The Constitution’s promises of freedom of speech and common defense can, at times, be at odds. One acute example of that tension is the prepublication review process, by which the government reviews written works by certain current and former employees to ensure that they do not contain classified or other sensitive information. While this process surely has its merits in preserving national security, it also presents authors with a bureaucratic thicket that is often difficult to navigate. This process is further complicated by the fact that the government can retroactively classify documents, meaning that information that authors might have thought was fair game is instead withdrawn from the public domain. The Supreme Court has addressed prepublication review only once, in Snepp v. United States. There, the Court validated the constitutionality of prepublication review but failed to articulate its reasoning in terms of established First Amendment doctrine. This Comment clarifies the standard of review applicable to prepublication review as an articulation of intermediate scrutiny.

Once that standard of review is established, this Comment applies it to the prepublication review process. With regard to substance, this Comment argues that, under intermediate scrutiny, the government does not have a sufficient national security justification to censor unclassified information during the prepublication review process. With regard to procedure, this Comment recommends that retroactive classification decisions during the prepublication review process should be subject to document-by-document review, that the burden-shifting framework to determine whether information is sufficiently public should begin by placing the onus on the government, and that authors’ legal claims arising from the process should not be mooted by completion of the review. Taken together, these clarifications and adjustments would subtly alter incentives to ensure that the prepublication review process equitably balances the interests of both the government and authors.



In the fall of 2019, John Bolton left his position as national security advisor to President Donald Trump after about seventeen months in the role. Bolton later recounted his experiences in his memoir, The Room Where It Happened. After acquiring a $2 million advance and drafting a 500-plus-page manuscript, he submitted the book to the White House for prepublication review.1 Prepublication review is the process by which government officials review some current and former employees’ writings, before they are published, to ensure that they do not contain sensitive information. The government found that Bolton’s first draft was “rife with classified information,” and he spent the next several months making modifications to mollify concerns.2 Due to these efforts, the White House official reviewing Bolton’s book “communicated to Bolton that she no longer considered the manuscript to contain classified material.”3 Despite these efforts, that informal near-approval was later overridden by official disapproval. Bolton went ahead with publication anyway, flouting his contractual obligations to put his book through the prepublication review process and prompting a lawsuit by the United States.4

Bolton’s experience with prepublication review was tumultuous but not unprecedented—or even uncommon.5 Another author described his own experience with the process as “contentious,” “fraught,” “inconsisten[t],” “lengthy,” and as inducing anxiety, belligerence, and self-censorship.6 In one particularly pointed email, a prepublication review official queried, “Why do you people insist on writing?”7 Nevertheless, the author said that he was “grateful” for the process insofar as it caught sensitive information that he had not fully understood was significant.8

These authors’ experiences illustrate the balancing act at play in the prepublication review process. Although the process has its benefits, its procedures often seem to be loaded against writers. This is partly because courts have done little to demarcate the boundaries of prepublication review. The Supreme Court’s sole decision on the topic, Snepp v. United States,9 dealt with a book published by a former CIA operative. The opinion neither identified the relevant First Amendment doctrines nor articulated the standard of review applicable to prepublication review. Instead, the Court rested its holding—that all proceeds from Snepp’s book would be disgorged to the government through a constructive trust—on the former agent’s breach of his fiduciary obligation to the CIA.10

Understanding these procedures, which filter the communication between intelligence officials and the public, is critical because officials’ writings are some of the only ways that the public gets a glimpse of the often shadowy intelligence community. An overly burdensome prepublication review process can discourage writings that are integral to the public’s ability to hold elected leaders accountable for activities that the public is unable to directly witness (because these activities happen behind closed doors or overseas).

This Comment makes two novel contributions to prepublication review scholarship. First, it identifies the standard of review applied to prepublication review by providing a comprehensive doctrinal review and parsing the language in Snepp. This analysis shows that prepublication review is subject to intermediate scrutiny.11 Second, this Comment places key elements of the prepublication review process under the intermediate scrutiny microscope. Based on that analysis, this Comment argues that unclassified information cannot be censored during prepublication review and makes a series of procedural recommendations that courts should apply. These changes would maintain the government’s substantial interests as an employer and in national security while reining in censorship techniques that are potentially not narrowly tailored to those interests.

Part I lays out the scope of executive authority over sensitive information. Part I.A describes one facet of that authority, the classification power. In particular, Part I.A focuses on retroactive classification, the process by which the government may classify information that was not classified when acquired. This power is important because it allows the executive to use the classification power reactively. That means that the power is an important fail-safe but also presents opportunities for abuse. Part I.B explains the rationale and procedures behind the prepublication review process. Part I.C puts forward the substantive, legitimacy, and logistical issues that arise when courts are faced with cases involving these powers. Part I.D argues that current executive authority and prepublication review jurisprudence create incentives that lead to a suboptimal cycle in which prepublication reviewers are overcautious, favoring even remote national security risks over speech. This pushes authors toward evasion, rendering the review process self-defeating and creating national security risks.

Part II distills the muddled and underdeveloped prepublication review doctrine. This discussion revolves around Snepp, the only Supreme Court case to deal directly with prepublication review. Part II outlines the three First Amendment doctrines that intersect in Snepp: unconstitutional conditions, government employee speech, and prior restraint. That Part then argues that, although the opinion did not explicitly articulate the standard of review that it employed, the Snepp Court was applying intermediate scrutiny.

Part III maps out this Comment’s recommendations. Part III.A describes why legislative solutions are less than ideal: they are hampered by both constitutional and political concerns. Part III.B outlines the constitutional limitations on executive actions regarding unclassified information. Part III.C charts a course for courts through a series of adjustments to procedure that would cabin the prepublication review process and retroactive classification power in accordance with intermediate scrutiny. These adjustments include document-by-document review of retroactive classification decisions that occur during the prepublication review process, flipping the burden-shifting framework to place the onus on the government because it is in possession of the information in question, and ensuring that objections to individual prepublication delays do not evade review through mootness.

I. Executive Authority to Control Sensitive Information

This Part describes the processes at play throughout this Comment—retroactive classification and prepublication review—and their treatment by courts. Section A describes the classification power generally because it is a critical input in the prepublication review process itself. Section B outlines the prepublication review process. Section C details courts’ typical treatment of cases involving these powers. Section D takes stock of the incentive structure created by the combination of broad executive powers over information and a prepublication review process relatively untouched by courts.

A. Classification

Congress has granted the director of national intelligence (DNI), an executive political appointee, broad authority to maximize intelligence gathering and protect intelligence sources.12 Pursuant to this authority, the president, vice president, agency heads, officials designated by the president, and some delegates of their authority are empowered to shield certain documents and information from public scrutiny.13 Executive Order 13,526 sets out specific reasons that information can be classified, but classification generally “shall not be considered . . . unless [the information’s] unauthorized disclosure could reasonably be expected to cause identifiable or describable damage to the national security.”14 In addition to this content-based restriction on what information is classifiable, there are procedural differences based on when the information is classified.

Original classification, in which the government classifies the information upon acquisition, is the typical process. Originally classified information is never in the public domain (the universe of information that is generally available “to any diligent seeker”)15 unless it is later declassified. Retroactive classification, in contrast, describes classification of information that is already in the public domain.

Retroactive classification comes in three forms: retroactive original classification, reclassification, and retroactive classification of inadvertently declassified documents.16 Retroactive original classification is the original classification process applied to information that was not classified when acquired.17 Reclassification entails classifying information that previously went through the formal declassification process; it is the only one of the three retroactive processes to face an additional hurdle beyond the requirements for original classification.18 Importantly, the information must be “reasonably recover[able] without bringing undue attention to the information” in order to be reclassified.19 Classification of inadvertently declassified information does not face the same additional hurdle because, intelligence officials argue, “the executive order refers to information declassified and released ‘under proper authority,’ so if the information was inadvertently declassified and released, it did not become public ‘under proper authority.’ Thus, the order’s limitations . . . do not apply.”20

Classification must be carried out pursuant to the criteria described in sections 1.1 and 1.4 of Executive Order 13,526.21 These sections describe the necessary conditions for classifying information, such as its importance to national security, relevance to specific subject matter, and so on.22 However, executive orders themselves may apply retroactively.23 This means not only that unclassified information may be retroactively classified but also that information that was unclassifiable at the time it was acquired can be retroactively classified. Specific classification decisions can be appealed to the Interagency Security Classification Appeals Panel, but documents undergoing prepublication review are specifically excluded from this appeals process.24 The only other time that the Executive Order mentions prepublication review is to exclude documents submitted during prepublication review from mandatory declassification review, the process by which nearly all classified documents are required to have their classified status reaffirmed or else be declassified.25 These ex-clusions are the result of the nondisclosure agreements that employees sign, which waive these procedures.

There are several formal restraints on the classification power. For example, classification powers may not be used to “conceal violations of law,” “prevent embarrassment,” or “delay the release of information that does not” meet the qualifications for classification.26 It is useful to keep in mind, though, that it is often difficult to divine the motivations behind a classification decision.27 Indeed, despite the formal classification requirements, “officials joke that ‘you could easily classify the ham sandwich.’”28 Estimates of the number of classified documents vary greatly, but even the lower end is in the billions.29 Of all classified documents, “somewhere between one-half of one percent and five percent of all classified information is in fact properly classified,” according to a former Department of Defense official.30 With so much information already classified—often improperly—and the classified universe constantly expanding, it has become increasingly difficult for writers on national security subjects to identify what terrain remains for them.

The classification power is enforced, at least in part, by the threat of criminal penalties posed by the Espionage Act.31 The Act criminalizes the knowing and willful communication (in any form, including publication) of four categories of classified information in a manner detrimental to the United States or beneficial to a foreign government.32 The four categories are broad, including “the communication intelligence activities of the United States or any foreign government”33 and any classified information “obtained by the processes of communication intelligence.”34

B. Prepublication Review

Prepublication review generally involves the “submission of information to an agency for the purpose of permitting such agency to examine, alter, excise, or otherwise edit or censor such information before it is publicly disclosed.”35 Many employees throughout the federal bureaucracy have signed contracts with prepublication review provisions,36 but intelligence officials’ agreements are, predictably, the most demanding and controversial.37

Federal employees who work with classified information are required to sign nondisclosure agreements that mandate prepublication review of future writings. There are two core form contracts. The first, Form 4414,38 requires that employees submit for review “any writing . . . including a work of fiction” that could implicate “Sensitive Compartmented Information” (SCI).39 SCI “is information about certain intelligence sources and methods . . . pertaining to sensitive collection systems, analytical processing, and targeting.”40 This contract applies indefinitely beyond the period of employment and requires that the author receive written authorization before publication.41

The second agreement, Form 312, requires employees not to disclose classified information unless they first receive official written approval.42 This is a fairly duplicative provision given that the Espionage Act criminally punishes identical conduct.43 What this contract adds is the requirement that the writer seek authorization if he is “uncertain about the classification status of information.”44 This clause has been interpreted as imposing a subjective burden; if the official himself feels any uncertainty, he is required to seek authorization.45 Seeking authorization has itself been interpreted to convey uncertainty.46 Part I.D explores the perverse incentives that this “certainty clause” creates.

These contracts also specify that the remedy in the event of a breach is a constructive trust, an instrument through which all the gains of an unauthorized work are disgorged to the government.47 The government can also seek criminal or other penalties on top of disgorgement.48 The internal reasoning of the contract (at least regarding the constructive trust) is that the classified information is the property of the United States,49 so the government is entitled to seize profits from its use. Alternatively, the confidence and “trust” that form the basis of the relationship between the employee and the government entitle the government to recover in trust.50

Given the highly sensitive nature of information handled by national security employees—and the accompanying property-rights and trust theories described above—courts agree that “the law would probably imply a secrecy agreement had there been no formally expressed agreement” because “[c]onfidentiality inheres in the situation and the relationship of the parties.”51 Of course, the contracts probably make enforcement simpler and put employees on notice of their obligations. Whether imposed by contract or inherent power, determining when contractual constraints on speech become unconstitutional is a complicated endeavor that this Comment undertakes in Part II.

