, e.g., Carroll, 393 U.S. at 176–77, 181–82 (restraining order preventing political rallies); Nat’l Socialist Party of Am. v. Vill. of Skokie, 432 U.S. 43, 43–44 (1977) (per curiam) (injunction preventing political party from marching and distributing certain materials); Vance v. Universal Amusement Co., 445 U.S. 308, 309 (1980) (per curiam) (injunction indefinitely preventing display of motion pictures under public nuisance statute). But although these cases may not have involved censorship schemes exactly 44 TWITTER, INC. V. GARLAND

like that in Freedman itself, it is obvious that they presented closely analogous speech restrictions. We are likewise not persuaded by the Second Circuit’s decision in John Doe, Inc. v. Mukasey, 549 F.3d 861, 876– 78 (2d Cir. 2008), which held that the pre-USA FREEDOM Act nondisclosure requirements for individual NSLs must comply with Freedman. Just as we have, the Second Circuit in Doe recognized that the individual NSL nondisclosure requirement is “not a typical prior restraint” because it “is not a restraint imposed on those who customarily wish to exercise rights of free expression, such as speakers in public fora, distributors of literature, or exhibitors of movies.” Id. at 876–77. But the court then rejected the analogy to the grand jury context on the theory that “[t]he justification for grand jury secrecy inheres in the nature of the proceeding,” whereas “secrecy might or might not be warranted” for national security letters. Id. The problem with this reasoning is that it fails to recognize that Freedman has not been extended to long-accepted confidentiality restrictions concerning government-provided information because of the differences between these types of confidentiality requirements and traditional prior restraints. C Even though Freedman’s specific procedural framework does not apply here, Twitter received considerable process— including some of the process that Freedman envisioned. This is hardly a case in which a would-be speaker was entirely frustrated by an administrative censor. We conclude that the process Twitter received was sufficiently “sensitive to the need to prevent First Amendment harms.” City of Littleton, 541 U.S. at 781.

TWITTER, INC. V. GARLAND 45

When Twitter circulated its proposed publication to the government, the FBI reviewed it and met with Twitter to discuss the issues before ultimately determining that certain information in Twitter’s publication could not be publicly released. We expect that going forward, the government will demonstrate comparable diligence when presented with these kinds of requests to ensure that free speech rights are adequately protected in the national security context. Twitter then filed this lawsuit just four weeks after the government informed Twitter that it could not publish the Transparency Report in full. Although Twitter shouldered the burden of filing the lawsuit, it had no apparent difficulty bearing that burden, and it was able to ensure that any speech restraint prior to judicial review was relatively brief. See Thomas, 534 U.S. at 321. That stands in contrast to the film context, in which Freedman concluded that it “may prove too burdensome” for certain speakers “to seek review of the censor’s determination” because movie distributors and exhibitors may have too little stake in displaying a single film in a particular location covered by a censorship scheme. 380 U.S. at 59. There was no similar incentive problem here. We have already held that “the Freedman burden-of- instituting proceedings safeguard does not apply” in the context of certain zoning and licensing schemes. Baby Tam & Co., Inc. v. City of Las Vegas, 247 F.3d 1003, 1008 (9th Cir. 2001); see also Dream Palace v. County of Maricopa, 384 F.3d 990, 1001 n.6, 1009–10 (9th Cir. 2004). We similarly conclude here that Twitter has not demonstrated why obligating the government to institute these proceedings was constitutionally mandated, or how it would have materially affected the resolution of this dispute. 46 TWITTER, INC. V. GARLAND

Once Twitter’s lawsuit was filed, the district court gave the case careful and diligent consideration. As Freedman requires, the government bore the burden of proof in demonstrating that the speech restriction was permissible. See Thomas, 534 U.S. at 321; Freedman, 380 U.S. at 58. Our review, and that of the district court, was conducted using strict scrutiny, which is the “most demanding test known to constitutional law.” City of Boerne v. Flores, 521 U.S. 507, 534 (1997). It is true, of course, that the proceedings in the district court and in this Court took considerable time to resolve. But Freedman itself noted that the timetable for judicial review may depend on the context of the restriction. See 380 U.S. at 60–61. And, as we held in NSL, “[n]either Freedman nor any other Supreme Court decision requires that judicial review be completed in a specified time frame.” 33 F.4th at 1079. In this case, we conclude that any delay was warranted and that “ordinary court procedural rules and practices” are generally sufficient “to avoid delay-related First Amendment harm.” City of Littleton, 541 U.S. at 782, 784; see also Dream Palace, 384 F.3d at 1003–04. The district court proceedings in this case required multiple rounds of classified and unclassified declarations. We cannot say that this process was unnecessary. Indeed, it was indispensable to our ultimate review. The specific protocols that govern judicial review of cases involving classified information, see Reagan, Keeping Government Secrets, supra, at 9–20, 22– 23—which here included judicial review and discussion of classified information in secure facilities—similarly led to more protracted proceedings. But this deliberative process was necessary in view of the national security sensitivity of the information at issue. We are also hopeful that having now resolved some of the complex legal issues underlying

