This article first appeared in Just Security.
Access to sensitive national security information comes with critical constraints and responsibilities. Government employees and contractors who work with classified information sign non-disclosure agreements (NDAs) pledging not to divulge that information. They also agree to participate in prepublication review—a process by which agencies review proposed publications, talking points, and resumes to ensure they do not contain classified material.
These employees and contractors do not, however, relinquish their First Amendment rights. Even as conditioned by NDAs, prepublication review is a content-based prior restraint on speech—one of the most severe infringements on First Amendment liberties. The process passes First Amendment muster only if it is reasonably tailored to the government’s interest in preventing national security harm; has narrow, objective, and definite standards to guide government reviewers and cabin official discretion; and includes robust procedural safeguards designed to mitigate the dangers of illegitimate censorship. Although a federal appellate court recently held otherwise, there is a strong argument that the current prepublication review system does not meet that standard.
Constitutionality aside, the system as operated across departments and agencies is rife with flaws. It requires too many people to submit their draft publications for review, including not just those with recent access to highly sensitive information but anyone who ever had access to any level of classified material. Many agencies require the submission of any document that touches on a broad array of topics, regardless of whether the document is likely to include or have been derived from classified information. Reviewers hold up publications for reasons unrelated to the protection of classified information—including potential embarrassment to agency officials. And the lack of any firm deadlines for agencies to complete their review leaves authors in limbo for months or even years.
These dysfunctions often serve to bar, delay, or deter publication—outcomes that not only curtail the authors’ freedom of speech, but also “lead to an impoverished public discourse” on national security matters, as Jack Goldsmith and Oona Hathaway have written. Alternatively, some authors give up on the process altogether and publish works without submitting them for prepublication review, increasing the risk of unintentional disclosures of classified information. And many individuals, including one of us, have declined national security positions in government because the lifetime prepublication review requirement would unduly constrain any post-government writing they wished to do.
In July, the Director of National Intelligence (DNI) issued a new Intelligence Community (IC) Directive that overhauls the current system and sets a uniform standard with which the IC’s prepublication review policies—which currently vary significantly by agency—must comply.
The DNI deserves great credit for tackling a thorny set of problems, especially given the pushback she almost certainly faced from some corners of the intelligence community. The resulting directive will make several important improvements to the system, including a new standard for what must be submitted and multiple provisions that will make the process more transparent and facilitate oversight. (As discussed below, we take a more positive view of the new standard than the one put forward by Goldsmith and Hathaway.) Unfortunately, however, the directive fails to make certain critical changes, including what might be the most important reform of all: a hard deadline for the prepublication review process. We thus second Goldsmith and Hathaway’s assessment that the reforms can and should be taken further.
Key Improvements to the Process
More appropriate standard for submission of materials. Perhaps the most important change made by the directive is the standard for what materials must be submitted. Going forward, those who hold or held access to classified information will be required to submit materials for prepublication review only if “a fully-informed person granted access to covered intelligence might reasonably deem [them] to contain or be derived from covered intelligence.” The directive defines “covered intelligence” as “classified intelligence and classified intelligence-related information.”
Goldsmith and Hathaway take issue with the “fully-informed person” criterion. They argue that this standard is essentially impossible to meet, as no individual author will have access to all the information that bears on whether what he or she wrote is classified. Although we greatly respect these authors and their writings on this topic, we must disagree. We read “fully-informed,” in this context, to mean that the (hypothetical) person not only had access to covered intelligence, but was familiar with the contents of the intelligence to which they had access—not that they are aware of every existing fact that might relate to that intelligence. We therefore do not see significant daylight between the standard in the directive and the one Goldsmith and Hathaway have proposed—namely, “writings that might reasonably contain or be derived from classified information should be subject to review.”
Thus interpreted (and the DNI can—and should—issue guidance making this interpretation clear), the new standard is a major improvement over current rules (summarized here). Many agencies require the submission of almost any material a current or former employee might produce. For instance, the Office of the Director of National Intelligence policy requires submission of “all official and non-official information intended for publication that discusses the ODNI, the IC, or national security.” The State Department requires current personnel to submit all proposed publications that “pertain to U.S. foreign policy or the Department’s mission.” The FBI requires submission of “[a]ll information created and acquired by current and former [FBI employees and contractors] in connection with official FBI duties.” The Department of Homeland Security similarly requires the submission of “information derived as a result of affiliation with DHS.”
