FAS | Government Secrecy | Library || Index | Search | Join FAS

Security Classification of Information: Table of Contents

Chapter 6.


The purpose of this chapter is to describe standards that can be used when the risks of information disclosure are balanced against the benefits. This balancing is the final major step in deciding whether information should be classified.

The following section of this chapter describes the rationale for balancing classification costs against classification benefits when making classification decisions.* Such balancing helps ensure that information is not classified unless information-disclosure risks exceed information-disclosure benefits. The next section discusses the need for standards to assist in this balancing process. Subsequent sections of this chapter describe existing standards (rules) that have been used to assist in balancing competing interests and thereby to guide decision-makers to reliable and consistent conclusions. Those rules include long-established rules of persuasion used in our legal system and balancing principles used when atomic energy information is declassified. The final section of this chapter proposes a set of standards to be used when information-disclosure risks and benefits are being balanced.


National Security Information

NSI, defined by E0 12356, is information whose "unauthorized disclosure, either by itself or in the context of other information, reasonably could be expected to cause damage to the national security."1 If unauthorized release of information reasonably could be expected to damage our national security, then it seems that EO 12356 requires that the information be classified. Even though classifying information may cause several kinds of adverse national impacts,2 EO 12356 does not explicitly require classification benefits (the damages to national security which are prevented by classifying the information) to exceed classification costs (the benefits to the nation of having the information unclassified) before information can be classified. That EO requires only consideration of the damage caused by unauthorized disclosure. However, some balancing of benefits and costs is implicit in many classification decisions. Certainly, information whose unauthorized disclosure would cause only minor damage to the national security should not be classified if major costs to our nation result from that classification.

Balancing information disclosure risks against benefits (classification benefits and costs) before reaching a classification decision is not a new idea. Balancing, for some NSI classification situations, was sanctioned by a 1964 DoD directive, as follows:

In certain circumstances, it may be necessary to weigh the benefits which would accrue to the United States generally from the unclassified use by other government agencies or commercial interest of information which is classified or otherwise classifiable under Executive Order 10501, against the military or technological advantage which would be gained or retained by classification or continued classification of the information. Where a net advantage to the United States distinctly can be ascertained beyond a reasonable doubt, that factor should be considered in reaching the classification determination. In such a case, final determination to withhold classification or to declassify shall be made only by the Secretary of Defense . . .3

A broader scope of the information disclosure benefits to be used in balancing classification costs and benefits was suggested in 1966:

The requirements for national defense in an absolute sense are not ends unto themselves but must be balanced against other necessarily competing requirements such as justice, liberty, and general welfare.4

A 1970 DoD report noted that "classification of information has both negative as well as positive aspects" and mentioned some of those negative aspects.5 Balancing, for all NSI classification decisions, was mentioned in 1971 as a new DoD classification policy element.6, 7, 8 This was described by a DoD representative in a 1971 hearing before a House of Representatives Subcommittee on Foreign Operations and Government Information as follows:

In October 1970 the Secretary of Defense announced the establishment of a new policy which provides that:
All classification decisions will be reached only after careful consideration of the advantages of open circulation after public release approval of the information against the advantages to a potential enemy. Even where security classification is clearly required, it should be retained for the minimum amount of time considering the degree of sensitivity, cost and probability of compromise.
This new policy means that security classification decisions will be reached only after consideration of competing advantages and disadvantages. In the past major emphasis for classification has generally been placed on the possible benefits of the information to potential enemies without consideration of the benefits which could accrue to the U.S. Government, industry and the domestic community, and our allies through open and effective technology dissemination. Now, both reasons must be considered in making the classification decision within that area.6

More recently, balancing has been implicitly sanctioned in a 1986 DoD handbook for writing classification guidance. That handbook mentioned the need to consider the "net national advantage" in deciding whether information should be classified as NSI.9 Also, DoD's June 1986 Information Security Program Regulation states that "both advantages and disadvantages of classification must be weighed" before making a classification decision.10 Thus, there appears to be support within DoD for balancing classification costs and benefits prior to making classification decisions.

