Skip to content
Our Secrecy Paradox
Go to my account

Our Secrecy Paradox

The government can’t stop intelligence leaks, but it stifles our national debate by preventing the publication of unclassified information.

Two decisions by the Justice Department in the last month reveal the inanity of the U.S. government’s secrecy system. The department’s new policy against seizing reporter communication information in leak investigations highlights government fecklessness in halting the flow of classified national security secrets to the public. And its cessation of legal actions started by the Trump administration against former National Security Adviser John Bolton over the publication of his memoir, while good for Bolton, highlights how successful the government normally is in preventing the publication of unclassified information. 

The executive branch cannot or will not stop leaks about the classified secrets that matter, and yet it imposes a broad prior restraint on the publication of unclassified information vital to public debate. This central paradox of the American secrecy system has a single root cause: massive overclassification of information by a tumefied intelligence bureaucracy.

The Justice Department backdown on reporter records grew out of leak investigations that began during the Trump administration. At least two sets of leaks appear to be in play. First, an April 2017 New York Times article revealed that the United States (with the help of a Dutch intelligence service, it was later disclosed) had penetrated elements of the Russian operation to interfere in the 2016 U.S. election. 

Second was a series of leaks related to the government’s investigation of the 2016 Russian interference, and possible connections between Russia and the Trump campaign. These leaks included intercepted conversations between the Russian Ambassador Sergey Kislyak and incoming National Security Adviser Michael Flynn; of Russian officials discussing their supposed influence over Flynn and “derogatory” information on Trump and his campaign staff; and of Kislyak reporting to Moscow that Jared Kushner, Trump’s son-in-law, wanted to open a secret communication channel during the presidential transition.

National security leaks are commonplace, but these were unusual. Never before had the detailed contents of foreign intelligence surveillance intercepts leaked to the public. Some of the leaks almost certainly revealed sensitive surveillance techniques employed by the United States or its allies against Russia. And some leaks violated the privacy rights of Americans that U.S. law tries hard to protect. These leaks were but a small part of the cascade of highly classified information gushing from the early Trump administration. As measured by intelligence community criminal referrals to the Justice Department, the number of serious leaks per year was almost three times higher in the first two years of the Trump administration compared to the Obama administration. 

Any administration would have aggressively investigated these leaks. Such investigations often begin with the collection of metadata—sender and recipient information from email and phone accounts, dates and times of communications, and the like. The metadata can help the government figure out who among the hundreds or thousands of people with access to leaked classified information communicated with the journalists who published the leak. Metadata alone usually cannot make a criminal leak case. But figuring out who spoke to whom is typically an early investigative step. Metadata analysis is often conducted, usually without controversy, on potential government suspects. Very occasionally, and more controversially, the investigators examine journalists’ metadata, which makes narrowing and identifying suspected leakers much easier. 

Journalists claim that these seizures violate fundamental constitutional rights. “If the First Amendment’s guarantee of a free press means anything, it has to mean that the government cannot seize reporters’ phone and email records in the hope of smoking out their confidential source,” scowled the New York Times editorial page last month in response to reports that the Trump administration had seized Times’ reporters’ phone records in 2017. Democrats claimed that the seizure of congressional records in a leak investigation was the latest example of dirty pool by the Trump Justice Department. Rep. Adam Schiff called the subpoenas Trump’s “cudgel against his political opponents and members of the media” and the latest examples of his “politicization of the department.” 

These are large exaggerations. The First Amendment does not prevent the government from getting these records. Most of the records were apparently obtained by grand jury subpoena, and some by a court order.* It turns out that, contrary to early reports, members of Congress were not targeted. Attorney General Merrick Garland declined to “cast[] blame” on anyone because the records were obtained “under a set of policies that have existed for decades.” There were sharp complaints about the so-called gag orders that the DOJ imposed, including under Garland’s watch. But these non-disclosure orders appear to be court-approved and fully consistent with the law. Finally, some have questioned whether the department followed its policies about when to notify journalists about the seizures, but no evidence has yet emerged that it did not.

The legality of the Trump-era records seizures did not, of course, make them a good idea or politically palatable. Journalists and transparency advocates insist that journalists need immunity from these seizures to serve their accountability function. Garland has long supported reporters’ rights, both as a federal judge and when he worked on early Justice Department guidelines to protect journalist records as a special assistant to Attorney General Benjamin Civiletti in the late 1970s. And President Biden clearly doesn’t approve of the seizure of such records. “It’s simply, simply wrong … I will not let that happen,” Biden said, seemingly off the cuff, following a press conference with Korean President Moon Jae-in. 