TWITTER, INC. V. GARLAND 47

this dispute, future disputes of this nature may move more quickly, in a manner consistent with the First Amendment and accounting for the unique needs that are attendant to the consideration of classified information. See City of Littleton, 541 U.S. at 782 (“We presume that courts are aware of the constitutional need to avoid ‘undue delay result[ing] in the unconstitutional suppression of protected speech.’” (quoting FW/PBS, 493 U.S. at 228)); id. (describing how “ordinary court procedural rules and practices, in Colorado as elsewhere, provide reviewing courts with judicial tools sufficient to avoid delay-related First Amendment harm”). Future litigants “remain free to raise special problems of undue delay in individual cases.” Id. at 784. In sum, although the specific Freedman procedures do not apply in these circumstances, Twitter received some Freedman-like protections, and it is entitled to due process when it wishes to disclose information like that at issue here—due process that Twitter received in this case. IV Twitter lastly argues that the government violated due process by refusing to allow Lee Rubin, Twitter’s lead outside counsel, access to the classified Tabb declaration and other classified materials that the government submitted in this case. This argument lacks merit. There is no general constitutional rule requiring the government to provide classified materials to an adversary in litigation. Nor is there a general constitutional rule allowing a party access to classified information by virtue of its decision to file a lawsuit that implicates that kind of information. That is true even if the party seeking the information has appropriate security clearances. As we have held, the government “might have a legitimate interest in 48 TWITTER, INC. V. GARLAND

shielding the materials even from someone with the appropriate security clearance.” Al Haramain Islamic Found., Inc. v. Dep’t of the Treasury, 686 F.3d 965, 983 (9th Cir. 2012) (“Al Haramain II”). From a procedural standpoint, our case law establishes that although “the Constitution does require that the government take reasonable measures to ensure basic fairness to the private party,” it “certainly does not require that the government take actions that would endanger national security.” Id. at 980. “[N]or does it require the government to undertake every possible effort to mitigate the risk of erroneous deprivation and the potential harm to the private party.” Id.; see also Kashem, 941 F.3d at 386. Our assessment of the required procedures—including who has access to what information—instead reflects “a case-by-case approach” that accounts for the fact that “the proper measures in any given case will depend on a number of factors.” Al Haramain, 686 F.3d at 984. As we have held, “the government may withhold classified information that truly implicates national security as long as it undertakes reasonable measures to mitigate the potential unfairness” to the plaintiff. Kashem, 941 F.3d at 380. In this case, the government submitted a declaration from Carl Ghattas, then-EAD of the FBI’s National Security Branch, which explained that under Executive Order 13,526, which governs the disclosure of classified information, the United States had determined that Rubin did not have the requisite “need to know” the classified information. The President’s Executive Order 13,526 defines “need to know” as “a determination within the executive branch in accordance with directives issued pursuant to this order that a prospective recipient requires access to specific classified information in order to perform or assist in a lawful and

TWITTER, INC. V. GARLAND 49

authorized governmental function.” Exec. Order No. 13,526, § 6.1(dd), 75 Fed. Reg. at 729. Mr. Ghattas concluded that “it does not serve a governmental function . . . to allow plaintiff’s counsel access to the classified FBI information at issue in this case to assist in representing the interests of a private plaintiff who has filed this civil suit against the government.” Mr. Ghattas contrasted Twitter’s outside counsel with federal judges, who are provided with classified information “necessary for the Court to perform its judicial function.” In response, Twitter maintains that Rubin does have a need to know the classified information in this case, so as to allow outside counsel fully to represent Twitter’s interests in this litigation. But under our precedents, this argument falls short. See Al Haramain II, 686 F.3d at 979 (directing application of the balancing test from Mathews v. Eldridge, 424 U.S. 319 (1976)). We have already determined that “the evidence is classified and truly implicates national security.” Kashem, 941 F.3d at 385. And we conclude that the process followed here mitigates the risk of an erroneous deprivation of Twitter’s First Amendment rights, see Mathews, 424 U.S. at 335, in a manner consistent with the government’s compelling interest in ensuring the confidentiality of classified information. See Kashem, 941 F.3d at 377–78, 382; Al Haramain II, 686 F.3d at 979–80. Twitter was provided with unclassified versions of the various declarations, which we have relied upon throughout this opinion. See Kashem, 941 F.3d at 385 (explaining that reasonable mitigation measures “may include disclosing the classified evidence to cleared counsel subject to a protective order or providing the complainant an unclassified summary of the classified evidence” (emphasis added)). The unclassified declarations provided Twitter with sufficient 50 TWITTER, INC. V. GARLAND

information by which to advance Twitter’s interests before this Court. The record amply demonstrates that Twitter’s capable counsel has vigorously advocated on behalf of its client. And although we appreciate Twitter’s concern that it cannot respond to what it does not know, Twitter’s interest in the classified information does not rise to the level of constitutional imperative. As we have made clear, “there is no general rule requiring both an unclassified summary and disclosure to cleared counsel.” Id. at 386 (citing Al Haramain II, 683 F.3d at 980); see also Al Haramain II, 683 F.3d at 983 (“We recognize that disclosure may not always be possible.”). The risk of erroneous deprivation is further mitigated by the extensive litigation process in this case, which involved multiple rounds of proceedings in the district court, multiple rounds of government submissions, and extensive in camera review of classified declarations in both the district court and this Court. In deciding Twitter’s challenge, we have “thoroughly and critically reviewed the government’s public and classified declarations,” Mohamed, 614 F.3d at 1086, under the demanding strict scrutiny framework. Our process here was not unusual. In the context of other similar judicial processes, we have conducted ex parte review of classified materials without finding a due process concern, even when those materials were critical to our resolution of the case. See, e.g., Kashem, 941 F.3d at 385; Mohamed, 614 F.3d at 1086. 5

5 Because we conclude that due process does not require the government to provide Twitter’s counsel with classified information, we do not reach the government’s argument that this information would be protected from disclosure under the state secrets privilege.