EO 12065, the predecessor to EO 12356, explicitly directed that a balancing test be used for some declassification decisions:

It is presumed that information which continues to meet the classification requirements in Section 1-3 requires continued protection. In some cases, however, the need to protect such information may be outweighed by the public interest in disclosure of the information, and in these cases the information should be declassified. When such questions arise, they shall be referred to the agency head, a senior agency official with responsibility for processing Freedom of Information Act requests or Mandatory Review requests under this Order, an official with Top Secret classification authority, or the Archivist of the United States in the case of material covered in Section 3-503. That official will determine whether the public interest in disclosure outweighs the damage to national security that might reasonably be expected from disclosure.*, 11

This balancing test caused problems to the government in litigation under FOIA12 requests for information.13 The major problem was said to be that judges began to be willing to "second guess on the question of whether information would damage the national security, notwithstanding their admitted lack of expertise in the particular subject matter area."14 Prior to EO 12605, judges were said to be unwilling to engage in such a review.14 Therefore, the balancing test was dropped from EO 12356 "not to take away our discretion to consider the competing interest but rather for us to avoid these problems in litigation."+, 14, 15

Support for balancing information disclosure risks and benefits has come from the Director of the Information Security Oversight Office (ISOO). This office is responsible for implementing and monitoring the nation's information security program established by EO 12356.16 The ISOO director has stated that balancing the competing interests for disclosing and protecting information was inherent in the classification process.14

Since balancing classification costs and benefits is inherent in the classification process, then it should explicitly be required in the applicable EO to ensure consistency throughout the government in classifying NSI.

Atomic Energy Information

Strong support for requiring a balancing process in classification decisions is found in U.S. policy for declassifying atomic energy information. RD, a category of classified atomic energy information, is the most stringently protected of our nation's classified information. RD was established by the Atomic Energy Act of 1946,17 and its protection was entrusted to AEC.18 Only an action by the five-member AEC could declassify RD. The AEC very early recognized the need to balance competing interests when making declassification decisions.*

A 1950 AEC report stated the following:

The secrecy classification to be applied to any item of information thus depends upon a balance between the value of that information to inimical interests and the value expected to accrue to the United States through its dissemination.19

A 1956 AEC report mentioned that, in classification decisions+ involving RD, "a fine balance must be struck between protecting information which is of national security significance and providing the maximum assistance to peaceful applications [of atomic energy]."20

More recently, the DOE policy has been expressed as follows:

The principle underlying classification and declassification policy for atomic energy information is to achieve a balance between two aims: (1) assuring the common defense and security by controlling the declassification of information concerning the military aspects of atomic energy; and (2) promoting the dissemination of scientific and technical information relating to the peaceful applications of atomic energy consistent with the common defense and security. However, the Atomic Energy Act states that the paramount objective of protecting the common defense and security must be observed. A determination on the declassification or downgrading of information can proceed only after a careful balancing of the risk to the common defense and security against the benefit to our own program of the proposed action.21 (emphasis added)


Classification decisions are not, in practice, strictly limited to a consideration of only the damage to national security that could reasonably be expected to be caused by disclosure of the information under examination for classification. The costs of classification to the nation (which are the national benefits of information disclosure) are also considered. A balancing of classification's costs and benefits may show greater national costs than benefits. In that case, the net national advantage will favor keeping the information unclassified even though there is a reasonable expectation that some damage could result from disclosure of that information. Information should be classified only when the resulting benefits to national security exceed the accompanying costs to the nation of classifying that information.


Although information-disclosure risks and benefits must be balanced before a proper classification decision is reached, the classification decision process is generally much less certain than the word balancing implies. Balancing suggests a finely tuned process, where slight differences in the materials being compared will tilt the balance to one side or the other. Balancing suggests the possibility of reaching completely objective classification decisions with scientific precision. Unfortunately, classification decisions frequently are not obvious. Information-disclosure risks and benefits cannot always be reduced completely to uncontroverted, quantified facts. There are always uncertainties, sometimes major, in the estimated risks and benefits.