Thus was born a new policy. A few weeks after Biden’s comment, Garland announced that he would craft new Justice Department rules that would be “the most protective of journalists’ ability to do their jobs in history.” Ten days later, Garland said he would support legislation to protect journalists’ records—a long-sought goal of the press—to make his policy prescriptions harder for future administrations to reverse. Sen. Ron Wyden soon proposed a bill—the Protect Reporters from Excessive State Suppression (PRESS) Act—that would protect journalist metadata unless the information is needed to prevent a terrorist act or to prevent imminent violence, significant bodily harm, or death.

These developments are the latest in a string of impressive victories by the press against the government since 9/11. Charlie Savage of the New York Times recently reported that “government leak hunters have been ratcheting up pressure on the ability of journalists to do their jobs for a generation—a push fueled by changing technology and fraught national-security issues that arose after the Sept. 11, 2001, attacks.” This is only half the story. Yes, leak investigations have grown with the sharp rise in leaks. And the government has occasionally tried to use coercive techniques with journalists in these investigations. But these intermittent efforts against journalists have been self-defeating. Blowback against them has led the government to grant the press greater freedoms from government scrutiny in leak investigations.

When the New York Times revealed the Bush administration’s warrantless wiretapping program in 2005, some argued that the newspaper could be prosecuted. This was a plausible contention. Section 798 of the Espionage Act prohibits harmful publication of “any classified information … concerning the communication intelligence activities of the United States.” And some noted constitutional law experts believe that a prosecution of journalists under this law would pass First Amendment scrutiny. The Bush administration floated the possibility but never followed through. It just seemed too controversial. 

That and subsequent government refusals to prosecute journalists for publishing information about “communication intelligence activities” left the press confident that it has effective immunity. Following the massive leaks by Chelsea Manning in 2010 and Edward Snowden in 2013, New York Times lawyers became “convinced that there are no legal consequences from publishing leaks,” according to Times Assistant General Counsel David McCraw in 2016. 

Some worry that the Trump administration’s indictment of WikiLeaks leader Julian Assange for procuring and publishing the classified documents stolen by Manning might be a precedent to prosecute American journalists. There are resemblances between what Assange did and what U.S. journalists do. But there are also differences, including Assange’s aims and threats against the U.S. government, his relatively lax curation practices, and his operation outside the United States. And indeed, in announcing the Assange indictment the Trump Justice Department confirmed the strong norm against prosecuting U.S. journalists for publishing classified secrets. “The department takes seriously the role of journalists in our democracy,” said Assistant Attorney General John Demers at the time. “It is not and has never been the department’s policy to target them for reporting. But Julian Assange is no journalist.”

The government also now refuses to force journalists to reveal their sources even though it has the power to do so. In 2005, Judith Miller went to jail for 85 days for refusing to reveal her sources to federal prosecutor Patrick Fitzgerald in the Valerie Plame affair. Nine years later, Attorney General Eric Holder declined to require New York Times journalist James Risen to testify in a leak case against ex-CIA officer Jeffrey Sterling even after a court ruled that Risen could be compelled to do so. The Trump administration did not even try to use this lawful but controversial and now-taboo technique.

The government has also been steadily tightening restrictions on seizing reporters’ metadata records. The Obama administration found itself in hot water after it was revealed that the Justice Department had seized metadata records from the Associated Press and Fox News chief Washington correspondent, James Rosen, in two different leak investigations. Under enormous political pressure by the press, and after President Obama ordered a review of Justice Department policy because “journalists should not be at legal risk for doing their jobs,” Holder backtracked. Working with elite journalists, he twice raised the bar on Justice Department guidelines for seizing journalists’ records. The Holder regulations are now seen as far too lax, as Garland’s pledge makes plain. 

It is hard to know where precisely the new policy will land. Garland has committed to a ban on using “compulsory process to obtain reporters’ source information when they are doing their jobs.” This leaves open the hard issue of who counts as a reporter and what counts as their job. Is WikiLeaks covered by the ban, for example? Whatever the answer to this question, there is no doubt that the new policy gives journalists more protection than before, thus continuing a two-decade trend.

These changing attitudes about targeting the press in leak investigations have developed during an era of unprecedented national security leaks going far beyond the ones about Trump. 