Prepublication review procedures vary by agency, but the first review is usually conducted by a prepublication review board. The board is typically made up of officials with the proper security clearances from a variety of agency divisions.52 Yet the board goes beyond a bare search for classified information. Instead, the board is tasked with knowing “the difference between what truly is sensitive and what is not.”53 As such, the board’s decision-making criteria are often murky to those on the outside because they go beyond the strictures of the classification requirements in Executive Order 13,526. Part III.B argues that the leap from classified to sensitive information is not justifiable on national security grounds. Appeals within the agency are available but are rarely used.54

C. Judicial Deference and Procedural Protections

From the internment of persons of Japanese descent during World War II to the travel ban imposed on nationals from predominantly Muslim countries, courts have traditionally been reticent to question the national security justifications forwarded by the executive branch.55 In the prepublication review context, courts have described their deferential positions in a number of ways—for example, “de novo with deference”56 or “presumption of regularity”57 —but the lesson is largely the same.

There are three main justifications for this practice. First, there is a substantive concern that courts are merely unable to determine the importance of certain information to national security—and that it is not their place to do so under the Constitution’s separation of powers.58 Judges, as generalists, are “ill-equipped to become sufficiently steeped in foreign intelligence matters” such that they could understand how “[t]he significance of one item of information may [ ] depend upon knowledge of many other items of information.”59 Second, there are legitimacy concerns. If courts authorized speech that led to a national security crisis, public backlash could be immense, potentially attracting congressional reform or simply reducing trust in the judiciary.60 Third, there are logistical issues. Nearly all judges, clerks, lawyers, and other people who interact with the court system on a daily basis lack security clearances or other authorization to work with classified information. Court buildings are not secured in the same way as other facilities that handle classified information. This feeds the other two justifications; when courts cannot even see the information that is the subject of litigation, they have no choice but to defer to those who can.61

Because courts are often unwilling or unable to informedly question executive decision-making in this context, they instead do what courts do best—impose procedural requirements. Perhaps the most popular is a timing requirement. Outside the national security context, the Supreme Court has subjected pre-approval delays on speech to heightened scrutiny because what looks like bureaucratic delay can just as easily be strategic foot-dragging that ensures that the material never attracts a large audience.62 One of the first court opinions dealing with prepublication review stated that “the maximum period for responding after the submission of material for [prepublication review] approval should not exceed thirty days.”63 That guideline has become part of the SCI agreement, which promises that authors will receive “a response” from the prepublication review authority within thirty working days of receiving the writing.64

Yet even this basic deadline requirement is violable. Courts consistently rule that claims for damages arising from a review period longer than thirty days are mooted when the prepublication process is completed by the time litigation ends.65 In United States v. Bolton (Bolton II),66 the court found that the thirty-day deadline was met even when the government backtracked, after the deadline, on its initial finding that the work did not contain classified information.67 The court determined that the deadline had been met because Bolton received an “initial response.”68 But this interpretation defangs the requirement because the contracts that Bolton signed required official written approval before he could proceed with publication. Part III.C.3 recommends alterations to both the timing and mootness inquiries.

D. Incentives Created by Broad Executive Informational Powers and Unchecked Prepublication Review

It is worth taking stock of the incentive structure created by the existing executive informational powers and prepublication review doctrine. The status quo works to the detriment of all parties by breeding risks to national security, personal livelihoods, and fundamental liberties.

As it stands, there are perverse incentives encouraging an author to be the first mover. If the prepublication review process is perceived as a black hole from which no book returns, authors might gamble by forgoing prepublication review entirely or simply publishing before receiving final approval as long as they are confident that their work contains no classified information.69 Because the certainty clause in Form 312—which requires that an author seek official approval whenever she is uncertain about the classification status of something she is writing about—has been interpreted to be a subjective requirement,70 an author might be able to plausibly plead that she was sure that her work did not contain classified information. Indeed, Bolton might have been better off had he not sought approval at all because doing so was interpreted as conveying his own uncertainty, creating potential criminal and contractual implications about his state of mind.

One other workaround would be for potential authors to become leakers instead. This is slightly different insofar as the author would no longer be lending credibility to the speech by stamping her name on it.71 Much of that loss in credibility is made up by going through an outlet like the New York Times, as in New York Times Co. v. United States (The Pentagon Papers Case).72 Furthermore, the line between leaker and whistleblower—a designation that comes with its own protections—can be blurry.73 Regardless of its enforcement options, the government surely does not want to set up a system that incentivizes leaking.74

That said, the risks that an author incurs by evading prepublication review are extremely high. There are potential criminal penalties as well as the near guarantee that a court will impose a constructive trust. Even if authors would be willing to take the risk to raise their public profiles or because they believe that publishing their work is in the public interest, publishers may not be so keen. The prospect of being sucked into the prepublication void, deprived of financial returns, or subjected to criminal penalties could strongly disincentivize publishers from working with former national security officials. That would deprive the public of a valuable source of information.

Of course, there are still impactful works published by former intelligence officials published each year,75 and it is difficult to prove that there could have been more if not for prepublication review. In all likelihood, the John Boltons and Hillary Clintons of the world will continue to publish notable work in the face of the prepublication review process. It is those on the margin of the decision whether to write—because they are uncertain about the classification status of some information, because they are unable to create a media fuss when their work is rejected by prepublication reviewers, or because of any number of other reasons—who might forgo writing, despite being the ones who might be able to tell readers the most about the day-to-day operations of the national security apparatus. That allows national security discourse to be even more elite driven in a country that is increasingly skeptical of elites.76 A more prosaic view of intelligence work might restore trust,77 but it is discouraged by existing prepublication review procedures.

Returning to evasion, even if authors evade the process, the government has demonstrated little appetite for criminal prosecutions. The government has never sought to criminally prosecute someone who violated the prepublication review process, even if their work might have included classified information.78 It is easy to see why when one considers the optics of appearing to jail political dissidents. Courts tend to have the same immediate reaction. In United States v. Bolton (Bolton I),79 the court expressed incredulity when it denied the government’s request for an injunction: “For reasons that hardly need to be stated, the Court will not order a nationwide seizure and destruction of a political memoir.”80

This further reveals why prepublication review is more problematic than ex post enforcement mechanisms like firings or criminal prosecutions. Not only does it implicate prior restraint doctrine, but it avoids a sticky political accountability problem. Instead of pursuing controversial criminal penalties after the speech in a way that the Framers might have recognized,81 the government is able to avoid high-profile public scrutiny through civil suits that do not attract the same ire.

II. Clarifying the Supreme Court’s Prepublication Review Jurisprudence

Because courts say so little about national security law—and what they do say is so deferential—many questions involving prepublication review have jumbled or inchoate answers. This Part aims to contribute to the literature by elucidating the standard of review applied in Snepp—the foundational Supreme Court case on prepublication review—to clarify the doctrine that guides the use of the powers described in Part I. Although the Court did not state what standard of review it was applying, this Part argues that the case is best explained by a synthesis of the unconstitutional conditions, government employee speech, and prior restraint doctrines. Sections A, B, and C, in turn, explain each of these doctrines and their relevance to Snepp. Section D argues that the combination of these doctrines, on balance, led the Court to apply a version of intermediate scrutiny.

In 1977, Frank Snepp published Decent Interval, a book about “certain CIA activities in South Vietnam.”82 The book was based largely on Snepp’s two tours of duty in Vietnam while working for the CIA.83 Snepp had signed an agreement similar to the ones described in Part I.B, but he published his book without submitting it for prepublication review.84 The government did not seek an injunction because the book had already been published.85 Instead, it sought “redress through more commonly utilized reme-dies”—a constructive trust on Snepp’s “ill-gotten gains.”86

In Snepp, the Court stated that “[t]he Government has a compelling interest in protecting [ ] the secrecy of information important to our national security” and that the prepublication agreement was “a reasonable means for protecting” that interest.87 The Court ultimately upheld the district court’s decision to impose a constructive trust on Snepp’s book because the “remedy is the natural and customary consequence of a breach of trust.”88

The opinion’s exact holding and precedential value are unclear, however, because Snepp is characterized by muddled legal reasoning and procedural idiosyncrasies. With regard to legal reasoning, there are two main deficiencies. First, the Court did not address the intersecting First Amendment doctrines at issue. The unconstitutional conditions, government employee speech, and prior restraint doctrines are all implicated by the case, but the Court cited none of the seminal cases on these subjects. Instead, the Court asserted that its “cases make clear that . . . the CIA could . . . protect substantial government interests by imposing reasonable restrictions on employee activities that in other contexts might be protected by the First Amendment.”89 Second, the Court did not clearly identify or categorize its standard of review. Pinpointing a standard is difficult given the opinion’s “cavalier”90 usage of constitutional terms of art: the opinion variously describes the government’s interest as “vital,” “substantial,” and “compelling.”91 These two problems are, of course, related. Without identifying the standards of review in related First Amendment doctrines, the Court had little basis to define one for the pre-publication review context.

The procedural quirks of the case might have contributed to the opinion’s underdevelopment. In addition to the dissent’s objection that the majority had answered the wrong question presented, the Court also decided the case without the benefits of merits briefing or oral argument.92 As Archibald Cox, former solicitor general and Watergate special prosecutor, put it at the time: “One would have supposed that the extent of the government’s authority to silence its officials and employees and thereby deprive the public of access to information about government activity was not too obvious to deserve deliberate judicial consideration.”93

Some have argued that Snepp is merely a peculiar case that should be limited to its facts.94 This approach is misguided. First, the cat is out of the bag: Snepp has been cited in more than two hundred cases, including thirteen Supreme Court opinions.95 As it stands, though, the lower courts tend to look to Snepp on an ad hoc basis, with significant variability regarding its implications for First Amendment law.96 Second, even if Snepp “leaves room for fact-sensitive analyses,”97 the opinion can—and should—be reconciled with First Amendment doctrine. Eschewing the sole Supreme Court data point on prepublication review would leave courts flying blind and the executive largely unencumbered. Translating Snepp into familiar First Amendment doctrine would provide the lower courts with a more systematic framework that more equitably balances the interests of the government and speakers in the prepublication review process.

This Part makes the novel argument that Snepp’s precedential value can be clarified by recognizing its position at the intersection of three First Amendment doctrines: unconstitutional conditions, government employee speech, and prior restraint. Sections A, B, and C explain each of these doctrines and how they are implicated in Snepp. This First Amendment soup does not automatically yield a single standard of review, but Section D argues that the countervailing concerns of each of these doctrines help explain how the Court arrived at what is best characterized as intermediate scrutiny.

A. Unconstitutional Conditions Doctrine Slightly Favors the Employee

The unconstitutional conditions doctrine limits the government’s power to require that someone refrain from otherwise constitutionally protected expression in order to receive a government benefit. The doctrine evolved from the rights-privilege distinction. The rights-privilege distinction is the idea that the government can attach whatever conditions it pleases to a government benefit because it need not provide it in the first place.98

The simple rights-privilege distinction fell by the wayside as the role of government benefits in society grew.99 In Perry v. Sinderman,100 the Supreme Court adopted the modern view of the unconstitutional conditions doctrine, specifically in the free speech context. There, the Court dealt with a state college professor who alleged that he had been terminated because of his public criticism of the college’s administrative policies.101 Although not necessarily required for its holding, the Court declared that “even though the government may deny [ ] the benefit for any number of reasons, . . . [i]t may not deny a benefit to a person on a basis that infringes his constitutionally protected interests—especially, his interest in freedom of speech.”102 Despite the breadth of this statement, the scope of permissible conditions has come to be defined by something more like a balancing test than a per se rule.