Because of uncertainties associated with classification costs and benefits, some guidance is needed to assist in the balancing of those costs and benefits. For example, in some situations the balancing guidance should require that there be significant apparent differences between costs and benefits before the information is classified. That is, when damage caused by unauthorized release of information may be severe, a balancing standard might require that the classifier be certain beyond a reasonable doubt that information disclosure benefits outweigh information disclosure risks before determining that the information is not classified. When the consequences of a wrong classification decision are less severe, the balancing rule might require that the classification decision be based on clear and convincing evidence or by the preponderance of the evidence, which are lesser standards of certainty. Use of balancing standards will help ensure that reasonable and consistent classification decisions are made. DoD has suggested a reasoned judgment standard—"Reasoned judgment shall be exercised in making classification decisions."10

The balancing rules suggested above for classification decisions are analogous to our legal system's rules for "weighing the evidence" in lawsuits. The following section of this chapter describes rules that are used by judges and juries to balance evidence when they are deciding the outcome of civil or criminal trials.


Civil and criminal trials represent circumstances where a "trier of fact" (a judge or jury) must weigh (balance) the evidence and decide a lawsuit. Those triers of fact must decide, based on evidence presented to them, the guilt or innocence of persons accused of crimes or the winners and losers in civil lawsuits. The facts (evidence) in those cases are seldom unequivocal. Margins of error in those facts must be anticipated in the search for an accurate decision. Our legal system has developed, over the centuries, balancing rules to help judges and juries reach decisions in lawsuits. Those rules also help ensure that those decisions are consistent.* Some of those rules provide guidance with respect to how convinced the trier of fact must be to reach a verdict. Other rules provide guidance on what a judge or jury must do if they are in doubt about a decision. Those rules are different for civil and criminal lawsuits.

As discussed in Appendix A on trade secrets and classification matters, principles adapted from the legal system are especially reliable for use in classifying information because the legal system's rules have been rigorously debated and have withstood the tests of time. Therefore, it is appropriate to consider our legal system's rules for balancing evidence and perhaps to adapt some of those rules for use in making classification decisions.*

For most civil cases, the standard of persuasion is by a preponderance or greater weight of the evidence.22, 23 This standard refers to the "convincing force of the evidence"24 and means that the "existence of the contested fact is more probable than its nonexistence."22 There are a few types of civil cases where the standard of persuasion is greater, where clear and convincing evidence is required. That higher standard is applied in those civil cases where there is special danger of deception or where public policy does not favor a particular type of claim.25

For trade secret matters, the proper standard of proof, which a trade secret owner- plaintiff must meet, is that "a trade secret must be established by a preponderance of the credible evidence introduced at trial."26 This is said to be "something more than equivocal balanced evidence" and something less than "clear and convincing evidence."27 Because it is hard to prove trade secret misappropriation by direct evidence (i.e., direct evidence is generally not available), the plaintiff usually has to rely on circumstantial evidence from which inferences can be drawn.28 When no direct evidence is available, some courts have said that the circumstantial evidence must be "clear and convincing."29

For criminal cases, where a person's liberty is at stake or where other significant penalties may be imposed on a defendant, the standard of persuasion is higher than for civil cases. Our society's policy is to minimize the probability of convicting an innocent person. For criminal cases, the generally accepted standard is "proof beyond a reasonable doubt."+, 30 The more rigorous persuasion standard required for convicting the defendant in criminal cases, as compared with deciding a winner in civil litigation, is also reflected in the degree to which direct and circumstantial evidence is relied on in criminal cases. If the prosecution relies on direct evidence to prove the guilt of the defendant, the standard of persuasion is proof beyond a reasonable doubt. However, if the prosecution has to rely on circumstantial evidence to establish guilt, then the standard of persuasion is sometimes required to be that "the evidence must be so conclusive as to exclude any other reasonable inference inconsistent therewith."31

In administrative hearings, the usual burden of persuasion is "a preponderance of the evidence." However, where major issues are at stake, such as in a deportation hearing, the standard is "clear, unequivocal, and convincing evidence."32