Edward Snowden disclosed hundreds of sensitive signals intelligence programs and techniques after exfiltrating an estimated 1.5 million documents in 2013. The Manning leaks, revealed in 2010, were at the time unique in their scale, exposing more than 700,000 documents. Since 9/11, journalists published many stories that contained highly classified information about covert actions, the identities of covert officers, and surveillance programs beyond the ones revealed by Snowden. Several other leaks reached the public without mainstream media mediation. A mysterious group called Shadow Brokers dumped online the National Security Agency’s most potent offensive cyber tools. (Whether they were stolen by outsiders or insiders remains unclear.) And WikiLeaks published descriptions of numerous CIA offensive cyber tools—most likely stolen by an insider—in what the government asserted “was the single biggest leak of classified national defense information in the history of the CIA.”

These leaks undermine expensive or vital programs and tools, dry up sources of information, make new classified operations harder, disincentivize information sharing by allies (like the Dutch intelligence service that turned over the Russia information that leaked), and can get people killed. The intelligence community is apoplectic about them. Why, then, does the government remain so leaky given the importance of the secrets and the costs of their disclosure? 

The main reason is that the secrecy system is too big. Whether a piece of information is deemed classified depends on a government official’s judgment about its degree of “damage” to national security if disclosed. Many matters, like the content of foreign intelligence intercepts, are properly classified. But government officials have broad discretion in wielding the secrecy stamp and large incentives to classify matters that should not be classified. “There is widespread, bipartisan recognition that the Government classifies too much information and keeps it classified for too long,” said the Public Interest Declassification Board in 2020. 

And too many people have access to these secrets. More than 4 million people are cleared for access to classified information. As former CIA Director Richard Helms noted, “the probability of leaks escalates exponentially each time a classified document is exposed to another person.” Such leaks are more likely to occur, and more likely to be in the public interest, when a program is morally or legally controversial, as so many have been since 9/11. The digitization of secrets makes them significantly easier for insiders and foreign adversaries alike to steal and distribute. The ever-growing dependence on private contractors with access to secrets has exacerbated these problems. 

Not all classified information leaks result from theft. Some are authorized and many more are peddled by senior officials in the gray area of uncertain authorization for any of a number of reasons. As Max Frankel famously argued a half-century ago, and as David Pozen has more recently explained in detail, senior government officials gain all sorts of benefits—sometimes personal, sometimes government-wide—from strategic leaks of classified information. They leak all the time and do not honor classified information rules. When President Obama defended Hillary Clinton’s lax handling of classified information by saying that “there’s classified, and then there’s classified,” he underscored the disrespect with which senior officials often treat the secrecy system.

The government has taken internal steps to address leaks, such as standing up anti-leak taskforces, implementing continuous security evaluations for cleared personnel, improving whistleblower protections, imposing partial restrictions on flash drives, installing wireless intrusion detection devices, and sometimes employing a “two-person rule” for accessing highly sensitive information and for copying classified information onto portable devices. 

But these are modest, halting actions given the gravity of the problem.  The government continues to create secrets on a gargantuan scale and to give very wide access to them, including to people in the private sector. And it has failed altogether to update its secrecy management systems. A “majority” of federal agencies “rely on antiquated information security management practices” that have “not kept pace with the volume of digital data that agencies create,” the Information Security Oversight Office reported in 2019.  

In sum, the government has proven itself fundamentally unserious about stopping what it claims are devastating national security leaks. 

This is the background against which to assess the latest round of government restraint in addressing press publication of national security leaks. One might see the Justice Department’s backdown on seizing journalist records as another sign of government unseriousness about stopping leaks. But the backdown will have the tiniest of impacts on leak investigations, and an even tinier impact on the leaks.

The time to protect secrets is before they leak. Once leaked, there is relatively little that the department can do. It has brought several successful prosecutions of leakers in the last decade. But compared to the avalanche of leaks, the number of prosecutions is small, and their general deterrence effect appears small as well. Leak investigations require large resources. The department can investigate only a fraction of the serious leaks reported by the intelligence community. In only a small fraction of these cases has the government sought access to journalist records. And only a fraction of the cases where journalist records are sought result in a successful prosecution.

It is unclear whether technological advances—encrypted communications, digital storage, digital tracking, and the like—favor the leakers or the investigators on balance. But leak investigations tend to succeed when the leaker does something imprudent, such as sending printed hard copies of highly classified documents to a publisher, which in turn sent them to the NSA; or sharing classified information with a girlfriend, who published the information; or divulging classified information to a reporter in an unencrypted email; or bragging to third parties about being the source of the leak.