While the rights-privilege bright line has been eliminated, it is still the case that government may impose some restrictions on constitutional rights as a condition to receive government benefits.103 When Snepp was decided, the predominant view of the doctrine was that it required courts to “balance competing public and private concerns,”104 which many state courts characterized as a “reasonable relationship” between the condition and the government interest.105 This standard was later refined to require that the condition have a “nexus” with and “rough proportionality” to the government’s legitimate interests.106 In other words, there must be “some sort of individualized determination that the required [condition] is related both in nature and extent to the impact of the proposed [private action].”107

The unconstitutional conditions doctrine has been criticized on the grounds that it is a contentless test holding on to too much of the rights-privileges view of state largesse.108 That may be so, and the doctrine has branched off to more specifically address the concerns that arise when the government aims to regulate the speech of its own employees, which the next Section describes. But the “reasonable relationship” framework begins to explain the background assumptions against which the Snepp Court was writing. The contract that Snepp signed conferred the benefit of employment, but that does not necessarily mean that the government could include any conditions it liked. Instead, the Court likely determined that the conditions bore a reasonable relationship—in kind and degree—to the government interests that were advanced by the condition. Although the government employee speech doctrine—the subject of the following Section—has developed into a standalone doctrine, it is best understood as a specific articulation of the unconstitutional conditions doctrine.

B. Government Employee Speech Doctrine Heavily Favors the Government in the National Security Context

The government employee speech doctrine is essentially a more specific articulation of the unconstitutional conditions doctrine. The government employee speech doctrine was articulated in three seminal cases: Pickering v. Board of Education,109 Connick v. Myers,110 and Waters v. Churchill.111 Each case reinforced the idea that the government’s power to limit employee speech is somewhat greater than its power to limit speech based on some other benefit and much greater than its power to limit the speech of the public writ large.

It is also worth noting at the outset that although the government’s interests as an employer and in national security are distinct, cases involving national security employees often consolidate the analysis by framing the question as about the “efficiency” of the employee in carrying out national security functions.112

Waters, although decided nearly fifteen years after Snepp, provides a simple two-step balancing framework that also fairly describes the holdings of Pickering (which predates Snepp) and Connick. For the employee speech to be protected, it must be “on a matter of public concern,” and “the employee’s interest in [speaking] must not be outweighed by . . . ‘the interest of the [s]tate, as an employer,’” in providing public services.113 This two-step approach is easier for the government to satisfy than typical review of speech restrictions. The precision principle, which requires that the government precisely define and specifically target only the speech that it is attempting to regulate, is absent.114 The harm and causation principles are also less rigorous.115 Rather than requiring a close causal nexus between the regulated speech and predicted harm, the Court gives “substantial weight to government employers’ reasonable predictions of disruption.”116

While courts permit significant restrictions on employee speech generally, the context in Snepp required an even more deferential analysis because it involved a former national security official. In particular, there are two critical questions drawn from Pickering and its progeny:117 Should the speaker be considered a member of the general public? Are personal loyalty and confidence important requisites for the job? The first is a threshold question; the government must show that it is restricting the speech of its employees qua employees to be entitled to the broader authority under this doctrine. Both questions also concern the extent of the government’s interests.

While a former employee is surely a more innocuous member of the general public than a current one, even that innocuity is questionable when, as a speaker, the former employee uses her former employment status as the basis for her credibility.118 This blurs the line between official and unofficial speech, which can have national security consequences such as misinterpretation by foreign leaders.119 Given that the confidentiality contracts also carry into retirement, the government has a plausible argument that there are certain job duties that extend beyond the period of employment, which brings the former employees closer to the realm of current employees, where the government has even more authority.120

As to the second question, the case law strongly supports the government’s need for confidence when dealing with its employees who handle confidential information.121 Furthermore, because of the need for discipline, highly hierarchical organizations that deal with public safety and national security are permitted even more leeway in restricting employee speech.122

Finally, in many cases, courts consider the interest of the public as an audience when evaluating the importance of the speech as a counterweight to the government’s interest.123 Such a justification is inapt in the national security context. When dealing with classified information, it is the very danger of the public getting information of extreme importance to national security that justifies the restriction. That is not to say that the public’s interest is unimportant, but this justification cannot advance the analysis because as the public’s interest increases with the importance of the information, the government’s interest will usually grow in roughly equal or greater measure. The government’s interest in national security directly concerns a public interest—unlike, say, the government’s interest as an employer, which requires more logical inferences to reveal a public impact—further complicating the inquiry. Relying on courts to identify the point at which the public interest in transparency exceeds the public interest in national security is a dangerous gambit—and one that courts are unlikely to take up.124 Indeed, courts have highly cabined the public’s “right to know,” if the right exists at all.125

The government’s power to impose conditions on employee speech is wide-ranging and is only made more so by the specific considerations in the national security context discussed above. The major factor cutting in favor of employee speech, though, is that prepublication review resembles a system of prior restraint.

C. Prior Restraint Doctrine Favors the Employee

Noticeably absent from the Snepp majority opinion is any discussion of prepublication review in light of the Court’s prior restraint doctrine. Prior restraints are what they sound like: government rules or orders that “forbid expression before it takes place.”126 There is a “heavy presumption” against prior restraints under the First Amendment;127 punishment after the fact, like that under the Espionage Act, does not face quite the same uphill battle. Even the narrowest original conceptions of the First Amendment were most concerned with prior restraint because it suppresses speech generation in the first place, contrary to the First Amendment’s general spirit of facilitating a marketplace of ideas.128

Even though prior restraints are strongly disfavored, they are not per se unconstitutional.129 Instead, prior restraint analysis requires an individualized evaluation of how the restraint operates in practice.130 In Bantam Books, Inc. v. Sullivan,131 for example, the Court invalidated the Rhode Island Commission to Encourage Morality in Youth, which notified the distributors of books and magazines when the Commission found its products “objectionable for sale, distribution or display to youths” and reported such “purveyors of obscenity” to the Attorney General.132 Although the Commission’s notices lacked the force of formal legal sanctions, the Court “look[ed] through forms to the substance” to determine that they “sufficiently inhibit[ed] the circulation of publications to warrant injunctive relief.”133 The scheme was held unconstitutional because the prior restraint did not come with the “most rigorous procedural safeguards” as “constitutionally required.”134 The Court noted that it had “tolerated such a system only where it operated under judicial superintendence and assured an almost immediate judicial determination of the validity of the restraint.”135 The Commission fell short because there was no judicial review, the publisher had no opportunity for a hearing, and both the Commission’s enabling statute and its criteria for determining “objectionableness” were “vague and uninformative.”136

Many of the same concerns were echoed in Freedman v. Maryland,137 which considered Maryland’s State Board of Censors, an organization that issued licenses to show films.138 There, the Court explained that a prior restraint is distinct from an ex post prosecution because a prior restraint puts the initial onus on the speaker and delays the involvement of a court.139 “And if it is made unduly onerous, by reason of delay or otherwise, to seek judicial review, the censor’s determination may in practice be final.”140 Building on Bantam’s judicial review requirement, Freedman’s contribution is to focus on the impact of delay, stating that “[a]ny [prior] restraint . . . must [ ] be limited to preservation of the status quo for the shortest fixed period compatible with sound judicial resolution.”141

These procedural critiques readily apply to prepublication review. Judicial review is hamstrung by the court’s inability to review relevant evidence.142 The DNI’s authority is derived from the broad and arguably vague mandate to “protect intelligence sources and methods from unauthorized disclosure.”143 There is some correspondence between the prepublication review authority and the author, but it is not much of a hearing, and authors are usually not privy to why certain information was removed.144 Finally, the process is certainly onerous and can lead to serious delays or complete withholding.145

As one might expect, though, this discussion is complicated by the long shadow of national security. One of the first discussions of prior restraint doctrine’s interaction with national security came in Near v. Minnesota ex rel. Olson.146 There, the Court struck down a state law that authorized “abatement” of any “malicious, scandalous and defamatory newspaper.”147 In dicta, however, the Court seemed to suggest that the circumstances might be different if national security were implicated. As examples of “exceptional cases” that might justify prior restraint, the Court stated that “[n]o one would question [ ] that a government might prevent actual obstruction to its recruiting service or the publication of the sailing dates of transports or the number and location of troops.”148 A version of this statement was subsequently put to the test—and essentially rejected—in the Pentagon Papers Case. There, the Court permitted the publication of a host of documents relating to a classified Defense Department study about American activities in Vietnam.149 Although this famous case determined that First Amendment concerns can outweigh national security ones, national security continues to be a relevant consideration when evaluating prior restraints.150

Beyond national security generally, there is some question about whether prepublication review is a system of prior restraint at all. Rather than enforcing the prepublication review obligation through an injunction in Snepp, the Court opted for a constructive trust.151 In other words, the book could be published, but the profits had to be disgorged. This was because the government was not seeking an injunction, which would have entailed finding and destroying copies of Snepp’s book. In practice, though, either mechanism is aimed at inducing adherence with the prepublication review regime. The difference in remedial formality would seem to be a thin distinction that the heavy presumption against prior restraint would not tolerate. Even if not, the Court also upheld the district court’s “injunction against future violations of Snepp’s prepublication obligation.”152

Snepp ultimately sidestepped prior restraint analysis by retreating to the freely signed contract that merely imposed prepublication review as a condition of employment. The D.C. Circuit has characterized the Supreme Court’s jurisprudence as permitting the government’s interests as an employer to “dominate the special concerns about prior restraints. This is especially so because Pickering can readily count those concerns in the course of the balance.”153 Although prior restraint comes with a heavy presumption against it, that presumption is rebutted somewhat by the national security and employment concerns presented above. It seems that in evaluating the interaction between the contractual conditions, the government’s interest as an employer, and the prior restraint, the Court balanced the doctrines and applied intermediate scrutiny.

D. Intermediate Scrutiny Balances These Three Doctrines

Although the Court may not have done so in a systematic way, Snepp seems to have mixed the above doctrines and arrived at intermediate scrutiny. The rough proportionality of unconstitutional conditions and the Waters balancing test applied to the government as employer require somewhat more than merely a rational link to a legitimate government interest, which characterizes rational basis review.154 But neither standard approaches the demands applied to prior restraint. The heavy presumption against prior restraint means that such restrictions are essentially subject to strict scrutiny,155 which requires that the law be the least restrictive means of furthering a compelling interest.156

The basic contours of intermediate scrutiny in the First Amendment context were laid out in United States v. O’Brien,157 in which David Paul O’Brien was convicted under a federal law that criminalized destruction of a draft card; he had burned his draft card during a public demonstration.158 The case stands for the proposition that content-neutral restrictions on speech are subject to intermediate scrutiny, and its multiprong test outlines the standard for intermediate scrutiny.159 First, a court must determine whether the law serves an important or substantial government interest.160 Second, the government interest must be “unrelated to the suppression of free expression.”161 Third, the restriction must be “no greater than is essential” to serve the government interest.162

The Court applied intermediate scrutiny in Brown v. Glines,163 which also involved national security and speech. Snepp was decided just a month later, and it cites Brown.164 In Brown, the Court dealt with a soldier on a domestic base. Albert Glines, a captain in the Air Force, was circulating a petition that he intended to send to congresspeople and the secretary of defense.165 While communication between soldiers and congresspeople was protected by a specific act of Congress, circulating a petition within one’s own base without the base commander’s approval violated Air Force regulations.166

Brown required that a regulation must “restrict speech no more than is reasonably necessary to protect the substantial government interest.”167 The Court upheld the restrictions on the soldier’s right to petition as “protect[ing] a substantial Government interest unrelated to the suppression of free expression.”168 And although the Court repeatedly emphasized the unique need for discipline on an army base,169 it also clarified that its holding extended beyond that narrow context.170 Snepp seems to import this reasoning, citing Brown for the proposition that Snepp’s contract was valid because it imposed a “reasonable restriction[ ]” to protect “substantial government interests.”171

Lower courts have also applied Brown in prepublication disputes.172 In McGehee v. Casey,173 another former CIA agent, Ralph McGehee, challenged the prepublication review process after the CIA reviewers determined that parts of an article that he sought to publish about CIA operations in El Salvador and elsewhere were secret.174 McGehee proceeded to challenge the constitutionality of using the “secret” classification category as grounds for censorship during prepublication review.175 Although the D.C. Circuit began its section entitled “The Constitutional Standard for Reviewing the Censorship Scheme” by recounting the facts and holding of Snepp, the court ultimately looked to Brown as the “best articulat[ion]” of the necessary relationship between the government interest and the speech restriction.176 The court held that censoring secret information during prepublication review was required to—and in that case did—serve a substantial government interest while being “narrowly drawn.”177 This language reinforces the idea that the standard of review for prepublication review is intermediate scrutiny.