A Supreme Court decision of relevance to establishing balancing standards for use in classification matters was rendered in a civil suit against the United States in which the plaintiff wanted the government to produce specific documents. The government asserted that those documents contained military secrets and could not be made public. In deciding whether those documents should be provided to the plaintiff, the Court's standard was whether there was a "reasonable danger" that release of the documents would reveal military secrets that should not be divulged for national security reasons.33

Table 6.1 summarizes some of the balancing rules used in our legal system to help decide the outcome of different types of civil and criminal lawsuits.

Table 6.1. A summary of some "balancing" rules used by our legal system

in deciding civil or criminal lawsuits


Applicable balancing rule or standard

Civil Matters

Ordinary civil litigation

Preponderance of the evidence

Litigation where there is a special danger of deception or where public policy does not favor a particular type of claim

Clear and convincing evidence

Trade secret litigation

Preponderance of the evidence

Trade secret litigation where proponent relies on circumstantial evidence

Clear and convincing evidence

Criminal Matters

Ordinary criminal cases

Proof beyond a reasonable doubt

Criminal cases where prosecution relies upon circumstantial evidence

Proof must be strong enough to exclude any other reasonable inference inconsistent therewith

Administrative Hearings

Ordinary administrative matters

Preponderance of the evidence

Administrative hearings where major issues are at stake (e.g., deportation)

Clear and convincing evidence


Balancing standards have been used when atomic energy information (which is "born classified") is being declassified since the earliest days that declassification of this information was considered. Balancing standards for declassification are, of course, particularly relevant for use in developing balancing standards for classification.

When the Manhattan Project's Tolman Committee developed the first guidelines for declassifying atomic energy information, this information was divided into three classes (areas):

Class I includes basic scientific information which has little direct application to the problems of production or military utilization. Class II includes certain basic scientific information which would be of great value to the development of science but which has a direct bearing on production or military utilization. It also includes technological information which would be of great importance for the peacetime utilization of atomic energy but which also has importance for production or military utilization. Class III includes information which has immediate application to the problems of military utilization but for the most part has little application to the development of science or to peacetime utilization.34

"Information recommended for immediate declassification" was contained in Class I (information developed from basic scientific research, little or no national security importance). Class II ("dual-use" scientific or technical information) contained that "information whose declassification would conduce to the national welfare and to long term national security, so that the date of declassification should depend on estimates as to the probability and imminence of war." Class III (information primarily useful only for military applications) contained "information not at present recommended for declassification and whose declassification should await a real reduction in the threat of atomic warfare."*, 34

The Tolman Committee's declassification principles (balancing standards) were essentially as follows (arranged from most stringent to least stringent) for the three classes (areas) of information:

Balancing principles are currently used by DOE when declassifying atomic energy information. The basic DOE declassification policy is to achieve a balance between two competing aims:

There are currently several major areas of atomic energy information—weapons design, isotope separation, naval reactors, etc. Each area has its own balancing standard for use when information in that area is considered for declassification. Use of a different balancing standard for each atomic energy information area reflects a conclusion that information in some areas is more important to the national security than information in other areas. Consequently, more-stringent standards should be used when information in those more- important areas is being declassified. The balancing standards used when atomic energy information is being declassified are described in more detail in Chapter 11.

A summary of the balancing standards that have been used (some are not now being used) by the AEC and DOE when balancing information disclosure risks and benefits in the different AEI areas (arbitrarily designated as 1 through 8) is given below.35, 36, 37 Those paraphrased standards, arranged in order from the most restrictive to the least restrictive, are as follows:

The standard for AEI 1 is clearly the most stringent—declassify only if such declassification would provide no help to inimical nations. The effect on our own progress is not a factor in this area. Use of this standard probably means that declassification should occur only if adversaries already have the information and they know that it is reliable information. This is perhaps the only situation where providing information to someone would not help them.