Requiring current and former employees to obtain prior approval for publications that are highly unlikely to contain or have been derived from any classified information goes far beyond what is necessary to protect national security, and thus encroaches on would-be authors’ First Amendment rights. The directive, as we interpret it, brings this aspect of the system in line with constitutional principles.
Protection for authors who participate in the process. Astonishingly, under current policies, an agency’s approval of a manuscript offers no legal protection to the author. If it is later determined that the publication includes classified information that the prepublication review process failed to flag, the author remains subject to administrative penalties. A current employee may face termination or loss of security clearance, while a former employee could be barred from future government employment.
Under the directive, “IC elements that approve requests for publication . . . shall not take adverse administrative action against submitters” who publish information with the agency’s approval. Of course, the directive does not, and cannot, prevent non-IC elements from penalizing such authors. In particular, the Department of Justice may still bring criminal charges under various statutes that prohibit the disclosure of classified information. However, the directive notes that “the Department of Justice would likely consider all relevant facts, including whether a publication or disclosure of information is within the scope of an approval, when making any relevant determinations.”
Alignment of agency policies with non-disclosure agreements. The legal basis for the requirement of prepublication review is the non-disclosure agreement (NDA) that government employees and contractors sign as a condition of access to classified information. For both legal reasons and to avoid confusion, it is important that the requirements set forth in the NDA match the requirements in the agencies’ prepublication review policies.
That is currently not the case. As noted above, many agencies require the submission of almost any proposed publication that relates to the work of the agency or national security more broadly. By contrast, the primary NDAs setting forth prepublication review requirements—Form 4414 and Standard Form 312—create much narrower obligations. By signing Form 4414, employees applying for access to the most highly classified material, known as “sensitive compartmented information” (SCI), “agree to submit for [pre-publication review] . . . any writing or preparation in any form, including a work of fiction, that contains or purports to contain any SCI or description of activities that produce or relate to SCI or that I have reason to believe are derived from SCI, that I contemplate disclosing to any person not authorized to have access to SCI or that I have prepared for public disclosure.” By signing SF 312, employees applying for access to any level of classified information agree to the following: “I understand that if I am uncertain about the classification status of information, I am required to confirm from an authorized official that the information is unclassified before I may disclose it” (emphasis added).
The directive resolves this problem by providing that “IC elements shall, within 180 days from signature of this ICD, revise all future NDAs to conform with this Directive.” The standard for submission under the revised NDAs will have to match the standard in the directive—a standard that falls between the ones articulated in current NDAs and the ones enshrined in current agency policies.
Limitation on access to submitted materials. A manuscript submitted for prepublication review should be circulated only to those officials whose participation in the review process is strictly necessary. There are two separate reasons for this. First and foremost, the prepublication review process presents an opportunity for political interference—i.e., delaying or denying approval for a publication because it criticizes political actors or agency policies. The more widely a manuscript is circulated—and particularly if it is sent to political appointees—the greater the opportunity for such interference.
Under the current system, political actors can and do intervene. The most obvious and egregious recent example is John Bolton’s book, “The Room Where It Happened.” The responsible reviewer in that case was a veteran information security professional. She spent literally hundreds of hours with another staff reviewer combing over successive revisions of the manuscript, as well as upwards of a dozen hours iteratively and exhaustively reviewing the manuscript on a line-by-line basis with the author. After she determined that the manuscript no longer contained any classified information, officials in the National Security Council and White House Counsel’s Office instructed her not to issue a final decision. A political appointee with no previous experience in pre-publication review or classification review then proceeded to conduct a separate review for the apparent purpose of delaying publication of a book critical of the president.
A second reason to minimize distribution is that draft publications are the authors’ intellectual property and should be protected as such. Anecdotally, some authors have learned from contacts within the reviewing agency that their manuscripts are circulating widely among agency personnel who have no role in performing the prepublication review.