Although at least one court has described the framework in Brown as intermediate scrutiny,178 Snepp has not received that label.179 Yet the language in Snepp is parallel to that in O’Brien and Brown in several respects: The Snepp Court speaks explicitly in terms of “substantial government interests.”180 The other two O’Brien requirements—that the government restriction be unrelated to free expression and no greater than essential—are somewhat more hidden but can be found.

For a restriction to be unrelated to suppressing speech, and thus be subject to intermediate rather than strict scrutiny, it must focus on the “noncommunicative aspect[s]” of the expression.181 In other words, it must not target speech based on its message or content. Defining a content-based restriction is notoriously tricky, but the Snepp Court avoided that inquiry entirely. Because the government had conceded that Snepp’s book did not contain any classified information,182 the Court distinguished between Snepp’s two obligations “not to divulge classified information and not to publish any information without prepublication clearance.”183 By grounding its holding in the latter requirement, the Court focused on the necessity of the prepublication review regime as a whole rather than the specific contents of Snepp’s book.184

As to the third requirement, although O’Brien’s “no greater than is essential”185 language may sound like the requirement associated with strict scrutiny, that cannot be right because the Court upheld the restriction based on a substantial—rather than compelling—government interest.186 Instead, O’Brien reveals that the tailoring required for intermediate scrutiny can run the gamut.187 In recent years, the Court has made clear that even in intermediate scrutiny First Amendment cases, the restriction “still must be ‘narrowly tailored to serve a significant governmental interest.’”188 But what constitutes narrow tailoring in the speech context ranges from requiring the least restrictive means to requiring a reasonable relationship between means and ends.189 Snepp seems to be toward the latter end of this spectrum, describing the prepublication review agreement as a “reasonable means” for protecting the government interest.190 This is consistent with the nature of intermediate scrutiny as something of a sliding scale. Because Snepp describes the government’s combined national security and employer interests in the prepublication review context as “compelling,” the Court required only that the system be a reasonable means of achieving those interests.191

Although the government has significant authority under intermediate scrutiny, the looseness of the standard as well as the lack of clear articulation in Snepp leave open the possibility of judicial adjustment when determining how tightly the government’s means must fit its ends. Although a version of intermediate scrutiny that emphasizes the government’s interests might tolerate a burdensome system of prepublication review, another version that focuses on the prior restraint aspect of the system might not. With a view to the precarity of this balance, Part III.C suggests procedural adjustments to ensure that prepublication review remains narrowly tailored to—or at least a “reasonable means” of achieving—the government’s substantial national security interests.

III. Applying Intermediate Scrutiny to Prepublication Review

This Part charts a course to keep prepublication review between the guardrails of intermediate scrutiny by more effectively balancing the interests of speakers and the government. While many scholars have recommended legislation, Section A suggests that congressional action that substantially limits executive freedom of movement in national intelligence might be unconstitutional in light of concerns about the separation of powers and the First Amendment. Section B then turns to the executive to apply the standard of review identified in Part II. Despite the branch’s broad authority in intelligence, that Section argues that the executive does not have a substantial national security interest in unclassified information. As such, unclassified information cannot be legitimately flagged for removal during the prepublication review process. Finally, Section C identifies procedural adjustments that courts could implement to bring prepublication review—particularly when leveraging the retroactive classification power—in line with intermediate scrutiny’s narrow tailoring and substantial interest requirements. Section C.1 recommends applying a heightened standard to retroactive classification decisions made during prepublication review. Section C.2 suggests restructuring burden shifting to put the responsibility for showing that information is classified on the government. Section C.3 argues for an exception to mootness for prepublication review claims to give substance to review deadlines.

A. Constitutional and Political Limitations on Legislative Action

Commentators have suggested a variety of congressional solutions, from essentially banning retroactive classification192 to strictly controlling prepublication review procedures.193 If Executive Order 13,526 is merely executing statutory commands, the thinking goes, then surely Congress would be able to control the executive’s actions in this realm. But this ignores the president’s constitutional authority. As long as courts agree with the president’s justification that prepublication review is critical to manage subordinates and national security intelligence operatives, the executive will have strong arguments that such procedures are squarely within the president’s Article II powers as commander in chief and head of the executive branch194 —even if the relevant statutory authority is hemmed in.

In fact, there was a bill proposed nearly forty years ago that sought to highly constrain prepublication review and similar executive restraints on intelligence officials.195 Of course, it is impossible to know exactly why the bill died in the House, but a witness at the subcommittee hearings expressed skepticism that the Supreme Court would permit Congress to oversee the executive in this way, given the deference frequently granted to the executive over national security affairs.196 If anything, the Court has become more favorable to executive power over the years, as two recent cases dealing with executive power over subordinates197 and the president’s personal documents198 demonstrate.

In addition to separation of powers concerns, there are First Amendment issues. The First Amendment only explicitly mentions Congress.199 Although the First Amendment has also been applied to the other branches, some courts have described the judiciary and executive as being subject to a more “flexible approach,” not only because they are not explicitly mentioned “but also because they lack legislative capacity to establish a pervasive system of censorship.”200 Admittedly, there are good reasons to doubt that Congress is better positioned than the executive to carry out a vast censorship regime201 —particularly of former executive officials. Still, this higher bar for congressional action is yet another reason why processes that potentially interfere with speech, like prepublication review, have long been kept in the executive branch.

Some commentators have also called on Congress to establish specialized courts to handle prepublication disputes,202 akin to those created by the Foreign Intelligence Surveillance Act of 1978203 (FISA) to handle surveillance warrant requests. Although potentially useful in the abstract, it seems unlikely that Congress will have any appetite for creating additional secretive, controversial courts in the near future—there is barely enough political will to keep the FISA courts intact.204

None of these critiques is a death knell to legislative action. Many proposals might still be worth pursuing as policy matters. They will have to overcome thorny constitutional and political problems, though, which is why this Comment emphasizes judicial decision-making.205

B. Constitutional Limits on Executive Power

Although there are substantial constitutional limits on the legislature’s ability to constrain the executive in both the national security and First Amendment contexts, the executive does not have free rein. Even though the Supreme Court did not squarely address the outer bounds of the executive’s compelling national security interest vis-à-vis the First Amendment in Snepp, this Section argues that the executive does not have a substantial national security interest in halting the publication of unclassified information.

1. Snepp leaves open whether the government may censor unclassified information during the prepublication review process.

Snepp states that prepublication review is constitutional even when the information under review is unclassified.206 This relates to the Court’s more sweeping claim that

[t]he problem is to ensure in advance, and by proper procedures, that information detrimental to national interest is not published. Without a dependable prepublication review procedure, no intelligence agency or responsible Government official could be assured that an employee privy to sensitive information might not conclude on his own—innocently or otherwise—that it should be disclosed to the world.207

This argument is closely related to the Court’s finding that Snepp’s evasion of the prepublication review process “exposed the classified information with which he had been entrusted to the risk of disclosure.”208 Of course, by the time the case reached the Court, there was no “risk” of disclosing classified information because the government had stipulated that Snepp’s book contained no classified information. Viewed from an ex ante perspective, though, the risk of disclosing classified information would surely be higher without authors going through the prepublication review process. As such, the Court was comfortable allowing a prepublication review system that subjected more harmless speech to regulation in order to potentially catch more harmful speech.

Whether unclassified information could be censored—not merely subject to review—is a question with which neither the Supreme Court nor the lower courts have fully grappled. Recall that Snepp’s holding is based on the idea that the government has a compelling interest in protecting national security.209 Snepp also states that “a former intelligence agent’s publication of unreviewed material relating to intelligence activities can be detrimental to vital national interests even if the published information is unclassified.”210 Despite the breadth of this statement, the lower courts have consistently asserted that “[t]he government has no legitimate interest in censoring unclassified materials,”211 strongly suggesting that the question was not foreclosed by Snepp. Indeed, given the Court’s focus on maintaining the efficacy of the prepublication review process as a whole, ad hoc censorship of unclassified information could be more damaging because authors might be incentivized to evade review if they perceive the process as broken.212

This divergence might also be explained by the perceived differences in the gestalt properties of intelligence in a particular case and who is in a position to understand those network effects. To the Snepp Court, the intelligence agencies, “with [their] broader understanding of what may expose classified information and confidential sources,” have the unique ability to determine whether information is detrimental to national security—even when it is unclassified.213 A similar phenomenon has been described by some lower courts, though not in the context of unclassified information.214 Perhaps the most famous case of this is United States v. Progressive, Inc.215 There, the court enjoined the publication of an article entitled “The H-Bomb Secret: How We Got It, Why We’re Telling It,” which assembled information necessary for building a nuclear weapon.216 Although that case is almost a caricature of equity balancing—free speech on the one hand versus nuclear annihilation on the other—courts generally consider even more banal threats to the confidentiality of intelligence sources to “endanger lives.”217

While the scope of the government’s authority to censor unclassified information remains a live question, the subsequent Section argues that such censorship would not be justifiable on national security grounds—even when multiple pieces of unclassified information are assembled to create a more sensitive work.

2. The government does not have a sufficiently substantial national security interest to halt the publication of unclassified information.

As the Supreme Court held in Snepp and Brown, suppressing government employees’ speech can sometimes be justified based on the compelling interest of national security.218 If the prepublication review system is to balance speech rights and national security in a manner that satisfies intermediate scrutiny, though, unclassified information cannot constitute a substantial national security interest. The retroactive classification power makes drawing this bright line possible.219 As long as the government has the power to classify information that it finds particularly threatening while a work is under prepublication review, there is no national security reason to allow government to withhold information that has not been put through the classification wringer.

Determining whether the government has a national security justification matters because Snepp’s somewhat relaxed tailoring requirement, “reasonable means,” is based in part on the government’s compelling interest in national security. Without a compelling national security justification, the presumption against prior restraint would probably overwhelm the government’s other interests, meaning that the government would have to rely on remedies after the speech (using the Espionage Act, for example) rather than preventing its utterance. Even if the government’s other interests as an employer (such as maintaining workplace harmony) were found to be sufficient to justify removing some unclassified information, that removal would at least need to meet a more demanding version of intermediate scrutiny that would require much narrower tailoring.