On first examination, the standard for AEI 2 does not appear to be very helpful. It allows declassification when clearly in the national defense interest but does not indicate what constitutes such interest. However, if this standard is interpreted to be "declassify only when the contribution to national defense clearly outweighs the risk of disclosure," then it is more useful. Assuming that this interpretation of the standard for AEI 2 is correct, then the standards for AEI 2, 3, 4, and 5 are quite similar. For each of those areas, the balancing standard is "clearly outweighs." The benefit to be considered becomes broader in going from AEI 2 to AEI 5—from "contribution to national defense," to "contribution to scientific and peaceful applications of atomic energy," to "contribution to our [general] progress" (where "general" is assumed to be part of the AEI 4 benefit description), to "advantage to ourselves." In all four of those areas, declassification should occur only when such action clearly helps us more than it helps our adversaries.

The standards for declassifying AEI 6 and 7 are similar. The balancing standard in both instances seems to be merely a preponderance of benefits over risks. In both instances it seems acceptable to help the adversary by the declassification. The difference appears to be that in AEI 6 advancing our technology is to be weighted heavier than helping an adversary, while in AEI 7 the two competing interests have equal weights.

AEI 8 seems to be the least stringent criterion. Information can be declassified if it is not vital to our defense activities or technologies but is important to basic science or industrial applications of nuclear power.


Proposed Standards for Balancing Information-Disclosure
Risks and Benefits

The major risks and benefits of information disclosure are discussed in detail in Chapter 5 and summarized in Table 5.1. Standards for balancing the risks and benefits of declassifying atomic energy information are described in the immediately preceding section of this chapter. An earlier section discusses rules (standards of persuasion) used in our legal system to assist judges and juries in weighing the evidence in civil and criminal cases. Based on those standards of persuasion from our legal system and from atomic energy declassification experience, five general standards (S1 through S5) have been developed for use in making decisions on the classification of information. Those general standards, to use when balancing the risks to national security of disclosing information against the benefits to the nation of having this information unclassified* (i.e., balancing the risks and benefits summarized in Table 5.1), are presented and discussed below.

This standard corresponds to a principle for the declassification of certain atomic energy information which stated that retarding the progress of an inimical nation (not providing any assistance to that nation) was more important than assisting our own progress. This standard requires a classifier to be very certain that release of information would not help an adversary before the classifier determines that the information can be unclassified. Information- disclosure benefits are not considered (i.e., are implicitly assigned zero weight) in classification decisions involving this type of information. There is no balancing of risks and benefits, only a determination that there are no information-disclosure (declassification) risks. This principle is a "presumed classified" principle of classification. There should be very few areas of information that are so important that they would fall in this category. NSI 4 and 9 (see Appendix B) would certainly be considered in this category,* as would some AEI areas. Certain information in other NSI areas would also be expected to fall in this category.

Under this second balancing standard, information-disclosure risks and benefits are balanced prior to reaching a classification decision. When the classifier is certain beyond a reasonable doubt that benefits outweigh risks, then the information can be unclassified. This is another "presumed classified" balancing standard because the information to which it applies is in areas that are very important to our national security.

The information areas covered by this third standard are the same as those covered by the second standard. The difference in the two standards is that the third standard includes situations where the information has known major commercial or basic science applications (it has major dual-use applications). For that situation, the less stringent "clearly outweighs" standard is more appropriate than the "beyond a reasonable doubt" standard. This third standard implicitly assigns more weight to technology transfer benefits than does the second standard.

This fourth balancing standard applies to information in areas of lesser importance to the national security than encompassed by the first three standards. The majority of information under consideration for classification as NSI is expected to be evaluated under this standard. This is a "presumed unclassified" balancing standard, consistent with a positive action being required to classify information as NSI.

This fifth standard is the same as the fourth standard with respect to the national security importance of the information under consideration for classification. The difference is that the fifth standard applies to information that would make a major contribution toward informing the public on an important national issue. In this situation the information should be classified only if information disclosure risks clearly outweigh information-disclosure benefits. This fifth standard represents a policy that a more rigorous standard should be required to classify information that would have a major impact on public debate of important governmental matters (i.e., more weight is assigned to the benefit of public discussions of the information). The presumption for having the information unclassified is higher than for normal circumstances.