The directive requires IC elements to ensure that “all non-official material in prepublication review requests are reviewed only by the necessary individuals authorized to assist in completing the prepublication review, including prepublication review personnel, government legal counsel, and subject-matter experts, and only for the purpose of prepublication reviews.” It also states that “[p]republication review personnel shall disclose to subject-matter experts only the minimum amount of non-official material necessary for them to review. Subject-matter experts may not further distribute or circulate material.” It emphasizes that “[u]nder no circumstances may material be shared with personnel . . . for reasons other than facilitating the prepublication review.”
While the directive unfortunately does not prohibit political appointees from participating in the review process, it does include explicit direction that “[p]republication reviews shall be timely, reasoned, impartial, and free from political or policy bias” (emphasis ours). More concretely (and therefore more usefully), political appointees would likely have to be deemed “subject-matter experts” to participate in the review process, and their access to submitted materials would be limited to those portions of the manuscript relevant to their expertise.
Enhanced transparency and record-keeping. Under the current system, government employees and contractors struggle to find concrete and consistent information about their prepublication review obligations. When they submit materials for review, the submissions often fall into a void; communications from many reviewing agencies are rare and unhelpful. Agencies lack systems to comprehensively and effectively track submissions, and they do not maintain the type of data that would facilitate oversight.
The directive contains several robust transparency and record-keeping provisions that would go far toward addressing these problems, including the following requirements:
- IC elements must provide training on prepublication review requirements and ensure that all personnel, upon entry-on-duty and exit-from-duty, are aware of their prepublication review obligations.
- IC elements must provide IC personnel exiting from duty a single set of written instructions for prepublication review.
- IC elements must maintain public-facing and internal websites regarding prepublication review that are accessible from the element’s main page and that include links to all relevant policies and instructions, as well as information about points of contact and instructions on how to submit material for review. ODNI must maintain its own website to serve as a centralized prepublication review resource for all IC elements and personnel.
- IC elements must promptly acknowledge the receipt of a prepublication review request and provide the submitter with information about the prepublication review process, including applicable deadlines and contact information for any relevant points of contact.
- If the review takes longer than 30 business days from the date of acknowledgment, the IC element shall immediately notify the submitter with a status update and reason for the delay and then provide an update every four weeks thereafter.
- IC elements must provide one consolidated and comprehensive written response for each prepublication review request, clearly indicating which portions of the non-official material may be disclosed without change or require modification prior to disclosure, and which portions are not authorized for disclosure.
- IC elements must document the reason for an approval, modification, or denial of each prepublication review request.
- IC elements must establish and maintain an up-to-date, searchable data system that tracks prepublication review requests and related metrics, including (among other information) the date of submission, any other agencies/components to which the material was referred, a brief description of any communications or meetings with the submitter, the date on which the submitter was given a final substantive response, and a summary of that response. IC elements must provide ODNI with the relevant data on an annual basis and as requested. ODNI, in turn, must oversee an audit based on this data for each element at least once every five years.
These requirements are essential to modernizing and standardizing the administration of prepublication review, as well as to enabling oversight and accountability. Whether and when agencies budget for and implement systems to meet these requirements, however, will bear continued watching.
Required appeals process. Currently, at least half of the agencies’ policies do not provide for appeals. Of course, reviewers are not infallible, and so it is important to provide an internal correction mechanism for erroneous review decisions. It is not sufficient to say that authors may bring lawsuits, as some might lack the time or resources to do so. The directive makes clear that all IC elements must establish an appeals process for prepublication review requests that have been denied.
Critical Missing Reforms
Lack of firm deadlines.Despite the many significant improvements it puts in place, the directive suffers from a fatal flaw: the failure to include firm, enforceable deadlines for prepublication review. Indeed, some current agency policies include stricter deadlines than the soft targets contained in the directive. To be sure, most policies acknowledge that deadlines might not be met in all cases—and in fact, deadlines are routinely honored in the breach, as there is no consequence for an agency’s failure to meet them. This is particularly true for book-length manuscripts, which can take years to review rather than the 30 days contemplated by multiple agency policies.