When describing its own requirements, the Office of the DNI says that unclassified information is not automatically publishable because, “[f]or example, privacy information, contractual information, or even unclassified sources and methods would likely be sensitive and therefore not publicly releasable.”220 Yet this loses sight of the purpose of prepublication review. Privacy and contractual information could be sensitive for individuals—imagine credit card numbers or leaked emails—but unless there is a strong link between that personal information and national security, there would not be a substantial national security interest in withholding that information. Instead, it should be the job of authors and publishers to ensure that their work does not unnecessarily expose personal information. The Constitution leaves the regulation of that kind of speech to the public. The backlash to the WikiLeaks info dump that included personal information demonstrates at least some of the wisdom of that approach.221 This is also a prime area to limit prior restraint in favor of ex post punishments. There are other legal avenues, namely tort law, that enable recovery against someone who discloses another’s personal information in a way that causes harm.222

This analysis is complicated somewhat by the fact that censorship of this kind is, generally, “unrelated” to the suppression of free speech, as the standard is articulated in O’Brien and Brown.223 Although there might not be an especially substantial government interest, the argument goes, the weightiness on the other side is even less. There is little to no value in an author including someone else’s sensitive personal information if it is not pertinent to the author’s purpose. That said, the government still does not have a substantial national security interest in suppressing the information. Again, the government might be able to assert its interests as an employer, but those are divorced from the national security foundations of prepublication review.

Furthermore, if information of this kind is truly sensitive enough to withhold, then it is sensitive enough to classify. Recall the descriptions from intelligence officials in Part I.A about the capaciousness of the classification power. Given the power to classify a broad array of sensitive information at practically any time, there must be a counterweight that preserves some safe harbor for speech. The liminal space between classified and sensitive information ought to be preserved for public speakers to prevent arbitrary censorship. This is bolstered by the fact that the government retains the power to pursue criminal prosecutions after the speech. Assuming that prepublication review is at least a level up from ex post prosecutions—even if not subject to strict scrutiny as a full-on prior restraint—there needs to be some reason why after-the-fact enforcement is not enough. Some justifications might exist, but national security cannot be one because the classification power sufficiently serves that interest.

A natural objection to this approach is that it would incentivize overclassification. Of course, just about everyone agrees that overclassification also describes the status quo, and there are plans in motion to reduce it.224 Additionally, the panoply of non–national security interests and the fail-safe of retroactive classification mean that the government would still be able to protect its informational interests without expanding original classification. Put another way, the problem of overclassification rises and falls largely independently of the prepublication review process. Taking the problem of overclassification as given, prohibiting prepublication reviewers from censoring outside the classified realm at least provides some bright-line check that could reassure authors.

Whether to permit suppression of unclassified information becomes especially fraught when considering censorship decisions based solely on the assembly of multiple unclassified pieces of information. According to Executive Order 13,526, the standard for classifying “[c]ompilations of items of information that are individually unclassified” requires that “the compiled information reveals an additional association or relationship” that is itself (1) classifiable and (2) not revealed by the individual pieces of information.225

In order to evaluate the threat to national security posed by a given work and determine whether the government has a substantial interest in its suppression, one ought to evaluate the information’s marginal effects. When a work reveals previously classified information, it provides a high value to adversaries because the value of that information was zero prior to its disclosure (assuming that classified information is inaccessible). When a work compiles unclassified information, however, the gain to the adversary is not nearly as high because the work merely eliminates the cost of collecting the already public information. It can be helpful to think of these as transaction costs. Rather than making a transaction possible by speaking into existence a new piece of information, the author is, at best, a broker between the adversary and the information.

The Progressive court considered its decision to enjoin publication as something approaching a no-brainer because of the sheer magnitude of the risk of nuclear proliferation.226 The marginal value of censorship in the case, however, demonstrates the inadequacy of the government’s interest in most compilation cases. It is difficult to imagine the middle of the Venn diagram that includes adversaries who are technically capable of building a nuclear weapon, willing to devote the necessary resources to its construction, and unable to assemble a series of publicly available documents on their own.227

C. Judicial Adjustments to Keep Prepublication Review Within the Bounds of Intermediate Scrutiny

With the limits on Congress and the executive in hand, a significant role remains for courts. While judges are rightly reticent to question national security justifications, there are many decisions within courts’ power and expertise that would alter the basic drumbeat of litigation. One should keep in mind that the doctrine in this area is quite limited. That has two implications: First, it means that changes would not upset deep-seated practices. Second, changes in just one or two cases could meaningfully alter conditions on the ground given that both the agencies and publishers are repeat players.

Although judicial action, like legislative action, implicates the separation of powers, the procedural adjustments recommended in this Part are distinguishable from the limits on legislative power discussed in Part III.A. The judicial changes would be instituted to ensure that the executive is abiding by the requirements of the First Amendment (albeit, under layers of judge-made doctrine). Congress, on the other hand, would be attempting to exercise some of its own constitutional authority, which is checked and balanced by similar or overlapping powers in the executive. Furthermore, the recommendations below are relatively modest procedural changes that are part of the judiciary’s unique expertise. Finally, as a practical matter, courts may simply be less likely to view judicial exercises of power in the name of the Constitution as violations of the separation of powers as compared to legislative action;228 unlike with Congress, there is essentially no one looking over the judiciary’s shoulder.

This Section recommends three changes that are basically procedural. Because providing adequate procedural safeguards aligns with the Court’s typical approach to prior restraints,229 courts should be comfortable implementing them. The basic purpose is to ensure that prepublication review and retroactive classification stay narrowly tailored to—or at least a reasonable means of protecting—substantial national security interests.230 First, retroactive decisions during prepublication review should be subject to a more demanding process. Luckily, the executive has one ready-made—document-by-document review. Second, courts should alter the burden-shifting framework to resemble other areas of law and to increase accountability by placing the onus on the government because it possesses the information. Third, courts should except prepublication review suits from mootness based on capability of repetition. Together, these suggestions retool judicial review of prepublication review to create somewhat more friction in the government’s decisions to censor during prepublication review, subtly shifting the balance of power toward authors.

1. Courts should subject retroactive classification decisions during prepublication review to document-by-document review.

The basic premise of this Comment is that if a speaker attempts to say something that is not classified, and then—once the author herself has alerted the government to her intention to say the thing—the government classifies the item, we ought to be skeptical of the government’s decision. Yet skepticism does not always mean invalidation. Instead, courts are well-equipped to deal in these gray areas, attempting to balance the equities. Helpfully, Executive Order 13,526 articulates a somewhat heightened procedure for classification decisions made after a Freedom of Information Act231 (FOIA) request:

Information that has not previously been disclosed to the public under proper authority may be classified or reclassified after an agency has received a request for it under [FOIA or a number of other means] only if such classification meets the requirements of this order and is accomplished on a document-by-document basis with the personal participation or under the direction of the agency head, the deputy agency head, or the senior agency official designated under section 5.4 of this order.232

The reasons that different procedures would apply in this context are highly analogous to those that apply to the prepublication review process. When a specific piece of information has been requested by a citizen under the Acts listed in the Executive Order, there are only a handful of avenues through which the government may say no. Classification is one of these. As such, when the government does not want to disclose a piece of information, there is a risk it will fall back on a procedure that is supposed to be the exception, not the rule.

Courts could require document-by-document review for instances in which information is retroactively classified and, as such, censored during the prepublication review process. The court could deem the extra procedure necessary to ensure that prepublication review remains narrowly tailored to the government’s interests. Courts have imposed similar procedural obligations in prepublication review disputes, such as the thirty-day requirement,233 right to retain counsel, and other “[r]easonable [p]rocedures.”234

That said, this process is less than satisfying because there is still no neutral, third-party oversight. The stricter review process here is simply applied by a higher-up with the same boss—the president. Yet without the development of specialized courts to overcome the logistical challenges described in Part I.C, it will require an internal official with the necessary clearances to review the information. Each intelligence agency has an inspector general, general counsel, and managing director who could be empowered to ensure adherence. The individual or body already tasked with prepublication review appeals could also manage this additional process.

This standard is no panacea, but it subtly alters incentives by imposing costs on sloppy agency action. At the very least, the requirement that a superior approve classification on a document-by-document basis should deter unnecessary classifications. Such review is time intensive. This would discourage both minor, overly cautious classification decisions as well as those that sweep in an entire book. In both instances, officials would be incentivized to classify less—or at least more precisely—because they would know that any retroactive decisions would be scrutinized. Attempts to strike at the well-protected heart of the executive’s control of national security are unlikely to get far. Pesky procedural requirements like these, however, can impact substantive decision-making on the margins.235

2. Courts should restructure burden shifting to put the onus on the government because it possesses the information.

As it stands, courts require the censored speaker to prove that the information she seeks to publish is in the public domain.236 The D.C. Circuit noted in McGehee that “[t]he CIA cannot reasonably bear the burden of conducting an exhaustive search to prove that a given piece of information is not published anywhere.”237 This is undoubtedly true, but it frames the problem in the wrong way. Classified information is not the default. As outlined in Part I.A, the executive must jump through hoops to classify information. Assuming that the information has not been formally classified, the speaker—subject to the discussion in Part III.B—has the right to say it.

While the universe of classified information is extensive, it is finite—unlike the constant stream of speech that enters the public domain. This is even clearer in the context of retroactive classification. If the government is attempting to classify a piece of information now, it is conceding that it is not already in the classified universe. As such, courts ought to force the government to first prove that the attempted speech is classified. Then, the author could attempt to prove that the information is sufficiently public such that censorship is improper. Such a procedure would be akin to other information-forcing rules based on who has better access to the information, such as res ipsa loquitur in tort law.238

There are complications that accompany this approach, though. First, courts attempting to verify that a document is properly classified are subject to logistical problems arising from reviewing classified information at all. As such, courts would have to rely on internal agency procedures. Just like with the
document-by-document procedure described above, courts could at least require some kind of document number or other identifier that has been reviewed by the agency’s general counsel or the like.239 Being able to link a document number to a later-declassified document could change policy over the long term. Although this might not present an especially satisfying remedy for the author given the extremely long timelines associated with declassification, it could create a paper trail so that future publishers, writers, academics, elected officials, and courts could study how prepublication review and retroactive classification are used in practice.

A second potential problem for a change in burden shifting is that courts have been deferential on the seemingly straightforward question of whether something is in the public domain. This is critical because Executive Order 13,526 requires that “the information [ ] be reasonably recover[able] without bringing undue attention to the information” in order to classify information that has already entered the public domain.240 This is an especially high bar to satisfy in the Internet Age.241 The recoverability limitation is only meaningful, though, if courts are comfortable determining the extent to which information has entered the public domain.

In some cases, that can be fairly easy—Bolton’s publisher had already distributed hundreds of thousands of physical and digital copies of his book to reviewers, booksellers, and the like.242 In Alfred A. Knopf, Inc. v. Colby,243 however, the court deferred to the CIA on the question whether a certain piece of information was “so widely circulated and [ ] so generally believed to be true, that confirmation by one in a position to know would add nothing to its weight.”244 These sorts of determinations ought to be squarely within a court’s wheelhouse. In fact, the intelligence agencies might have a less reasonable view of whether a particular piece of information is in the public domain because they could read too much into a single piece of information with the entire web of classified information in mind. Courts, on the other hand, have no choice but to look at information only in the context of the unclassified universe. If an author can point to certain public sources, a judge is well positioned as a generalist to determine whether it is reasonable to think of the information as in the public domain. This would also remedy the perverse incentives created by only considering something to be in the public domain if an author forces it there, as in Bolton.

3. Courts should give teeth to prepublication review deadlines by finding that violations are capable of repetition yet evade review.