These standards might be affected by the international situation existing when the classification decision is made. That is, the balancing standards might become more stringent (tending toward not disclosing information) if we are at war, or if war seems imminent, as contrasted to an existing or expected relatively peaceful international situation. *, + Alternatively, more weight might be assigned to national security benefits in time of war or international crises, consistent with a society's different values when at war than in peacetime. 38

Another factor that might affect the stringency of the balancing standard used in a particular situation is quality of the evidence on which the classifier must base the classification decision. If the evidence is circumstantial (e.g., intelligence information about an adversary's scientific or technical progress, which frequently is not direct evidence), then perhaps a more stringent standard of persuasion should be used as contrasted to a standard that would be used when the evidence is direct (e.g., information in scientific papers published by an adversary as evidence of the adversary's scientific or technical progress). This would be consistent with our legal system's principles. For example, in certain criminal prosecution situations, the legal system imposes a more stringent balancing standard when circumstantial evidence is presented than when direct evidence is presented. More stringent standards are also imposed on circumstantial evidence than on direct evidence in certain trade secret litigation matters. Those differing standards were discussed previously in this chapter.

For referencing ease, the proposed standards for balancing information-disclosure risks and benefits when classification of that information is being determined are presented in Table 6.2.

Table 6.2. Proposed standards for use in balancing risks and benefits

of information disclosure when classification determinations are being made

Standard number

Description of standard


When the information concerns an area that is crucial to our national security, then the information must be classified unless the classifier is certain beyond a reasonable doubt that release of the information would not help an adversary.


When the information concerns an area that is very important to our national security, but not crucial, and if the information-disclosure benefits outweigh the information-disclosure risks beyond a reasonable doubt, then the information should not be classified.


When the information concerns an area that is very important to our national security, but not crucial, and when the information has major commercial or basic science applications in addition to national defense uses, if the information-disclosure benefits clearly outweigh the information-disclosure risks, then the information should not be classified.


When the information concerns an area that is important to our national security and if the information-disclosure risks outweigh the information-disclosure benefits (a mere preponderance of risks over benefits), then the information should be classified.


When the information concerns an area that is important to our national security and when the information is of major significance to public debate of important national issues, then the information should be classified only if the risks to national security clearly outweigh the benefits of information disclosure.

Table 6.3 summarizes matters of proof in classification decisions (facts to be proven, burden of persuasion, standard of persuasion, and identification of the person who can make the classification decision). This table encompasses all categories of classified information. Further details about RD, FRD, and transclassified NSI are given in Chapter 11.

Table 6.3. Matters of Aproof@ in decisions on classification of information.

Category of classified information

Elements of the classification decision

Restricted Data


Formerly Restricted

Data (FRD)

National Security Information (NSI)





from RD

Originally Classified

by Executive Order

Facts to be proven

That the information falls within the definition of RD [42 U.S.C. '2014(y)] and has not been declassified or transclassified to FRD or NSI

That the information is RD, that it can be transclassified to FRD because it relates primarily to the military utilization of atomic weapons, and that it can be adequately safeguarded as NSI [42 U.S.C. '2162(d)]

That the information is RD, that it concerns the atomic energy programs of other nations, that transclassification is necessary to carry out the provisions of '102(d) of the National Security Act of 1947, and that it can be adequately safeguarded as NSI [42 U.S.C. '2162(e)]

That disclosure of the information reasonably could be expected to cause damage to the national defense or foreign relations of the United States [EO 12356, Preamble]

Burden of persuasion

On the Department of Energy (DOE) Director of the Office of Classification [DOE Order 5650.2B, Chap. II.A, '4.a(5)]

Jointly on DOE and the Department of Defense (DoD) [42 U.S.C. '2162(d); DOE Order 5650.2B, Chap. II.A, '2.c]

