Skip Navigation

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.