As noted in Part I.C, the Supreme Court has recognized that delay can effectively quash speech: it can deter writers from seeking to publish, derail writers and publishers along the way, or simply make the writing untimely. As such, requirements for timely review by prepublication review boards ought to be robustly enforced by courts. Courts have typically drawn the line at thirty days, and the Office of the DNI has articulated that as a guideline too.245

The problem, though, is that authors have little recourse to challenge delays. In at least two cases, authors had their claims under the Administrative Procedure Act246 (APA) mooted because their prepublication reviews had been completed by the time the district court had reached a decision.247 Both authors alleged that the reviewing agencies had acted arbitrarily and capriciously, in violation of the APA, by failing to issue a prepublication review decision in a timely manner.248

Mootness is the doctrine that prevents a court from dealing with issues that it can no longer remedy. It is closely tied to standing doctrine, which requires that the plaintiff have suffered an injury in fact that is fairly traceable to the defendant and will be redressed by a favorable decision.249 These doctrines exist because Article III of the Constitution requires that courts hear only “Cases” and “Controversies,” which courts have interpreted to mean only live disputes.250

There are narrow exceptions to mootness, however, which courts should apply to prepublication review litigation in order to make timing deadlines meaningful. It would be an ironic disservice of justice to allow executive delay to slide only because the judiciary is even slower. The relevant mootness exception here would be that deadline violations are “capable of repetition, yet evad[e] review.”251 In the two cases in which the claims were mooted, the courts found that prepublication review disputes did not meet the standard for the repetition exception, which requires both exceptional circumstances and that a named plaintiff make a reasonable showing that he will again be subject to the illegal conduct.252 More succinctly, the courts determined that they could provide “no further relief" once the agency had made a final prepublication review decision.253

The first requirement, that the situation is exceptional, can be a slippery concept. Lawsuits involving abortion are perhaps the most prominent examples.254 These situations are exceptional because the circumstances giving rise to the action “exist only for a short, fixed time period and [ ] may be over by the time the litigation” reaches the appellate stages.255 The same logic maps onto the prepublication review process with perhaps even greater force because the government could furtively control the pace of the prepublication review in accordance with the pace of litigation.256 Constitutional concerns are also implicated in both contexts—Fourteenth Amendment substantive due process there257 and First Amendment speech here.

The second requirement, that a named plaintiff make a reasonable showing of potential for repeat harm, would also be met as long as the plaintiff plans to write more than one piece on the same subject of dispute. Much of the plaintiff’s writing is presumably going to be on her area of expertise, and even if she might not write multiple treatises on the subject, op-eds and other writings are also within the ambit of prepublication review. Speedy review of multiple pieces is particularly important when the time window in which the plaintiff’s knowledge is pertinent to public discourse is narrow, as often happens with national security topics. In sum, the court would be providing relief insofar as it would be laying out the requirements for future prepublication reviews. Although this might normally sound in the register of advisory opinions, the exception outlined in this Section provides an outlet to avoid mootness and provide prospective relief.

Still, courts are understandably uncomfortable with holding prepublication review boards to strict deadlines for fear that rushing the process might let some critical piece of information slip through the cracks. Indeed, there is a delay built into Executive Order 13,526 to deal with this possibility and bolster retroactive classification. The Order requires officials who are not vested with classification authority to notify those who are when the officials come across information that they believe should be classified.258 This provision requires this classification decision alone to be made within thirty days,259 and then the work must still progress through the rest of the prepublication review process. Instead, prepublication review boards should be vested with original classification authority as long as any retroactive decisions are reviewed according to the procedure outlined in Part III.C.1. Collapsing these bureaucratic layers should expedite the process.


Courts and commentators have recognized the tension between national security and free speech since the Founding, but national security has gained significant territory at the expense of speech in recent decades. The executive has obtained broad authority over national intelligence information, including the classification and prepublication review powers. The judiciary is nearly always deferential to these judgments—and for good reason. Logistical and constitutional constraints on the judiciary leave it little choice but to defer to the executive. That said, this Comment has outlined ways in which the judiciary can bring the prepublication review process into a more equitable balance with First Amendment speech protections in order to satisfy intermediate scrutiny—particularly when the unique risks of retroactive classification are involved.

Although Snepp (the Supreme Court’s sole case on prepublication review) is abstruse, the standard of review essentially resembles intermediate scrutiny. Extensive powers combined with a fairly flexible standard of review incentivize behavior that is inimical to the goals of all parties. Unfortunately, Congress has not shown interest in this area for over three decades; even if it were so inclined, the degree to which it could constitutionally intervene is unclear. That is not to say executive power is unbounded—the branch should at least be unable to censor unclassified information based on national security justifications.

Navigating the underdeveloped doctrine, courts have an important role to play in erecting procedural guardrails to narrowly tailor executive authority to substantial interests, as required by intermediate scrutiny. Courts ought to ensure that retroactive classification is not frictionless and that the prepublication review process proceeds according to clear rules. These include document-by-document review of retroactive classification decisions during prepublication review, restructuring burden shifting for determining whether information is in the public domain, and excepting prepublication review litigation from mootness.