Jointly on DOE and the Director, Central Intelligence Agency (CIA) [42 U.S.C. '2162(e); DOE Order 5650.2B, Chap. II.A, '2.c]

On the individual classifying the information

Standard of persuasion

Reasonably determined to fall within the definition of RD

Reasonably determined to be related to military utilization of atomic weapons and can be safeguarded as NSI

Reasonably determined to concern the atomic energy programs of other nations, necessary to carry out the provisions of '102(d) of the National Security Act of 1947, and can be safeguarded as NSI

Reasonable [EO 12356, Preamble]

Also, should depend on the NSI area (see text, this chapter, and Table 6.2)

Person who can make classification decision

Director, Office of Classification [42 U.S.C. '2162(a); DOE Order 5650.2B, Chap. II.A, '4.a(5)]

DOE and DoD, jointly [42 U.S.C. '2162(c); DOE Order 5650.2B, Chap. II.A, '2.c]

DOE and CIA, jointly [42 U.S.C. '2162(e); DOE Order 5650.2B, Chap. II.A, '2.c]

An authorized individual [EO 12356, '1.2]

An Authorized Original Classifier [DOE Order 5650.2B, Chap. V.A]

Presumptions in Balancing Information-Disclosure
Risks and Benefits

There are several situations where presumptions need to be established to aid in reaching classification decisions. One such situation is when the uncertainties in the information- disclosure risks and benefits are so great that there is no evident choice between classifying or not classifying the information. In that case, the presumption should be that the status quo will prevail. For example, information that is under consideration for classification as NSI is initially presumed to be unclassified except for a few NSI areas (see Chapter 5). Therefore, that information should remain unclassified unless the classifier can determine that classification benefits exceed classification costs. * Conversely, atomic energy information is "born classified" and should not be declassified unless the benefits of declassification exceed the risks of declassification. This principle is consistent with the rules of our legal system, where the burden of persuasion is on the party that wants "to change the present state of affairs."39

Another such situation occurs when the classifier has to determine whether adversaries already have the information under consideration for classification. It is usually difficult to prove whether an adversary has the information unless the adversary has published that information.+ The classifier usually has to rely on circumstantial evidence such as intelligence information to decide whether the adversary already has the information. ++ Who has the burden of proof in this situation? Does the proponent for classification have to prove that the adversary does not have the information, or does the proponent for not classifying the information have to prove that the adversary has the information? Trade secret law is helpful in answering this question.

In trade secret litigation, the trade secret owner does not have to prove that others in the trade do not have the trade secret information which is in dispute. Trade secret information is valuable and closely guarded by its possessors (just as adversaries guard their national security information). Therefore, it is difficult for a trade secret owner to prove that no competitor has the same, or similar, information.40 Similarly, it is difficult for a classifier to prove that an adversary does not have the information. Therefore, consistent with trade secret law principles, if there is not adequate proof that an adversary has the information, then the presumption should be that the adversary does not have the information.**

Weighting Factors for Information-Disclosure Risks and Benefits

In the discussions on balancing information-disclosure risks and benefits, it was generally assumed that all of those risks and benefits had equal weight. That is, the discussions assumed (with a few implicit exceptions) that the risks to national security of information disclosure (e.g., assistance to other nations in developing new armaments) carried the same weight as the benefits of information disclosure (e.g., technology transfer benefits). It might be appropriate to assign national security risks more weight than purely economic benefits. Perhaps weighting factors are implicit in the choice of balancing principles. This matter of using different weighting factors for the different types of information-disclosure risks and benefits needs further consideration.