Most leakers of highly classified information are not so imprudent, and communicate with journalists, if at all directly, with greater care. Metadata analyses that try to connect leakers with journalists often fail. And even if a suspected leaker is tied to the journalist, the tie alone does not a prosecution make. The needed evidence of the content of a leak is hard to come by absent a leaker mistake. And many front-page “leaks” are in fact an aggregate of tiny disclosures by many people, any one of which might not be prosecutable if discovered.

For these reasons, the marginal benefit to the Justice Department of seeking journalist metadata on the overall leaking problem is negligible. And the cost to realize this negligible benefit has proven to be large for the department, which is excoriated every time it uses these techniques. Journalists have an incentive to proclaim that the sky is falling when their records are seized. But politicians tend to back up the complaints—not just when they are directed at the Trump administration, but when they were directed against prior administrations as well. At the very least the political pressures appear to be moving in one direction—against using any legal process against journalists who report even the most sensitive classified information. The Garland Justice Department was roundly criticized for its brief continuance of the Trump press processes, but it received no obvious pushback from either party when it announced the ban on seizures. 

Underlying these changing norms is a broad recognition among the political class that many national security leaks serve the public interest on balance by holding the government at least partially accountable for the massive array of secret operations it has conducted since 9/11. Many of the most controversial leaks—involving CIA black site and interrogation programs, government signals collection against Americans, and more—resulted in political and judicial pushback, and reform of the underlying programs. Journalistic disclosure of wrongdoing is “vital to the functioning of our democracy,” said Garland in recent testimony. “It’s part of how you have faith in a democracy.”

Not every highly classified leak discloses wrongdoing or serves the public interest. But not every secret deserves to be classified, and the same government that overclassifies often plays fast and loose with its secrets. Journalists and the executive branch both err in self-serving ways. As the trends of the last two decades show, our democracy increasingly puts a strong thumb on the scale of allowing journalists to publish leaks without consequence as a counterpoint to the executive branch’s choice to act so extensively in secret, and often controversially.


The government is unwilling or unable to halt the flow of classified secrets to the public, but it has done a better (though legally dubious) job of halting the flow of unclassified information that former government officials want to publish. It does so through its control of a process known as “prepublication review.” 

Millions of people with access to highly classified information have promised by contract to allow the government to review future writings before publication to discern whether their works contain such information. In theory, this lifelong duty makes sense. The government has a genuine interest in ensuring that former officials do not disclose classified information in their writings, whether purposely or accidentally. But the process it uses is broken. The contractual duty to submit manuscripts for prepublication review is typically broad and vague, and the government holds all the cards in enforcing the duty.

The government has enormous discretion in determining what is classified in a manuscript and even in determining what must be submitted for review. As I have learned from personal experience, and as many others can attest, the government’s prepublication decisions are sometimes informed by political considerations and “extreme arbitrariness,” as former CIA Director Stansfield Turner put it. The government often refuses to abide by its own contracts and regulations on prepublication matters. This is especially so of the legal time limits on its review, which the government regularly exceeds, usually without informing the author when its review will be complete, thus making publication deadlines hard to meet or plan. 

While some questionable prepublication decisions are politically motivated, most result from inadequate resources. Too many matters are deemed classified, and too many people going back decades have a contractual duty to preclear publications for the government to conduct reviews efficiently. More and more people have sought preclearance review, overwhelming the system. For example, the CIA reviewed 42 manuscripts in 1977, 148 in 1980, between 200 and 400 each year from 1980-2003, more than 1,800 in 2010, and approximately 2,500 in 2011. That number is surely higher today. It is very difficult for understaffed review boards to pore over all these pages in a short time frame to determine which of the government’s innumerable classified secrets they might contain. Mistakes and delays are the inevitable result.

Whatever the reason, the government, in the name of protecting national security, often improperly censors or delays publication of speech that is vital to democratic deliberation about national security policy. And it backs up this censorship with threats of criminal penalties or civil suits to seize author royalties. Most former officials acquiesce to the government’s censorship (or are chilled from writing in the first place) because they don’t want to run these risks and don’t want to jeopardize future government opportunities that turn on security clearances. 

John Bolton was a rare exception; he fought back. Bolton wrote a memoir of his time as national security adviser called The Room Where It Happened. After a lengthy prepublication review, and many cuts to his manuscript, the career intelligence official in charge of the NSC process informed Bolton that she “had completed her review and was of the judgment that the manuscript draft did not contain classified information.” 