  • 1First Amended Complaint at 1, United States v. Bolton (Bolton I), 468 F. Supp. 3d 1 (D.D.C. 2020) (order denying temporary restraining order and preliminary injunction) (No. 20-cv-01580).
  • 2Id. at 1–2.
  • 3Bolton I, 468 F. Supp. 3d at 3.
  • 4First Amended Complaint, supra note 1, at 17–25.
  • 5See, e.g., Natasha Bertrand, ‘I’ve Had It Take Years’: Bolton’s Book Could Be Tied Up past November, Politico (Jan. 29, 2020), (noting that the prepublication review process may be drawn out when politically contentious material is under review); Kevin Casey, Note, Till Death Do Us Part: Prepublication Review in the Intelligence Community, 115 Colum. L. Rev. 417, 445–49 (2015).
  • 6Thomas Reed Willemain, A Personal Tale of Prepublication Review, Lawfare Blog (Jan. 10, 2017),
  • 7Id.
  • 8Id.
  • 9444 U.S. 507 (1980) (per curiam).
  • 10Id. at 510.
  • 11For an explanation of intermediate scrutiny, see infra Part II.D.
  • 12See 50 U.S.C. § 3024.
  • 13Exec. Order No. 13,526, 3 C.F.R. 298, 298–300 (2010).
  • 143 C.F.R. 300.
  • 15United States v. Progressive, Inc., 467 F. Supp. 990, 993 (W.D. Wis. 1979).
  • 16See Jonathan Abel, Do You Have to Keep the Government’s Secrets? Retroactively Classified Documents, the First Amendment, and the Power to Make Secrets out of the Public Record, 163 U. Pa. L. Rev. 1037, 1053–59 (2015).
  • 17See id. at 1056–58.
  • 18See 3 C.F.R. 302–03.
  • 193 C.F.R. 302.
  • 20Abel, supra note 16, at 1058 (quoting 3 C.F.R. 302–03).
  • 213 C.F.R. 298, 300.
  • 22Id.
  • 23See Afshar v. Dep’t of State, 702 F.2d 1125, 1136–37 (D.C. Cir. 1983) (justifying the court’s application of a classification standard promulgated as part of an executive order that was signed while the case was on appeal by pointing to the executive’s need to respond quickly to changing national security interests).
  • 24See 3 C.F.R. 298, 303.
  • 25See 3 C.F.R. 311–12.
  • 263 C.F.R. 302.
  • 27Cf. Geoffrey R. Stone, Perilous Times: Free Speech in Wartime 75 (2004) (pointing to the Sedition Act of 1798 to illustrate that “[s]uppressing speech because it is dangerous to the national interest is one thing; suppressing it because it threatens a partisan interest is something else entirely. . . . [I]t is often difficult to tell the difference.”).
  • 28Abel, supra note 16, at 1057 (quoting Telephone Interview with William J. Bosanko, Chief Operating Officer, Nat’l Archives & Recs. Admin., Former Dir., Info. Sec. Oversight Off. (Nov. 1, 2013)).
  • 29Id. at 1049 (citing Peter Galison, Removing Knowledge, 31 Critical Inquiry 229, 230–31 (2004)) (noting that experts believe the government possesses between four billion and one trillion classified documents).
  • 30Susan Nevelow Mart & Tom Ginsburg, [Dis-]Informing the People’s Discretion: Judicial Deference Under the National Security Exemption of the Freedom of Information Act, 66 Admin. L. Rev. 725, 753 (2014) (citing U.S. Government Information Policies and Practices—Security Classification Problems Involving Subsection (b)(1) of the Freedom of Information Act (Part 7): Hearings Before the Subcomm. of the H. Comm. on Gov’t Operations, 92d Cong. 2296 (1972) (statement of Arthur Schlesinger, Jr., Albert Schweitzer Professor of Humanities, City Univ. of N.Y.) (recounting statistics provided by William Florence, a former Department of Defense official)).
  • 31Pub. L. No. 65-24, 40 Stat. 217 (codified as amended in scattered sections of 18, 22, and 50 U.S.C.).
  • 32See 18 U.S.C. § 798(a).
  • 3318 U.S.C. § 798(a)(3).
  • 3418 U.S.C. § 798(a)(4).
  • 35Federal Polygraph Limitation and Anti-Censorship Act of 1984, H.R. 4681, 98th Cong. § 7361(6); see also Edgar v. Haines, 2 F.4th 298, 304 (4th Cir. 2021) (“Prepublication review [ ] requires current and former employees to submit materials intended for publication to their agencies to enable the agencies to redact, in advance of publication, classified or otherwise sensitive information.” (quotation marks omitted)).
  • 36See, e.g., Alex Abdo, Jameel Jaffer, Meenakshi Krishnan & Ramya Krishnan, How a New Administration—and a New Congress—Can Fix Prepublication Review: A Roadmap for Reform, Just Sec. (Nov. 24, 2020), (“[S]ubmission requirements vary considerably by agency, and they are imposed through a confusing and sometimes conflicting tangle of contracts, regulations, and policies.”).
  • 37See Casey, supra note 5, at 430–51; cf. infra Part I.D (discussing former national security officials’ avoidance of prepublication review).
  • 38Nat’l Counterintel. & Sec. Ctr., Sensitive Compartmented Information Nondisclosure Agreement (2013),
  • 39Id. ¶ 4.
  • 40U.S. Dep’t of Com., Access to Sensitive Compartmented Information (SCI),
  • 41Nat’l Counterintel. & Sec. Ctr., supra note 38, ¶ 4.
  • 42Off. of the Dir. of Nat’l Intel., Classified Information Nondisclosure Agreement (2013),
  • 43See supra Part I.A.
  • 44Off. of the Dir. of Nat’l Intel., supra note 42, ¶ 3.
  • 45See United States v. Bolton (Bolton II), 496 F. Supp. 3d 146, 160–61 (D.D.C. 2020) (order denying motion to dismiss).
  • 46See id.
  • 47See Off. of the Dir. of Nat’l Intel., supra note 42, ¶ 5 (“I hereby assign to the United States Government all royalties, remunerations, and emoluments that have resulted, will result or may result from any disclosure, publication, or revelation of classified information not consistent with the terms of this Agreement.”); Nat’l Counterintel. & Sec. Ctr., supra note 38, ¶ 12 (similar).
  • 48Off. of the Dir. of Nat’l Intel., supra note 42, ¶ 4; Nat’l Counterintel. & Sec. Ctr., supra note 38, ¶ 6.
  • 49Off. of the Dir. of Nat’l Intel., supra note 42, ¶ 7; Nat’l Counterintel. & Sec. Ctr., supra note 38, ¶ 8.
  • 50See Snepp, 444 U.S. at 510–11, 511 n.6.
  • 51United States v. Marchetti, 466 F.2d 1309, 1316 (4th Cir. 1972); see also Snepp, 444 U.S. at 511 n.6.
  • 52See, e.g., Casey, supra note 5, at 434, 437 (describing the makeup of the CIA’s and NSA’s prepublication review boards).
  • 53Id. at 441 (quoting John Hollister Hedley, Reviewing the Work of CIA Authors: Secrets, Free Speech, and Fig Leaves, Stud. Intel., Spring 1998, at 75, 82–83).
  • 54See id. at 434–35, 437; Off. of the Dir. of Nat’l Intel., Pre-Publication Review—Frequently Asked Questions 3 (2020).
  • 55See Korematsu v. United States, 323 U.S. 214, 218 (1944); Trump v. Hawaii, 138 S. Ct. 2392, 2409 (2018). But see Youngstown Sheet & Tube Co. v. Sawyer, 343 U.S. 579, 589 (1952).
  • 56E.g., McGehee v. Casey, 718 F.2d 1137, 1148 (D.C. Cir. 1983) (“We conclude that reviewing courts should conduct a de novo review of the classification decision, while giving deference to reasoned and detailed CIA explanations of that classification decision.”).
  • 57E.g., Alfred A. Knopf, Inc. v. Colby, 509 F.2d 1362, 1368–69 (4th Cir. 1975).
  • 58See Marchetti, 466 F.2d at 1318.
  • 59Id.
  • 60See Mart & Ginsburg, supra note 30, at 748.
  • 61This is not to say that courts never deal with classified information. See, e.g., infra text accompanying notes 202–04 (discussing FISA courts). For example, the Classified
    Information Procedures Act, 18 U.S.C. app. 3, prescribes security procedures when a criminal trial involves classified information. See 18 U.S.C. app. 3 § 9.
  • 62See Freedman v. Maryland, 380 U.S. 51, 58–59 (1965).
  • 63Marchetti, 466 F.2d at 1317.
  • 64Nat’l Counterintel. & Sec. Ctr., supra note 38, ¶ 5.
  • 65See, e.g., Stillman v. CIA (Stillman II), 517 F. Supp. 2d 32, 36 (D.D.C. 2007); Boening v. CIA, 579 F. Supp. 2d 166, 172 (D.D.C. 2008).
  • 66496 F. Supp. 3d 146 (D.D.C. 2020).
  • 67See id. at 157; Bolton I, 468 F. Supp. 3d 1, 3 (D.D.C. 2020) (order denying temporary restraining order and preliminary injunction).
  • 68Bolton II, 496 F. Supp. 3d at 157.
  • 69See, e.g., Casey, supra note 5, at 448; Bolton I, 468 F. Supp. 3d at 3; Greg Miller, Panetta Clashed with CIA over Memoir, Tested Agency Review Process, Wash. Post (Oct. 21, 2014),
  • 70See supra text accompanying notes 44–46.
  • 71See Knopf, 509 F.2d at 1370–71.
  • 72403 U.S. 713 (1971).
  • 73See, e.g., Brittany Gibson, All the President’s Whistleblowers, Am. Prospect (Oct. 18, 2019),
  • 74Treatment of leakers involves another can of constitutional worms beyond the scope of this Comment. For a discussion of the topic, see generally Heidi Kitrosser, Free Speech Aboard the Leaky Ship of State: Calibrating First Amendment Protections for Leakers of Classified Information, 6 J. Nat’l Sec. L. & Pol’y 409 (2013).
  • 75See, e.g., Jeff Stein, Four Books About the C.I.A.’s Exploits and Secrets, N.Y. Times (Aug. 29, 2019),
  • 76See Kurt Andersen, How America Lost Its Mind, The Atlantic (Dec. 28, 2017),
  • 77Cf. Steven Aftergood, DNI Says Build Trust in Intelligence Through Transparency, Fed’n Am. Scientists: Secrecy News (Apr. 3, 2018),
  • 78See Bradley P. Moss, Will the Justice Department Prosecute John Bolton?, Lawfare Blog (Sept. 25, 2020),
  • 79468 F. Supp. 3d 1 (D.D.C. 2020) (order denying temporary restraining order and preliminary injunction).
  • 80Id. at 6.
  • 81See Stone, supra note 27, at 40–41.
  • 82Snepp, 444 U.S. at 507.
  • 83United States v. Snepp, 456 F. Supp. 176, 178 (E.D. Va. 1978).
  • 84See Snepp, 444 U.S. at 507.
  • 85See Snepp, 456 F. Supp. at 177.
  • 86Id. at 177, 182.
  • 87Snepp, 444 U.S. at 509 n.3.
  • 88Id. at 515.
  • 89Id. at 509 n.3.
  • 902 Rodney A. Smolla, Smolla & Nimmer on Freedom of Speech § 15:25 (3d ed. 1996).
  • 91Snepp, 444 U.S. at 509 n.3, 512; id. at 519 (Stevens, J., dissenting).
  • 92See Kitrosser, supra note 74, at 432–33.
  • 93Archibald Cox, Foreword: Freedom of Expression in the Burger Court, 94 Harv. L. Rev. 1, 9–10 (1980).
  • 94See, e.g., Kitrosser, supra note 74, at 432.
  • 95See, e.g., United States v. National Treasury Employees Union, 513 U.S. 454, 465 (1995); United States v. Aguilar, 515 U.S. 593, 606 (1995).
  • 96See, e.g., Zook v. Brown, 865 F.2d 887, 890 (7th Cir. 1989); Weaver v. U.S. Info. Agency, 87 F.3d 1429, 1439 (D.C. Cir. 1996); Edgar v. Coats, 454 F. Supp. 3d 502, 531–32 (D. Md. 2020).
  • 97Kitrosser, supra note 74, at 432.
  • 98See William W. Van Alstyne, The Demise of the Right-Privilege Distinction in Constitutional Law, 81 Harv. L. Rev. 1439, 1440, 1442–45 (1968).
  • 99See id. at 1461–62.
  • 100408 U.S. 593 (1972).
  • 101Id. at 594–95.
  • 102Id. at 597.
  • 1032 Smolla, supra note 90, § 7:15.
  • 104Van Alstyne, supra note 98, at 1449.
  • 105Dolan v. City of Tigard, 512 U.S. 374, 390–91 (1994) (collecting cases).
  • 106Id. at 386, 391.
  • 107Id. at 391.
  • 108See, e.g., Cass R. Sunstein, Why the Unconstitutional Conditions Doctrine Is an Anachronism (with Particular Reference to Religion, Speech, and Abortion), 70 B.U. L. Rev. 593, 604 (1990).
  • 109391 U.S. 563 (1968).
  • 110461 U.S. 138 (1983).
  • 111511 U.S. 661 (1994).
  • 112Weaver, 87 F.3d at 1441.
  • 113Waters, 511 U.S. at 668 (quoting Connick, 461 U.S. at 142).
  • 114See 2 Smolla, supra note 90, § 18:8.
  • 115Id.
  • 116Waters, 511 U.S. at 673.
  • 117See, e.g., Gustafson v. Jones, 290 F.3d 895, 909 (7th Cir. 2002) (listing seven factors considered by the Pickering Court).
  • 118Cf. Alfred A. Knopf, Inc. v. Colby, 509 F.2d 1362, 1370 (4th Cir. 1975) (suggesting that reports by officials are particularly credible to the public).
  • 119See Weaver, 87 F.3d at 1442.
  • 120See id.
  • 121See, e.g., Snepp, 444 U.S. at 510–11; United States v. Marchetti, 466 F.2d 1309, 1313 (4th Cir. 1972); Knopf, 509 F.2d at 1370.
  • 122See 2 Smolla, supra note 90, § 18:7.
  • 123See, e.g., Pickering, 391 U.S. at 571.
  • 124See supra Part I.C.
  • 125Marchetti, 466 F.2d at 1318–19 (Craven, J., concurring).
  • 1262 Smolla, supra note 90, § 15:1.
  • 127Bantam Books, Inc. v. Sullivan, 372 U.S. 58, 70 (1963).
  • 128Cf. Stone, supra note 27, at 40–41 (explaining that an early interpretation of the First Amendment was that it codified an English rule forbidding prior restraints on speech but allowed punishment for sedition after publication).
  • 129See Times Film Corp. v. City of Chicago, 365 U.S. 43, 47–49 (1961).
  • 130See Kingsley Books, Inc. v. Brown, 354 U.S. 436, 441–42 (1957).
  • 131372 U.S. 58 (1963).
  • 132Id. at 59–63.
  • 133Id. at 67.
  • 134Id. at 66, 64.
  • 135Id. at 70.
  • 136Bantam, 372 U.S. at 71.
  • 137380 U.S. 51 (1965).
  • 138Id. at 52–53.
  • 139Id. at 57–58.
  • 140Id. at 58.
  • 141Id. at 59.
  • 142See supra text accompanying notes 58–61.
  • 14350 U.S.C. § 3024(i)(1).
  • 144See, e.g., Willemain, supra note 6.
  • 145See supra Parts I.B, I.D; infra Part III.C.3.
  • 146283 U.S. 697 (1938). The very first discussion (although not explicitly using the term “prior restraint”) was in Schenck v. United States, 249 U.S. 47, 52–53 (1919).
  • 147Near, 283 U.S. at 701–02, 723 (quoting Act of Apr. 20, 1925, ch. 285, § 1(b), 1925 Minn. Laws 358, 358).
  • 148Id. at 716.
  • 149The Pentagon Papers Case, 403 U.S. at 714.