TWITTER, INC. V. GARLAND 51

Twitter protests that this case is different because “[u]nlike some of the litigants that have sought access to classified evidence over the years, Twitter is not a designated terrorist organization or foreign national whose access (even through cleared counsel) might legitimately raise national security concerns.” This argument fails. Twitter confuses whether the government could have allowed Twitter access to classified information with whether due process mandates that result. For the reasons we have explained, it does not. The process afforded to Twitter was constitutionally sufficient, even without its having received classified materials. Under these circumstances, the government was not required to draw distinctions among different types of litigants, as Twitter suggests, which could require potentially fraught predictions as to whether disclosure of classified materials to one group as opposed to another posed greater risks. Nor is this selective differentiation among litigants a task that is proper for the judiciary to undertake. See United States v. Ott, 827 F.2d 473, 477 (9th Cir. 1987) (“Congress has a legitimate interest in authorizing the Attorney General to invoke procedures designed to ensure that sensitive security information is not unnecessarily disseminated to anyone not involved in the surveillance operation in question, whether or not she happens for unrelated reasons to enjoy security clearance. We reject the notion that a defendant’s due process right to disclosure of FISA materials turns on the qualifications of his counsel.”); see also Al Haramain II, 683 F.3d at 983. In a case such as this, requiring courts to evaluate the perceived trustworthiness of individual litigants in their receipt of classified information would invite a weighing of interests that is beyond our role. 52 TWITTER, INC. V. GARLAND

That is not an inquiry we have undertaken before, and we do not do so now. * * * The government may not fend off every First Amendment challenge by invoking national security. But we must apply the First Amendment with due regard for the government’s compelling interest in securing the safety of our country and its people. We hold here that, both as a matter of substance and procedure, the government’s restriction on Twitter’s speech did not violate the First Amendment. The judgment of the district court is

AFFIRMED.

VANDYKE, Circuit Judge, concurring in the judgment: I agree with the majority’s conclusion in this case, and most aspects of its analysis, with our only significant disagreement being whether we need to rely on classified materials to resolve this case. I conclude that the unclassified materials are sufficient to meet the government’s burden. Rather than attempt to parse how that difference might change the analysis, I simply provide my own analysis below.

I. DISCUSSION

“[O]ne of the most difficult tasks in a free society like our own is the correlation between adequate intelligence to guarantee our nation’s security on the one hand, and the preservation of basic human rights on the other.” S. Rep. No. 95-604, pt. 1, at 4 (1977) (quoting former United States President Jimmy Carter). It’s a longstanding legal axiom

TWITTER, INC. V. GARLAND 53

that a government can safeguard liberty only if it has some latitude to narrowly restrict speech that endangers national security. See, e.g., 1 William Blackstone, Commentaries *126; 3 Joseph Story, Commentaries on the Constitution of the United States §§ 1874, 1878, at 731–33, 735–37 (1833). But it is as well-recognized that the First Amendment’s protection to speak freely about matters of public concern is “an opportunity essential to the security of the Republic,” and “a fundamental principle of our constitutional system.” New York Times v. Sullivan, 376 U.S. 254, 269 (1964) (quotation marks and citations omitted). This case requires us to address the intersection of those two weighty concerns: free speech and national security. More specifically, Twitter has brought an as-applied constitutional challenge asking whether the Government can constitutionally prevent it from disclosing in its Report classified information it obtained only through its involvement in the Government’s national security investigations. The First Amendment provides that “Congress shall make no law ... abridging the freedom of speech, or of the press.” U.S. Const. amend. I. But “the Supreme Court has concluded that some restrictions on speech are constitutional, provided they survive the appropriate level of scrutiny.” In re National Security Letter, 33 F.4th 1058, 1070 (9th Cir. 2022) (“NSL”). Our court analyzes nondisclosure requirements pertaining to national security in three steps. See id. at 1071 (evaluating 18 U.S.C. § 2709(c)’s nondisclosure requirement). First, we “determine whether the nondisclosure requirement is content based or content neutral.” Id. Second, “[i]f the nondisclosure requirement is content based, we then consider whether it survives strict scrutiny.” Id. Third, we “determine whether the 54 TWITTER, INC. V. GARLAND

nondisclosure requirement constitutes the type of restraint for which the procedural safeguards are required and, if so, whether it provides those safeguards.” Id. A. The Government’s Restrictions Result from the Statutory Framework. At the outset, it is important to first define the precise speech restrictions at issue here. On appeal, Twitter argues that “the origin of the restraint on [its] aggregate reporting is the FBI’s discretionary ‘classification’ of [its] … Report … and its continued assertion of that classification under Executive Order 13526 and 50 U.S.C. § 1874.” In contrast, the Government argues that “[t]he obligation of the recipients of national security process to protect the secrecy of classified information relating to that process stems from their statutory nondisclosure obligations” (emphasis added). The Government is right. It is the statutory nondisclosure requirements pertaining to electronic communication service providers’ (ECSPs’) receipt of national security process that prevent Twitter from disclosing the information it seeks to publish. The text of the statutory nondisclosure provisions at issue requires ECSPs to “protect [the] secrecy” of the investigation. 50 U.S.C. §§ 1805(c)(2)(B) (electronic surveillance), 1824(c)(2)(B) (physical searches), 1842(d)(2)(B)(i) (pen registers or trap and trace devices), 1881a(i)(1)(A) (persons abroad); see also 18 U.S.C. § 1862(d)(2) (mandating that appropriate recipients of a request for records not “disclose to any person … that the [FBI] has sought or obtained records pursuant to an order under this section”); 18 U.S.C. § 2709(c)(1)(A) (containing the NSL nondisclosure requirement mandating that “no [ECSP] … shall disclose to