Experts have proposed that agencies adopt deadlines that correspond to the number of words in the submission. For instance, the Knight First Amendment Institute at Columbia University suggests that agencies should be required to review manuscripts of less than 1,500 words within three days; manuscripts of 1,500–10,000 words within 14 days; and manuscripts of more than 10,000 words within 30 days. To make deadlines enforceable, experts have also proposed that a failure to meet deadlines should constitute de facto approval for publication—or, similarly, a waiver of the agency’s right to take administrative action against the author.
The directive does not take this approach. For shorter pieces like op-eds or blog posts, the directive states only that “[r]elatively short, time-sensitive requests (e.g., letters to the editor, some oral statements) shall be handled as expeditiously as practicable.” For other submissions, the directive states that IC elements “shall strive to complete a prepublication review request within 30 business days of receipt, or within 90 days for lengthy publications (e.g., books, large manuscripts), to the maximum extent practicable.” Of course, this language expressly contemplates that the review might exceed the stated time periods. There is no backstop—no maximum length of time for a review that has gone beyond the target deadlines in the directive; if the agency review takes longer than 30 days, the agency need only update the author every four weeks.
In the past, a major source of delay has been referral of manuscripts to other agencies that may have equities in the document. Agencies have little incentive to make referrals in a timely manner, and agencies that receive referrals do not appear to follow the deadlines that would apply if they had received the submission in the first instance. The directive does require referrals to be made within 10 or 15 business days of receipt, depending on the length of the publication. But it does not impose firm deadlines on the agencies that receive the referrals. Instead, it requires those agencies to “make every effort to respond to the receiving IC element within 15 business days of receipt, or within 45 days for lengthy publications (e.g., books, large manuscripts).” If the agencies miss those target deadlines, “they shall notify the receiving IC element every 10 days until completion of the review, or every 30 days for lengthy publications.”
The ability to delay approval indefinitely, which the directive leaves in place, is at the root of the system’s dysfunction. It greases the skids for political interference, as agencies need not point to actual classified information to interfere with publication; all they have to do is wait. It causes many would-be authors—particularly those who seek to weigh in publicly on time-sensitive matters—to simply give up on publishing. In other cases, authors decide to go forward without submitting their work for prepublication review, thus increasing the chances of an inadvertent disclosure of classified information. Finally, the lack of meaningful deadlines and the resulting lengthy periods of review cannot be squared with authors’ First Amendment rights.
A slight mitigating factor is the directive’s requirement for data-keeping and reporting. If agencies are continuing to blow past deadlines for review, that fact will be apparent in the data the agencies must report to ODNI, and it will presumably be a focus of the ODNI’s periodic audits. That transparency could help to build public support and pressure for creating firmer deadlines, empowering a future DNI to overcome the resistance that the current DNI no doubt faced on this point from intelligence agencies.
Overbroad designation of required submitters. Given the severity of prior restraints on speech, prepublication review should only apply to people who have access to highly sensitive national security information. Access to information that is classified at lower levels, as well as information that is over a decade old and has likely lost some or all of its sensitivity, should not trigger this requirement. Of course, authors who were not required to participate in the review process would not be free to publish classified information; they would still be subject to sanctions, including potential criminal prosecution, if they did so.
The Knight Institute has recommended that prepublication review should be mandatory only for employees and contractors who have held a Top Secret/SCI clearance within the past 10 years. Goldsmith and Hathaway, who formerly served as national security officials, have endorsed this recommendation. Other authors would still be able to avail themselves of the process if they wished. Indeed, if the system were reformed to operate a timely manner and to focus only on classified information, many authors would likely take advantage of that option in order to have the added security of agency approval.
The directive does not follow this recommendation or otherwise narrow the pool of people subject to the prepublication review requirement. Instead, it mandates prepublication review for “any individual who previously or currently receives, handles, stores, or processes classified intelligence and classified intelligence-related information.”
Failure to appropriately limit the scope of review. A major problem with the current system is that agencies often withhold approval for publication—or require authors to remove certain material—for reasons other than preventing the disclosure of classified information. Authors have reported that agencies object to material that would prove embarrassing to agency officials; information that agencies deem inaccurate or privileged; information that intrudes on personal privacy; information (even if unclassified) made public by allegedly unauthorized disclosures; and information that the agencies consider “sensitive” for reasons other than national security.