At that point Trump political officials exploited the government’s discretionary power over prepublication review to try to stop Bolton from publishing. The new senior director for intelligence at the NSC, Michael Ellis, a political appointee, immediately started his own review and determined that the book still contained classified information after all. This was a suspicious conclusion since Trump was openly furious about the book, the book had already been cleared by an experienced career official, and Ellis was a Trump loyalist who didn’t complete his mandatory training “in proper classification (including avoidance of overclassification)” until after he ruled that Bolton’s book contained classified information.

Almost everyone would have acquiesced, delayed publication, cut the newly identified classified information, and waited indefinitely for government clearance to publish. But Bolton went ahead and published without permission, and the government sued him. Judge Royce Lamberth rejected the government’s request to enjoin publication, but he tipped his hand on the government’s request for Bolton’s royalties when he said he was “persuaded that Defendant Bolton likely jeopardized national security by disclosing classified information in violation of his nondisclosure agreement obligations.” The government also opened a criminal investigation against Bolton.

Last month the Garland Justice Department ended the civil lawsuit and dropped the criminal investigation. It remains unclear why it did so, but it was probably related to Judge Lamberth’s decision to make Trump officials “answer questions under oath about how they handled Mr. Bolton’s book and how Mr. Trump pressured them,” as the New York Times put it.  Those answers may have revealed a corrupt process behind the Ellis machinations. Whatever it was, the Garland Justice Department decided that it could not defend the Trump administration’s actions against Bolton, in pointed contrast to matters on which it has sided with Trump-era legal positions.

This was a victory for Bolton, but the government’s prepublication review process won as well. The failed case against Bolton was an exception that highlighted the rule. The lawsuit failed because political actors screwed around more obviously than usual in the process in a very high stakes contest against an experienced former government official with the resources and incentives to call their bluff. But the Bolton saga also makes clear that the government has very powerful tools to enforce its often-arbitrary prepublication process—a process that Judge Lamberth endorsed despite its flaws before the government dropped the case. Many will continue to kowtow to the flawed process because the costs of not doing so are so high. Public deliberation about the operation of government in important areas related to war and national security will continue to suffer.


This American secrecy system is unique among democracies. No other democracy has an intelligence bureaucracy and secrecy system that approaches the size of the U.S. government’s. No other democracy leaks classified information to nearly the same degree, or provides such vigorous protections for journalists to publish classified secrets. And no other democracy imposes such broad prior constraints on former official publications in the name of protecting classified secrets. The protection for journalists can be justified as a counterpoint to the size and manipulation of the secrecy system. But every other element is dysfunctional.

The solutions to these dysfunctions have been well rehearsed. The Defense Department’s Coolidge Committee warned in 1956 that “overclassification has reached serious proportions,” and urged reforms to raise barriers to classification and reduce barriers to declassification. Variations of these warnings and recommendations have been repeated for decades by one government study after another. Yet the secrecy system grew and grew. “Generations of critics have risen to attack, bemoan, lampoon, and correct the excesses of government secrecy,” notes Steven Aftergood of the Federation of American Scientists. “Only rarely have they had a measurable and constructive impact.” 

The same is true for pre-publication review. The path to improvement is pretty straightforward: clarify, tighten, and make uniform the criteria for prepublication review, impose binding deadlines to complete review, significantly increase agency resources to conduct review, and require reasons and an informal appeal when an agency censors. A 2016 House Intelligence Committee report expressed concern about the pathologies of prepublication review, suggested reforms, and issued a nonbinding directive to the Director of National Intelligence to “issue an IC-wide policy regarding pre-publication review” to address these reforms within 180 days of the enactment of legislation related to the report. Five years later, the DNI still has not issued the policy. 

The U.S. secrecy system has grown too big to fix. It is too sprawling across too many agencies and private contractors who are too interested in maintaining their secrecy prerogatives as they defend the nation. Congress could in theory fix some of the system’s problems by legislation, but has never come close to doing so. The person who could order the needed root and branch changes is the president, who possesses constitutive control over the secrecy system. But the reform transition would be disruptive to the national security system, and would redirect resources and focus from the everyday imperatives of defense. National security would benefit overall from these reforms, probably a lot. But unfortunately, it is hard to imagine a president who would fight the bureaucratic fights and absorb the significant transition costs. 

So the government will continue to leak sensitive classified secrets. And at the same time it will continue to suppress important non-classified speech about the operation of government in a context, national security, where informed public debate is essential. Only in America.

Correction, July 15: The article has been updated to clarify that the journalist records obtained by the Justice Department were pursuant to court order and grand jury subpoena. 

Jack Goldsmith is a Harvard law professor and senior fellow at the Hoover Institution, and worked in the George W. Bush administration.