  • 150See, e.g., United States v. Progressive, Inc., 467 F. Supp. 990, 995 (W.D. Wis. 1979).
  • 151See 2 Smolla, supra note 90, § 15:25.
  • 152Snepp, 444 U.S. at 509; see also Marchetti, 466 F.2d at 1318.
  • 153Weaver, 87 F.3d at 1440.
  • 154See, e.g., Williamson v. Lee Optical of Okla., Inc., 348 U.S. 483, 487–88 (1955) (explaining that, under rational basis review, “the law need not be in every respect logically consistent with its aims to be constitutional. It is enough that there is an evil at hand for correction, and that it might be thought that the particular legislative measure was a rational way to correct it.”).
  • 155See Se. Promotions, Ltd. v. Conrad, 420 U.S. 546, 552–60 (1975) (describing the narrow exceptions and procedural requirements that the Court’s doctrine requires for a prior restraint to pass muster).
  • 156See Williams-Yulee v. Fla. Bar, 575 U.S. 433, 442 (2015).
  • 157391 U.S. 367 (1968).
  • 158Id. at 369–70.
  • 159See 2 Smolla, supra note 90, §§ 9:10, 9:17.
  • 160See O’Brien, 391 U.S. at 376–77.
  • 161Id. at 377.
  • 162Id.
  • 163444 U.S. 348 (1980).
  • 164See Snepp, 444 U.S. at 509 n.3.
  • 165Brown, 444 U.S. at 350–51.
  • 166Id.
  • 167Id. at 355.
  • 168Id. at 354.
  • 169Id. at 354, 357.
  • 170Brown, 444 U.S. at 356 n.13.
  • 171Snepp, 444 U.S. at 509 n.3.
  • 172See, e.g., McGehee v. Casey, 718 F.2d 1137, 1142 (D.C. Cir. 1983); Zook, 865 F.2d at 890; Weaver, 87 F.3d at 1440.
  • 173718 F.2d 1137 (D.C. Cir. 1983).
  • 174Id. at 1139.
  • 175Id. at 1140.
  • 176Id. at 1140, 1142.
  • 177Id. at 1143–44.
  • 178Weaver, 87 F.3d at 1440.
  • 179See id. at 1441.
  • 180Snepp, 444 U.S. at 509 n.3.
  • 181O’Brien, 391 U.S. at 381–82.
  • 182Snepp, 444 U.S. at 510.
  • 183Id. at 508 (emphasis in original).
  • 184Id. at 511.
  • 185O’Brien, 391 U.S. at 377.
  • 186Id.
  • 187See 2 Smolla, supra note 90, § 9:17.
  • 188McCullen v. Coakley, 573 U.S. 464, 486 (2014) (quoting Ward v. Rock Against Racism, 491 U.S. 781, 796 (1989)).
  • 189Id. at 477, 486 (citing Ward, 491 U.S. at 791, 798).
  • 190Snepp, 444 U.S. at 509 n.3.
  • 191Id.
  • 192See, e.g., Abel, supra note 16, at 1097.
  • 193See, e.g., Casey, supra note 5, at 452–54.
  • 194U.S. Const. art. II, § 1–2.
  • 195See Federal Polygraph Limitation and Anti-Censorship Act of 1984, H.R. 4681, 98th Cong. (1984), reprinted in H.R. Rep No. 98-961, at 4 (1984) (seeking “to ban the use and enforcement of prepublication review requirements by Federal agencies against their employees”).
  • 196See Federal Polygraph Limitation and Anti-Censorship Act: Hearing Before the Subcomm. on Civ. & Const. Rights of the H. Comm. on the Judiciary, 98th Cong. 21 (1984) (statement of Michael E. Tigar, Law Professor, Univ. of Tex.). Despite this skepticism, Tigar concluded that “restrictions on freedom of speech should [not] be tolerated because of a guess about what the Supreme Court will do.” Id.
  • 197See Seila Law, LLC v. CFPB, 140 S. Ct. 2183, 2197 (2020) (holding that a for-cause restriction on the president’s ability to remove an agency’s sole director violated constitutional separation of powers).
  • 198See Trump v. Mazars USA, LLP, 140 S. Ct. 2019, 2035–36 (2020) (requiring that courts evaluate the sufficiency, tailoring, and supporting evidence of congressional purpose and burdens on the president before permitting Congress to subpoena the president for private, unofficial documents).
  • 199U.S. Const. amend. I.
  • 200United States v. Marchetti, 466 F.2d 1309, 1314 (4th Cir. 1972).
  • 201See, e.g., Terry M. Moe & Scott A. Wilson, Presidents and the Politics of Structure, 57 Law & Contemp. Probs. 1, 15–17 (1994) (comparing the presidency’s institutional strengths to Congress’s).
  • 202See, e.g., Mart & Ginsburg, supra note 30, at 752 (suggesting specialized courts as a possible solution to overclassification); Grayson M.P. McCouch, Note, “Naming Names”: Unauthorized Disclosure of Intelligence Agents’ Identities, 33 Stan. L. Rev. 693, 711 n.95 (1981) (proposing the creation of a specialized court analogous to the U.S. Tax Court to handle prepublication review disputes).
  • 203Pub. L. No. 95-511, 92 Stat. 1783 (codified as amended in scattered sections of 50 U.S.C.).
  • 204See Elizabeth McElvein, The Political Landscape of FISA Reauthorization, Lawfare Blog (Feb. 25, 2020),
  • 205See infra Part III.C.
  • 206See Snepp, 444 U.S. at 507.
  • 207Id. at 513 n.8 (emphasis in original).
  • 208Id. at 511.
  • 209See supra text accompanying note 87; Part II.D.
  • 210Snepp, 444 U.S. at 511–12 (emphasis added).
  • 211McGehee, 718 F.2d at 1141; see also Bolton II, 496 F. Supp. 3d at 157 (“[P]republication review does not allow the government to permanently restrain a former employee from publishing unclassified information.”); Marchetti, 466 F.2d at 1317 (“We would decline enforcement . . . to the extent that [the secrecy oath] purports to prevent disclosure of unclassified information, for, to that extent, the oath would be in contravention of his First Amendment rights.”).
  • 212See supra Part I.D; Casey, supra note 5, at 448; Jack Goldsmith & Oona Hathaway, The Scope of the Prepublication Review Problem, and What to Do About It, Lawfare Blog (Dec. 30, 2015),; see also supra Part III.A.
  • 213Snepp, 444 U.S. at 511–12.
  • 214See, e.g., Marchetti, 466 F.2d at 1318.
  • 215467 F. Supp. 990 (W.D. Wis. 1979).
  • 216See id. at 998–1000.
  • 217Bolton II, 496 F. Supp. 3d at 149, 159 (referring to “the classified information about intelligence sources and methods known as sensitive compartmented information,” including Bolton’s memoir).
  • 218See supra Part II.D.
  • 219See supra text accompanying notes 16–20.
  • 220Off. of the Dir. of Nat’l Intel., supra note 54, at 2.
  • 221See Emma Grey Ellis, WikiLeaks Has Officially Lost the Moral High Ground, Wired (July 27, 2016),
  • 222See generally, e.g., Alicia Solow-Niederman, Beyond the Privacy Torts: Reinvigorating a Common Law Approach for Data Breaches, 127 Yale L.J.F. 614 (2018), (discussing privacy torts and the breach of confidentiality tort as potential remedies to data breaches).
  • 223O’Brien, 391 U.S. at 376–77; Brown, 444 U.S. at 354.
  • 224See, e.g., Inspector Gen., U.S. Dep’t of Def., Follow up to DoD Evaluation of Over-Classification of National Security Information 1–2 (2016) (evaluating the Department of Defense’s progress toward implementing its inspector general’s recommendations to identify policies that contribute to misclassification).
  • 2253 C.F.R. 303.
  • 226See Progressive, 467 F. Supp. at 995 (“Faced with a stark choice between upholding the right to continued life and the right to freedom of the press, most jurists would have no difficulty in opting for the chance to continue to breathe and function as they work to achieve perfect freedom of expression.”).
  • 227This relatively simple story is complicated by the fact that the government claimed that the article contained information that remained classified even if it was in the public domain. Id. at 993. That said, the government’s position was “that whether or not specific information is ‘in the public domain’ or has been ‘declassified’ at some point is not determinative” because the court should instead “analyze what the practical impact of the prior disclosures are.” Id. Independent of whether the government was correct about its classification arguments, the case remains a useful lens through which to evaluate what constitutes “practical impact” and substantial interest.
  • 228See, e.g., Jonathan L. Entin, Separation of Powers, the Political Branches, and the Limits of Judicial Review, 51 Ohio St. L.J. 175, 186–209, 217 n.291 (1990) (describing cases where the Supreme Court reviewed the allocation of power between branches). But see Cannon v. Univ. of Chi., 441 U.S. 677, 730–31, 747–49 (1979) (Powell, J., dissenting).
  • 229See supra Part II.C.
  • 230For a refresher on where “narrowly tailored” and “reasonable means” are coming from and what they mean for the intermediate scrutiny analysis, see Part II.D.
  • 2315 U.S.C. § 552.
  • 2323 C.F.R. 303.
  • 233See supra text accompanying notes 62–64; infra Part III.C.3.
  • 234Stillman v. Dep’t of Def. (Stillman I), 209 F. Supp. 2d 185, 217–21 (D.D.C. 2002) (order granting in part and denying in part motion to compel access), rev’d on other grounds sub nom. Stillman v. CIA, 319 F.3d 546 (D.C. Cir. 2003).
  • 235Cf. Karl N. Llewellyn, The Bramble Bush 11 (11th prtg. Oxford Univ. Press 2008) (1930) (“You must read each substantive course, so to speak, through the spectacles of the procedure. For what substantive law says should be means nothing except in terms of what procedure says that you can make real.”).
  • 236See McGehee, 718 F.2d at 1141 n.9; Boening v. CIA, 579 F. Supp. 2d 166, 170–71 (D.D.C. 2008).
  • 237McGehee, 718 F.2d at 1141 n.9.
  • 238See, e.g., Restatement (Third) of Torts: Physical & Emotional Harm § 17 cmt. i (Am. L. Inst. 2010).
  • 239One worry is that the government would simply invoke the “Glomar response,” in which the government claims that admitting whether the information exists at all is itself a classified item. Phillippi v. CIA, 546 F.2d 1009, 1011–12 (D.C. Cir. 1976). Glomar responses are most often used in the FOIA context but would be inapt in prepublication review cases because the source of the information is different. In a FOIA request, a member of the public is requesting a specific piece of information from the government; a Glomar response means that the government neither confirms nor denies whether it has such information. In prepublication review, the writer is the one forwarding a specific piece of information and the government is flagging it based on its claim that such information is in its possession and is classified.
  • 2403 C.F.R. 302.
  • 241See Bolton I, 468 F. Supp. 3d at 6.
  • 242See id.
  • 243509 F.2d 1362 (4th Cir. 1975).
  • 244Id. at 1370–71. (“It is true that others may republish previously published material, but such republication by strangers to it lends no additional credence to it. [Former intelligence officials] are quite different, for their republication of the material would lend credence to it.”). Note that this reasoning is especially strange with respect to unclassified information. If the former official did not have the qualifications sufficient to lend credence to her statements, why speak at all? If this were the framework, it would seem that former officials would only be allowed to speak on topics that they know nothing about. Surely that cannot be what the First Amendment encourages—let alone the only thing it permits.
  • 245See Off. of the Dir. of Nat’l Intel., supra note 54, at 3.
  • 246Pub. L. No. 79-404, 60 Stat. 237 (1946) (codified as amended in scattered sections of 5 U.S.C.).
  • 247See Boening, 579 F. Supp. 2d at 172; Stillman II, 517 F. Supp. 2d 32, 36 (D.D.C. 2007).
  • 248See Boening, 579 F. Supp. 2d at 172; Stillman II, 517 F. Supp. 2d at 36.
  • 249See Lujan v. Defs. of Wildlife, 504 U.S. 555, 560–61 (1992).
  • 250U.S. Const. art. III, § 2; see also, e.g., Powell v. McCormack, 395 U.S. 486, 496 & n.7 (1969).
  • 251S. Pac. Terminal Co. v. Interstate Com. Comm’n, 219 U.S. 498, 515 (1911).
  • 252See Boening, 579 F. Supp. 2d at 172 (citing City of Los Angeles v. Lyons, 461 U.S. 95, 109 (1983)) (determining that the plaintiff’s situation failed to satisfy either requirement); see also Stillman II, 517 F. Supp. 2d at 36 (determining that the chance of future harm was not “more-than-speculative” (quoting Clarke v. United States, 915 F.2d 699, 701 (D.C. Cir. 1990))).
  • 253Stillman II, 517 F. Supp. 2d at 36.
  • 254See Stephen Wermiel, Battling over Mootness, SCOTUSBlog (Aug. 29, 2019),
  • 255Id.
  • 256That the government controls the pace of review might make the completion of prepublication review while litigation is pending look more like the “voluntary cessation” exception to mootness than capability of repetition. See County of Los Angeles v. Davis, 440 U.S. 625, 631 (1979) (quoting United States v. W.T. Grant Co., 345 U.S. 629, 632 (1953)). Voluntary cessation requires that there is no reasonable expectation that the issue will recur and that the cessation of the activity completely resolves the controversy. See id. The core difference between voluntary cessation and capability of repetition is that the former requires some action by at least one of the parties. If it were clear that the government were deliberately completing the prepublication review in order to moot the litigation, it would probably fall under voluntary cessation. From the outside, though, it appears that prepublication reviews simply come to a natural end during litigation. Either way, it would be excepted from mootness.
  • 257See, e.g., Roe v. Wade, 410 U.S. 113, 164 (1973).
  • 258See 3 C.F.R. 300.
  • 259See 3 C.F.R. 300.