TWITTER, INC. V. GARLAND 55

any person that the [FBI] has sought or obtained access to information or records under this section”). Disclosing any information about a particular national security investigation, including aggregated information that incorporates the occurrence of that investigation, directly undermines its “secrecy” by, at the very least, revealing its existence. Moreover, the only provision that permits the disclosure of any information pertaining to the receipt of national security process whatsoever, 50 U.S.C. § 1874, presents itself as an exception to the nondisclosure requirements accompanying the receipt of individual orders and subpoenas. See § 1874(a) (“A person subject to a nondisclosure requirement accompanying an order or directive under this chapter or a national security letter may, with respect to such order, directive, or national security letter, publicly report the following information using one of the [provided] structures ….” (emphases added)). Section 1874’s explicit incorporation of the nondisclosure requirements pertaining to the receipt of individual orders and subpoenas further enforces the statutory requirement to generally prohibit the disclosure of any information pertaining to the receipt of national security process, whether individualized or in the aggregate, except for information falling within the limited boundaries articulated in § 1874. Reading the nondisclosure requirements together, as one must, these provisions prohibit the disclosure of aggregate information pertaining to the receipt of national security process that falls outside of the limited bounds articulated in § 1874. See Food & Drug Admin. v. Brown & Williamson Tobacco Corp., 529 U.S. 120, 133 (2000) (“It is a fundamental canon of statutory construction that the words of a statute must be read in their context and with a view to 56 TWITTER, INC. V. GARLAND

their place in the overall statutory scheme.” (citation and internal quotation marks omitted)). Because the information Twitter seeks to disclose undisputedly falls outside of § 1874’s permissible boundaries, the statutory nondisclosure requirements prohibit the disclosure of the information Twitter seek to publish here. 1 B. Traditional Strict Scrutiny Applies. Returning to our court’s tripartite analysis, I easily conclude that the Government’s restrictions are content based and warrant the application of strict scrutiny. “A government’s restriction on speech is content based if a law applies to particular speech because of the topic discussed or the idea or message expressed.” NSL, 33 F.4th at 1071 (citation and internal quotations omitted). “[A] regulation or law that restricts speech based on its topic, idea, message, or content is ‘content based’ on its face, and is accordingly subject to strict scrutiny.” Id. at 1071–72. And when a non- disclosure requirement, “[b]y its terms … prohibits speech about one specific issue,” then “[s]uch a restriction targets speech based on its communicative content.” Id. at 1072 (citation, internal alterations, and quotation marks omitted). In this case, the Government’s restrictions are clearly content based. The unclassified FBI declarations reveal that the Government redacted certain information in Twitter’s Report because the “message expressed,” if published, would reasonably be expected to endanger national security.

1 Moreover, the logical extension of Twitter’s argument is that no statutory nondisclosure requirement exists for the disclosure of aggregate information pertaining to the receipt of national security process, which effectively renders the exceptions for amounts of aggregate reporting articulated in 50 U.S.C. § 1874(a) meaningless.

TWITTER, INC. V. GARLAND 57

See id.; see also Unclassified Tabb Decl. ¶ 5 (concluding that the redactions are “properly classified, and that its unauthorized disclosure reasonably could be expected to result in serious damage to the national security”); Unclassified Steinbach Decl. ¶ 5 (same). For example, in the Unclassified Tabb Declaration, Tabb testified that disclosure of the redacted information “would allow adversaries of the United States … significant insight into the U.S. Government’s counterterrorism and counterintelligence efforts and capabilities, or, significantly, the lack thereof, and into particular intelligence sources and methods.” Unclassified Tabb Decl. ¶ 16. By the FBI’s own attestations, therefore, it was precisely the content of the redacted information that could endanger national security if disclosed and accordingly justified the classification of that information. In addition to the executive branch’s own characterization of its classification of the redacted information in Twitter’s Report as content based, our court has already determined that at least part of the statutory nondisclosure framework at issue here is content based. NSL, 33 F.4th at 1063. In NSL, the panel reasoned that 18 U.S.C. § 2709(c) “prohibits speech about one specific issue: the recipient may not disclose to any person that the [FBI] has sought or obtained access to information or records by means of an NSL.” Id. at 1072 (citation and internal quotation marks omitted). “Such a restriction targets speech based on its communicative content, and restricts speech based on its function or purpose.” Id. (citation and internal quotation marks and alterations omitted). The panel therefore concluded that 18 U.S.C. § 2709(c) was content based on its face. Id. 58 TWITTER, INC. V. GARLAND

NSL controls the analysis of the statutory nondisclosure framework at issue here. As to the NSL nondisclosure requirement, NSL explicitly dictates that 18 U.S.C. § 2709(c) is content based. Id. And as to the other nondisclosure requirements pertaining to FISA orders, NSL’s rationale leads to the same conclusion: just like 18 U.S.C. § 2709(c), the nondisclosure requirements for FISA orders “prohibit[] speech about one specific issue: the recipient may not disclose to any person that the [government] has sought or obtained access to information or records by means of” a FISA order. Id. (citation and internal quotation marks omitted); see also 50 U.S.C. §§ 1805(c)(2)(B), 1824(c)(2)(B), 1842(d)(2)(B)(i), 1862(d)(2), 1881a(i)(1)(A). The statutory nondisclosure requirements at issue here are nearly identical to those the panel considered in NSL, just at a higher level of generality. But this is largely a distinction without a difference. Because both the executive branch’s classification of the redacted information in Twitter’s Report and the statutory nondisclosure requirements at issue “target speech based on its communicative content,” strict scrutiny applies. NSL, 33 F.4th at 1072 (citation and internal quotation marks omitted). Neither party disputes that the Government’s restrictions are content based. But they both nonetheless argue that a standard other than strict scrutiny governs. The Government argues that a standard of review more akin to intermediate scrutiny applies, 2 whereas Twitter argues that some extra-

2 Specifically, the Government argues that a standard more akin to intermediate scrutiny applies because the redacted information in Twitter’s report concerns information obtained solely through Twitter’s participation in confidential government activities. But in NSL, we

