Skip Navigation
Research Report

Reducing Overclassification Through Accountability

Published: October 5, 2011

Classification is one of the most important tools our government has to keep us  safe. But many secrets protected by the classification system pose no danger to the nation safety. In fact, “overclassification” jeopardizes national security. Excessive secrecy prevents federal agencies from sharing information internally, with other agencies, and with state and local law enforcement, making it more difficult to draw connections and anticipate threats. This report focuses on how to ensure that classifiers comply with existing criteria for classifying documents.


The authority to classify documents exists to protect information that could threaten national security if it got into the wrong hands. It is one of the most important tools our government has to keep us safe. But many secrets “protected” by the classification system pose no danger to the nation’s safety.

On the contrary, needless classification—“overclassification”—jeopardizes national security. Excessive secrecy prevents federal agencies from sharing information internally, with other agencies, and with state and local law enforcement, making it more difficult to draw connections and anticipate threats. The 9/11 Commission found that the failure to share information contributed to intelligence gaps in the months before the September 11, 2001, attacks, cautioning that “[c]urrent security requirements nurture overclassification and excessive compartmentation of information among agencies.”

Overclassification also corrodes democratic government. Secret programs stifled public debate on the decisions that shaped our response to the September 11 attacks. Should the military and CIA have used torture to extract information from detainees in secret overseas prisons and at Guantánamo Bay? Should the National Security Agency have eavesdropped on Americans’ telephone calls without warrants? Even leaving aside the legality of these measures, whether to use torture or to forego the use of warrants are questions that, in a democracy, properly belong in the public sphere. Classification forced the nation to rely on leaked information to debate these questions, and to do so well after torture and warrantless surveillance programs were in place.

Overclassification is rampant, and nearly everyone who works with classified information recognizes the problem. In 1993, Senator John Kerry, who reviewed classified documents while chairing the Senate Select Committee on POW/MIA Affairs, commented, “I do not think more than a hundred, or a couple of hundred, pages of the thousands of [classified] documents we looked at had any current classification importance….” And two years later, Donald Rumsfeld, while noting that disclosure of truly sensitive information can put lives at risk, acknowledged, “I have long believed that too much material is classified across the federal government as a general rule.”

Government statistics bear out these assessments. When a member of the public asks an agency to review particular records for declassification (through a process called “mandatory declassification review”), 92 percent of the time the agency determines that at least some of the requested records need not remain classified. But the number of documents reviewed through this process pales in comparison to the universe of documents that, though they may not require classification, remain unreviewed—and thus classified—for many years.

A major theme of this report—and a source of frustration to those who have studied the classification system—is the persistent gap between written regulation and actual practice. Chief executives since Franklin Delano Roosevelt have issued executive orders on classification. Classification authority emanates primarily from these orders, which have long purported to impose common-sense limits, such as a ban on using classification to conceal embarrassing information about government officials. And the current order—Executive Order 13,526, which President Obama issued in December 2009—includes further limits, such as a requirement that records not be classified if significant doubt exists about the need for secrecy. In practice, however, such limits too often fall by the wayside. As a Senate Commission chaired by Daniel Patrick Moynihan found, “Any policy, including on classification and declassification, is only as good as its implementation.”

This report focuses on improved implementation, i.e., how to make sure that classifiers comply with existing criteria for classifying documents. It does not address ways in which the classification system could be improved by changing those criteria, such as revising agency classification guides—which govern many classification decisions—to eliminate classification categories that are outdated, unnecessary, or imprecise; requiring classifiers to weigh national security risks against the public interest in disclosure; or amending the National Security Act of 1947 to clarify that “intelligence sources and methods” may be classified only if their disclosure would harm national security. Measures to improve the substantive criteria for classification will form a critical piece of any successful reform effort, and their omission from this report should not be taken as an assessment of their relative importance. But the widespread failure of classifiers to comply with existing rules suggests that changing them will have little effect until we understand and address the persistent gap between rules and reality.

This report concludes that the primary source of the “implementation gap” is the skewed incentive structure underlying the current system—a structure that all but guarantees overclassification will occur. Numerous incentives push powerfully in the direction of classification, including the culture of secrecy that pervades some government agencies; the desire to conceal information that would reveal governmental misconduct or incompetence; the relative ease with which executive officials can implement policy when involvement by other officials, members of Congress, and the public is limited; the pressure to err on the side of classification rather than risk official sanctions or public condemnation for revealing sensitive information; and the simple press of business, which discourages giving thoughtful consideration to classification decisions. By contrast, there are essentially no incentives to refrain from or challenge improper classification. After all, classification is an easy exercise that can be accomplished with little effort or reflection; those who classify documents improperly are rarely if ever held accountable—indeed, there is no reliable mechanism in place to identify them; classifiers receive insufficient training in the limits of their authority; and those who have access to classified information are neither encouraged to challenge improper classification decisions nor rewarded for doing so.

In order to succeed, any effort to reduce overclassification must address this problem of skewed incentives. The final chapter of this report sets forth a reform proposal that would rebalance existing incentives, primarily by introducing accountability into the classification system. The proposal consists of six main parts:

  • When classifying documents, officials would be required to complete short electronic forms in which they would provide explanations for their classification decisions.
  • In each agency with classification authority, the Office of the Inspector General would conduct “spot audits” of classifiers, identifying those who exhibit serious tendencies to overclassify and subjecting them to periodic follow-up audits.
  • Successive unsatisfactory audit results would result in mandatory escalating consequences for the individual classifier, agency management, and the agency itself.
  • Agencies would be required to spend at least eight percent of their security classification budgets on training and to obtain approval of their training materials from the government office that oversees classification.
  • Derivative classifiers (those who carry forward classification decisions made by others) would be “held harmless” if they failed to classify information whose status was ambiguous.
  • Agencies would establish procedures to allow authorized holders of classified information to challenge classification decisions anonymously, and those who brought successful challenges would be given small cash awards.

We recommend that this proposal be implemented as a pilot project at one or more agencies. This could be accomplished largely if not exclusively through executive order and implementing regulation. The results of the project should be closely tracked and evaluated to assess both its benefits and its costs. If the proposal yields the expected dividends, it could be expanded through legislation. One thing is certain: the status quo is untenable. The classification system must be reformed if we are to preserve the critical role that transparent government plays in a functioning democracy.