The directive affirms that “[t]he purpose of prepublication review is to ensure the protection of classified government information.” However, it goes on to state: “While not the purpose of prepublication review, during the course of such review an IC element may also identify unclassified but sensitive information that might be otherwise restricted under law or federal regulation. If so, that information may also be flagged for the submitter consistent with the timelines described herein.” This, in itself, is not problematic; indeed, it may be viewed as a courtesy to the author. The directive further provides, however, that reviewers who identify such information may “request its removal”—suggesting, although not explicitly stating, that the agency may withhold approval for publication of such information.
Unclassified information that is protected by law or regulation is a narrower category than the universe of unclassified information agencies currently seek to censor. Nonetheless, allowing agencies to prevent publication of such information is inappropriate. Courts have identified only one interest that is sufficiently compelling to justify the prior restraint of prepublication review: protection of national security. Indeed, they have made clear that agencies may not use the prepublication review process to prevent the publication of unclassified information. To the extent laws and regulations prohibit the disclosure of certain types of information for reasons other than safeguarding national security—for instance, to protect privacy or honor copyright—enforcement of those prohibitions must rely on after-the-fact penalties rather than prior restraints.
As noted, the directive is somewhat ambiguous as to whether reviewers may withhold authorization to publish unclassified information that is protected by law or federal regulation. If the intent of the directive was merely to allow reviewers to flag such information for authors’ benefit—rather than to require its removal, and withhold authorization if the information is not removed—the DNI should follow up with clarifying guidance.
Failure to require meaningful dialogue with authors. Former government employees who have participated in prepublication review have noted the lack of constructive engagement on the part of reviewers. Most notably, reviewers are not required to—and often do not—provide their reasons for denying approval or requesting the removal of information. They reject authors’ requests for in-person meetings, or they agree to meetings but refuse to allow authors’ attorneys to attend. In some cases, authors may be able to reword passages that reviewers have flagged as sensitive in ways that convey the same information without implicating national security, but they are not given the opportunity.
As noted above, the directive does require timely acknowledgment of an author’s request for review and regular updates about the status of the requests, including reasons for any delay. When it comes to the substance of the review, however, the directive underperforms. While IC elements must provide authors with written responses that “clearly indicat[e] which portions of the non-official material may be disclosed without change or require modification prior to disclosure, and which portions are not authorized for disclosure,” it does not require the agencies to provide the reasons for those decisions. Instead, it includes what appears to be a solely internal documentation requirement: “IC elements shall document the reason for an approval, modification, or denial of each prepublication review request.”
The directive also states that “IC elements shall, to the maximum extent practicable, ensure that there is discretion to work with the submitter to devise ways for the submitter to communicate the non-official material in unclassified terms.” The term “discretion,” however, necessarily means that the IC elements may choose not to work with the submitter to develop workable alternatives. And there is no requirement that agencies agree to meet with authors, if authors request such a meeting, let alone allow attorneys to attend the meeting. Without such an opportunity to confer, the discretion to work with the author is unlikely to be operationalized.
It seems likely that the intent of the directive was to foster a better dialogue between reviewers and authors, even though the language implementing that intent falls short. If so, this is another area where further guidance from the DNI could be useful. The DNI should consider issuing Intelligence Community Policy Guidance clarifying that agencies must provide authors with the reason for each requested redaction or revision; that agencies should meet with authors if a meeting is requested; and that agencies should work with authors to find ways to communicate classified material in unclassified terms unless it is manifest that no such workaround is possible.
* * *
There is much in the DNI’s prepublication review directive to applaud, including meaningful substantive changes and important transparency provisions. But there is still work to be done. Most crucially, without real deadlines, prepublication review will continue to frustrate both the authors who participate in the system and the First Amendment rights they should enjoy. The directive is proof of concept that fundamental change to the system is possible. ODNI should build on this change in the future to complete the project of prepublication review reform.
Elizabeth (Liza) Goitein (@LizaGoitein) is Co-Director of the Liberty & National Security Program at the Brennan Center for Justice at NYU Law.
John Fitzpatrick provides strategic advisory services in management of corporate, government, and international security, cyber, counterintelligence, and related matters.