TWITTER, INC. V. GARLAND 59

strict form of strict scrutiny articulated in New York Times Co. v. United States, 403 U.S. 713, 714 (1971) (per curiam) (“Pentagon Papers”), applies. But both parties ignore NSL’s application of the traditional form of strict scrutiny to materially similar nondisclosure requirements. NSL even went so far as to determine that the same argument Twitter raises—that Pentagon Papers instructs the application of a more demanding form of strict scrutiny when evaluating nondisclosure requirements—is “meritless”:

The recipients argue that the NSL law should be held to a higher standard than strict scrutiny. According to the recipients, a content-based restriction imposed by a system of prior restraint is permissible only if (1) the harm to the governmental interest is highly likely to occur; (2) the harm will be irreparable; (3) no alternative exists for preventing the harm; and (4) the restriction will actually prevent the harm. This argument is meritless. No Supreme Court or Ninth Circuit opinion has articulated such a test, nor do the three cases cited by the recipients support it. The brief per curiam opinion in [Pentagon Papers] did not specify a test that should be applied to prior restraints.

evaluated constitutional challenges to the nondisclosure requirement in 18 U.S.C. § 2709(c) under strict scrutiny, even though the plaintiffs in that case also received the information at issue only from their involvement in confidential government investigations. NSL, 33 F.4th at 1071–72. 60 TWITTER, INC. V. GARLAND

NSL, 33 F.4th at 1076 n.21. Given that not even the Pentagon Papers per curiam majority clearly established the test advocated by Twitter, and given the material similarities between the nondisclosure requirements at issue in NSL and this case, the traditional form of strict scrutiny is the correct standard for evaluating the Government’s restrictions. 1. The Government’s Restrictions Satisfy Strict Scrutiny. Having determined that the traditional form of strict scrutiny applies, the next step is to determine whether the Government’s restrictions satisfy this heightened standard. “Under strict scrutiny, restrictions may be justified only if the government proves that they are narrowly tailored to serve compelling state interests.” Id. at 1070 (citation and internal quotation marks omitted). Both requirements are met here. First, the restrictions serve a compelling state interest. Both the regulatory and statutory nondisclosure frameworks at issue undisputedly operate to prevent the disclosure of the redacted information in Twitter’s Report for the purpose of national security. See E.O. 13,526; NSL, 33 F.4th at 1073 (“Here, the recipients do not dispute that the nondisclosure requirement directly serves the compelling state interest of national security ….”). Our court has “readily conclude[d] that national security is a compelling government interest. Indeed, … everyone agrees that the Government’s interest in combating terrorism is an urgent objective of the highest order.” NSL, 33 F.4th at 1072 (citation, internal alterations, and quotation marks omitted). “By the same token,” our court also has determined that “keeping sensitive information confidential in order to protect national security is a compelling government interest.” Id. Given that the

TWITTER, INC. V. GARLAND 61

Government’s restrictions undisputedly rest on national security interests, strict scrutiny’s first requirement is met here. The next step, therefore, is to “turn to the question [of] whether the [Government’s restrictions are] narrowly tailored.” Id. Even though a “restriction is not narrowly tailored if less restrictive alternatives would be at least as effective in achieving the legitimate purpose that the statute was enacted to serve,” our court has observed in this very context that strict scrutiny does not require the content-based restriction to be “perfectly tailored.” Id. (citation and internal quotation marks omitted). “Accordingly, a reviewing court should decline to wade into the swamp of calibrating the individual mechanisms of a restriction.” Id. (citation, internal quotation marks, and alterations omitted). My review is particularly informed in this context by the Supreme Court’s frequent admonition that courts must provide the “utmost deference” to Congress’s and the executive branch’s factual judgments pertaining to national security matters. See, e.g., Holder v. Humanitarian Law Project, 561 U.S. 1, 34, 36 (2010) (determining that Congress’s and the executive branch’s judgments on national security matters are “entitled to significant weight”); see also Dep’t of Navy v. Egan, 484 U.S. 518, 529– 30 (1988) (observing that “the courts have traditionally shown the utmost deference to Presidential responsibilities” regarding military and national affairs (internal quotations mark omitted)); CIA v. Sims, 471 U.S. 159, 180 (1985) (“[I]t is the responsibility of the [executive branch], not that of the judiciary, to weigh the variety of complex and subtle factors in determining whether disclosure of information may lead to [national security harm].”). In justifying its restrictions on speech in the national security context, the Government 62 TWITTER, INC. V. GARLAND

must provide “reasonable specificity” and “demonstrat[e] a logical connection between the deleted information and the reasons for classification.” Wilson v. CIA, 586 F.3d 171, 185 (2d Cir. 2009) (citation omitted). It need not, however, provide “detail, specific facts, and specific evidence,” nor “conclusively link all the pieces in the puzzle before [courts] grant weight to its empirical conclusions.” Humanitarian Law Project, 561 U.S. at 34–35 (citation internal quotation marks omitted) (rejecting as “dangerous” the dissent’s proposed requirement that the Government justify constraints on speech with detailed factual explanations). Given the “significant weight” a court must afford to the Government’s national security factual findings, I would hold that the Government’s unclassified declarations— specifically, the Unclassified Tabb Declaration— sufficiently demonstrate that the Government’s restrictions on Twitter’s speech are narrowly tailored. See Dep’t of Navy, 484 U.S. at 527. As discussed at length in that declaration, the Government only redacted various pieces of information that the USA FREEDOM Act did not exempt from preexisting non-disclosure requirements, that “would disclose specific numbers of orders received, including characterizing the numbers in fractions or percentages, and would break out particular types of process received.” “Information at a more granular level than described in the USA FREEDOM Act remains classified, because it would provide a roadmap to adversaries revealing the existence of or extent to which Government surveillance may be occurring at Twitter or providers like Twitter.” Unclassified Tabb Decl., ¶ 15; see also Sims, 471 U.S. at 176–77 (“A foreign government can learn a great deal about the [executive branch]’s activities by knowing the public sources of information that interest the [executive branch].

TWITTER, INC. V. GARLAND 63

The inquiries pursued by the [executive branch] can often tell our adversaries something that is of value to them.”). Specifically, disclosure of the “granular aggregate data” that Twitter seeks to publish “would assist adversaries in avoiding detection by and in carrying out hostile actions against the United States and its interests.” Unclassified Tabb Decl. ¶¶ 8 n.2, 9. Tabb further averred that the three restrictions to which Twitter objects—(1) no disclosure beyond permitted ranges; (2) beginning the lowest band with zero; and (3) reporting a band for every type of process received—“were designed specifically to minimize the harms that could reasonably be expected to result from disclosure” of aggregate national security process data. Id. at ¶ 17. As Tabb explained, limiting the disclosure of information to the reporting bands permitted by the Act allows the Government to conceal trends in collection over time, which prevents foreign adversaries from knowing which platforms are “safe” for use and obscures the Government’s evolving intelligence collection capabilities. See id. at ¶¶ 16–18. And starting the lowest bands at zero instead of one prevents foreign adversaries from ascertaining with any certainty whether the Government was, or has recently started, collecting from a given platform. Reporting at least the lowest band for all types of national security process similarly conceals the types of collection in which the Government is engaged on a given platform, from which adversaries can deduce information about the capabilities and limitations of the Government’s collection abilities. Id. at ¶¶ 17–23. If “[a]rmed with the kind of detailed information about Twitter’s receipt of national security process contained in Twitter’s draft … Report,” Tabb explained, “adversaries reasonably can be expected to take operational security 64 TWITTER, INC. V. GARLAND

measures to conceal their activities, alter their methods of communication to exploit secure channels of communication, or otherwise counter, thwart or frustrate efforts by the Government to collect foreign intelligence and to detect, obtain information about, or prevent or protect against threats to the national security.” Id. at ¶ 19. By way of the detailed Unclassified Tabb Declaration, I conclude that the Government has sufficiently “indicate[d] the nature of the apprehended harm” and provided ample bases demonstrating that “the link between disclosure and risk of harm is substantial” in the unclassified record before us. John Doe, Inc. v. Mukasey, 549 F.3d 861, 881 (2d Cir. 2008). While these bases may not “link all the pieces in the puzzle,” they are commensurate with the level of detail provided in affidavits that the Supreme Court has determined, in the national security context to suffice in supporting strict scrutiny. See Humanitarian Law Project, 561 U.S. at 29–33, 35. The Government’s restrictions also fall squarely within its pursuit of national security: the redactions are neither overinclusive, because they only target precisely the type of aggregate information that both the executive branch and Congress have deemed to pose a harm to national security if disclosed, nor underinclusive, because the statutory framework prevents any disclosure of national security process outside of 50 U.S.C. § 1874’s aggregate reporting bands. In other words, Twitter remains free to disclose anything it wants other than precisely the national security process information—including most (but not all) aggregate national security process data—that Congress and the executive branch have authoritatively concluded will compromise important national security interests. Our court must give strong deference to the Government’s factual findings on national security. Doing so, it is evident that the

TWITTER, INC. V. GARLAND 65

restrictions on Twitter’s speech are narrowly tailored to the compelling interest of protecting national security and safeguarding classified information. See Dep’t of Navy, 484 U.S. at 527. Twitter’s contrary arguments are unpersuasive. Relying on The Florida Star v. B.J.F., 491 U.S. 524, 540 (1989), it argues that the redactions do not satisfy strict scrutiny because the Government failed to conduct an “individualized inquiry” as to whether the redacted information should be disclosed. But the record reveals that the Government did, in fact, individually assess the harms that reasonably could result from the disclosure of the classified information. Indeed, in the Unclassified Tabb Declaration, Tabb repeatedly referred to the specific information redacted in Twitter’s Report when he concluded that the disclosure of that specific information would provide foreign adversaries “a clear picture not only of where the Government’s surveillance efforts are directed … but also of how its surveillance activities change over time, including when the Government initiates or expands surveillance capabilities or efforts involving providers or services that adversaries previously considered ‘safe.’” Unclassified Tabb Decl. ¶ 7. The record fatally undercuts this argument. Twitter’s remaining arguments lack merit. Twitter argues that the Government’s restrictions are not narrowly tailored because they lack “durational limitation.” But as already described, the statutory nondisclosure frameworks provide for judicial review, which includes a review of any durational limitations (or lack thereof). See 18 U.S.C. § 3511(b)(1)(C); 50 U.S.C. § 1881a(i)(4). The EO also provides that classification determinations automatically expire by default after 10 years. See EO §§ 1.5(a)–(d), 3.1(a), 3.5(a)–(c). Twitter also proffers various 66 TWITTER, INC. V. GARLAND

disagreements with the Government’s assessment that the disclosure of the redacted information in the Report would harm national security. But “[a]t bottom, [Twitter] simply disagree[s] with the considered judgment of Congress and the Executive” on their assessments regarding national security. Humanitarian Law Project, 561 U.S. at 36. “That judgment, however, is entitled to significant weight, and we have persuasive evidence before us to sustain it.” Id. Twitter’s factual disagreements with the Government’s national security assessments fail under the significant deference we must provide to the Government’s factual claims about national security risks. In sum, the Unclassified Tabb Declaration provides a sufficient rationale to determine that the Government’s restrictions survive strict scrutiny. 2. Freedman Does Not Apply. Having determined that the Government’s restrictions survive strict scrutiny, the majority then rightly considers Twitter’s argument that the Government’s restrictions present “the sort of content-based restriction on speech which must have the procedural safeguards identified by the Supreme Court in Freedman.” NSL, 33 F.4th at 1076 (citation omitted). In Freedman, the Supreme Court held that a statute prohibiting the exhibition of films prior to a censorship board’s approval constituted an invalid prior restraint. 380 U.S. at 60. In doing so, the Court established three “procedural safeguards designed to obviate the dangers of a censorship system.” Id. at 58. These safeguards include:

1. any restraint prior to judicial review can be imposed only for a specified brief

TWITTER, INC. V. GARLAND 67

period during which the status quo must be maintained; 2. expeditious judicial review of that decision must be available; and 3. the censor must bear the burden of going to court to suppress the speech and must bear the burden of proof once in court.

NSL, 33 F.4th at 1071 (citing Freedman and other cases applying Freedman). Since Freedman, our court has recognized that “[t]he Supreme Court has generally focused on two types of government schemes requiring [Freedman’s procedural] safeguards: censorship schemes and licensing schemes.” Id. at 1076. In NSL our court also observed that the same statutory nondisclosure requirement that comprises part of the same nondisclosure framework at issue in this case “does not resemble [the] government censorship and licensing schemes” that triggered Freedman’s procedural safeguards. Id. at 1077. Unlike the censorship scheme addressed in Freedman, 18 U.S.C. § 2709(c)—the statutory provision that prevents NSL recipients from disclosing the fact that they received such a request—only “prohibits the disclosure of a single, specific piece of information that was generated by the government: the fact that the government has requested information to assist in an investigation addressing sensitive national security concerns.” Id. (emphasis added). I’m not the first to observe that the concerns that animated Freedman arose in a very different context: the Second Circuit has similarly acknowledged that “§ 2709(c) limits certain speech in advance but is not a typical example of a regulation for which procedural safeguards are 68 TWITTER, INC. V. GARLAND

required.” Id. at 1076 (discussing John Doe, Inc., 549 F.3d at 876) (internal quotation marks omitted). Building on that thought, the NSL panel explained that, “unlike an exhibitor of movies, the recipient of a nondisclosure requirement did not intend to speak and was not subject to any administrative restraint on speaking prior to the Government’s issuance of an NSL.” Id. at 1077 (emphasis added) (citation, internal alterations, and quotation marks omitted). “Rather than resembling a censorship or licensing scheme, [18 U.S.C. § 2709(c)] is more similar to governmental confidentiality requirements that have been upheld by the courts.” Id. at 1078 (citing Butterworth v. Smith, 494 U.S. 624, 634–36 (1990) (upholding in part a law requiring witnesses to maintain the confidentiality of the grand jury process); Seattle Times Co. v. Rhinehart, 467 U.S. 20, 37 (1984) (upholding a restriction on disclosure of information obtained through pretrial discovery). 3 But the NSL panel stopped short of explicitly deciding “whether [18 U.S.C. § 2709(c)] must provide procedural safeguards,” because in that case the panel determined that the government had satisfied all the requisite procedural safeguards regardless of whether Freedman applied. NSL, 33 F.4th at 1079. NSL’s discussion regarding the inapplicability of Freedman’s procedural safeguards is well-reasoned, and it must govern here. Just like the nondisclosure provision at

3 In Butterworth, the Supreme Court declined to invalidate part of a state statute that prohibited a witness from disclosing the testimony of another witness—which the former witness only learned of through her participation in confidential government legal processes. See 494 U.S. at 632–36. Similarly, in Seattle Times Co., the Supreme Court upheld a restriction on the disclosure of information obtained through pretrial discovery—which, again, it only obtained through its participation in confidential procedures. 467 U.S. at 37.

TWITTER, INC. V. GARLAND 69

issue in NSL, and similar to the confidentiality requirements at issue in Butterworth and Seattle Times, the Government’s restrictions only prevent “the disclosure of … specific piece[s] of information … generated by the government: the fact that the government has requested information to assist in an investigation addressing sensitive national security concerns.” Id. at 1077. Specifically, in this case, the Government prevented the disclosure of information pertaining to whether and how often the Government compelled Twitter to produce various types of information about its users. See Unclassified Tabb Decl. ¶ 7. The nature of this Government-generated information is likely far more sensitive than information disclosed during civil discovery or grand jury proceedings. See Butterworth, 494 U.S. at 634–36; Seattle Times Co., 467 U.S. at 37. That neither Butterworth nor Seattle Times applied Freedman makes sense, given that confidentiality requirements pertaining to information gathered solely through participation in confidential government procedures do not pose the risk of “freewheeling censorship” that Freedman was designed to prevent. See NSL, 33 F.4th at 1077 (citation omitted); see also Se. Promotions, Ltd. v. Conrad, 420 U.S. 546, 559 (1975). Moreover, “unlike [the] exhibitor of movies” considered in Freedman, Twitter “‘did not intend to speak and was not subject to any administrative restraint on speaking prior to the Government’s issuance of [the national security process].’” NSL, 33 F.4th at 1077 (internal alterations omitted) (emphasis in original) (quoting John Doe, Inc., 549 F.3d at 880). This distinction holds true for the Supreme Court cases Twitter relies on in support of its argument that Freedman applies here. See, e.g., Vance v. Universal Amusement Co., 445 U.S. 308, 317 (1980) (per curiam) 70 TWITTER, INC. V. GARLAND

(determining that “the absence of any special safeguards governing the entry and review of orders restraining the exhibition of named or unnamed motion pictures … precludes the enforcement of these nuisance statutes against motion picture exhibitors” (emphasis added)); Nat’l Socialist Party of Am. v. Vill. of Skokie, 432 U.S. 43, 43–44 (1977) (per curiam) (determining that Freedman applied to an injunction prohibiting the “marching, walking or parading in the uniform of the National Socialist Party of America” (internal alterations omitted)); Bantam Books, Inc. v. Sullivan, 372 U.S. 58, 61– 62, 71 (1963) (determining that the procedures of a state commission, whereby it notified book distributors that certain books were “objectionable for sale, distribution or display to youths under 18 years of age” and reminded them of the commission’s “duty to recommend to the Attorney General prosecution of purveyors of obscenity,” were “radically deficient”). Twitter identifies no decision, and I am aware of none, where a court held that the Government may not prohibit the disclosure of classified information—let alone classified information obtained solely through participation in confidential government investigations—in the absence of Freedman’s procedures. Instead, “[r]ather than resembling a censorship or licensing scheme, [the Government restrictions here are] more similar to governmental confidentiality requirements that have been upheld by the courts.” NSL, 33 F.4th at 1078. In accordance with NSL’s well-reasoned rationale, I would conclude that Freedman’s procedural requirements do not apply here. Even if some process similar to that required by Freedman was required, the process Twitter received is not far removed from Freedman’s framework. Although

TWITTER, INC. V. GARLAND 71

Twitter initiated this lawsuit, nothing prevented it from seeking prompt judicial review in federal court of the Government’s decision prohibiting it from disclosing certain information about national security process. Cf. id. at 1080 (“Freedman focused on minimizing the burden to the film exhibitor to ‘seek judicial review’ of the state’s denial of a license; it did not focus on which party bore the initial burden. Here, the burden on a recipient is de minimis, as the recipient may seek judicial review simply by notifying the government that it so desires.” (internal citation omitted)). As in NSL, the judicial process available to Twitter, which it has apparently been able to utilize without too much difficulty, satisfies any Freedman-type requirements that might properly apply here. See id. at 1079–80 (concluding that various provisions of 18 U.S.C. § 3511 provided for the requisite “specified,” “brief,” and “expeditious” period of judicial review contemplated by Freedman); see also 50 U.S.C. § 1881a(i)(4) (permitting review of FISA orders by the Foreign Intelligence Surveillance Court). At the end of the day, even if Freedman’s procedural protections had applied to the Government’s restrictions and the parties had operated under that framework, it would not have materially changed the outcome of this case. The parties would still have become embroiled in a lawsuit regardless of who initiated it; they would still have raised the same legal arguments on the merits; intervening statutory developments would still have altered those arguments and delayed a final resolution; and the parties would still have proceeded to dispositive motions. 72 TWITTER, INC. V. GARLAND

3. Due Process Does Not Entitle Twitter’s Counsel to Classified Declarations. Lastly, I would conclude that procedural due process does not require that Twitter’s counsel be provided access to classified information. When assessing due process challenges that implicate national security interests, a court must “apply the balancing test set forth in Mathews v. Eldridge, 424 U.S. 319 (1976).” Al Haramain Islamic Found., Inc. v. U.S. Dep’t of Treasury, 686 F.3d 965, 979 (9th Cir. 2012) (partial citation omitted); see also Hamdi v. Rumsfeld, 542 U.S. 507, 528–29 (2004) (plurality) (determining that the Mathews balancing test provides the “ordinary mechanism that we use for balancing such serious competing interests” as due process rights and national security). “[T]o determine whether administrative procedures provided to protect a liberty or property interest are constitutionally sufficient,” Mathews instructs us to consider three factors:

First, the private interest that will be affected by the official action; second, the risk of an erroneous deprivation of such interest through the procedures used, and the probable value, if any, of additional or substitute procedural safeguards; and finally, the Government’s interest, including the function involved and the fiscal and administrative burdens that the additional or substitute procedural requirement would entail.

TWITTER, INC. V. GARLAND 73

Kashem v. Barr, 941 F.3d 358, 377–78 (9th Cir. 2019) (internal quotation marks omitted). And when due process claims implicate classified information:

Courts should adopt a case-by-case approach [in] determining what disclosure of classified information is required, considering, at a minimum, the nature and extent of the classified information, the nature and extent of the threat to national security, and the possible avenues available to allow the designated person to respond more effectively to the charges.

Id. at 382 (citation and internal quotation marks omitted). Applying the Mathews factors, Twitter asserts a general interest in adversarial proceedings “in order to effectively vindicate its First Amendment rights.” But Twitter’s private expressive interest here is relatively weak because, as I note above, Twitter seeks to disclose classified information the Government shared only as a necessary part of conducting national security investigations. When balanced against the Government’s compelling interest in national security, the relatively low risk of erroneous suppression under the carefully tailored nondisclosure regime, and the heavy burden of providing access to classified information to Twitter’s counsel, the due process balance weighs against disclosure here. Cf. Al Haramain Islamic Found., Inc., 686 F.3d at 979–80; Kashem, 941 F.3d at 378. Moreover, “even assuming cleared counsel were available to the plaintiffs and that it was error not to disclose the additional reasons to such counsel, [Twitter] ha[s] not shown that [it was] prejudiced.” Kashem, 941 F.3d at 383. And any prejudice argument 74 TWITTER, INC. V. GARLAND

would face a particularly steep uphill battle, given that I believe we easily could have, and indeed should have, decided this case on the unclassified record alone. No due process violation arises here.

II. CONCLUSION

The Government’s prevention of Twitter from publishing classified, redacted information satisfies strict scrutiny, and Freedman’s procedural protections do not apply in this case. Due process also does not demand that Twitter’s counsel obtained access to classified information. I therefore agree with the majority to affirm the district court.