Requirements for De Minimis Damage to National Security

A principle that might be useful in making classification decisions would be to establish a de minimis amount of damage to the national security, a threshold level somewhere above zero, which must be exceeded before information would be further considered for classification. Whenever expected damage from information disclosure was less than this threshold amount, the information would not be further considered for classification. This de minimis principle is based on the fact that every decision to classify information results in certain minimum classification costs. Not classifying information avoids those classification costs. Consequently, the benefits of not classifying information are at least as large as those de minimis costs. Therefore, the damage to national security from information disclosure should at least be greater than those de minimis costs before information should be classified as NSI. Adopting this principle is essentially establishing a policy that certain "trifling" damages to national security will be accepted, * that the United States does not want to expend its resources to avoid such minimal damage. +

Such a principle has its analogy in the legal system, where the courts do not want to expend their limited resources on minor matters. The adage "de minimis non curat lex" expresses the legal system's reluctance to use its resources on small or trifling matters. Trade secret law incorporates this principle because information is not protected as a trade secret if it is only a "trivial advance" or its value is "not significant"; an "unusual competitive advantage" is usually required.41, 42 A de minimis principle has also been suggested as a pragmatic approach with respect to acceptance of de minimis societal risks and focusing limited resources on greater-than-de-minimis risks. 43 A de minimis risk concept has also been used, or suggested for use, in federal regulatory activities (e.g., in determining permissible exposure levels to carcinogens).44

Establishing a principle to accept a certain amount of damage to the national security from disclosure of information, before that information is classified as NSI, is consistent with established U.S. policy for declassifying RD or FRD. RD is our nation's most stringently protected classified information. The Atomic Energy Act of 1954 states that RD can be declassified if it "can be published without undue risk to the common defense and security." 45 Implicit in that declassification policy is the acceptance of some risk when RD is declassified. The statute does not require "no risk" to national security before RD can be declassified. Therefore, our national policy is to accept a level of damage, not to exceed "undue," when RD is declassified.

FRD is the next most stringently protected of our nation's classified information. FRD is not protected as stringently as is RD but is better protected than NSI. FRD can be declassified when it "may be published without constituting an unreasonable risk to the common defense and security."46 Accepting a less-than-unreasonable risk (i.e., accepting a reasonable risk) presumably means accepting more damage than from an undue risk. Consequently, if a scale of damages to the national security caused by publication of information were established, starting with zero damage and increasing in damage toward the right, then the acceptable damage (the "not undue" damage) caused by publication of RD (it was RD before it was declassified) would be somewhat to the right of zero and the acceptable damage (the "not unreasonable" damage) caused by publication of FRD would be even further to the right (more damage).

EO 12356 states that information can be classified as NSI if its disclosure would cause damage to the national security. Because in practice NSI is less stringently protected than either RD or FRD, and because some damages are acceptable from the disclosure of declassified RD and FRD, then it can be concluded that some damages are acceptable before information should be classified as NSI. The damages acceptable before information is classified as NSI are greater than those acceptable from declassification of FRD. That amount of acceptable damage can be thought of as a de minimis level of damage.

Figure 6.1 [not included here] illustrates graphically the relative difference in the threshold level of damage (de minimis damage) reasonably expected to be caused by publication of the three different categories of classified information. The threshold level of damage increases from RD to FRD to NSI, consistent with the relative stringency of information security for those three categories. Also shown in Fig. 6.1, as a matter of some interest, is the information that may be disclosed to persons holding "L" or "Q" clearances (assuming that they have a need-to- know). Persons with L clearance may have access only to Confidential RD but may have access to both Confidential and Secret FDR and NSI. Assuming that granting an individual a particular clearance has some relationship to the amount of damage the government is willing to risk by giving that individual some classified information, then this figure tends to indicate that the threshold level of damages for Secret RD is about the same as the threshold level of damages for Top Secret FRD or NSI. Figure 6.1 also tends to indicate that for all categories and levels of classified information, the threshold level of damages for the three categories has the same value for a particular level (e.g., Top Secret) only for Top Secret FRD and Top Secret NSI (although it is possible that the threshold levels could be the same for Secret FRD and Secret NSI). There are other interesting ramifications of Fig. 6.1. One might ask why an L-cleared person should be trusted with Secret NSI, the unauthorized release of which would cause serious damage to the national security, but not with Secret RD, the unauthorized release of which would cause the same amount of damage to national security. It certainly reflects the very high care with which DOE protects RD, which is consistent with the national security importance ascribed to RD by congress when the Atomic Energy Acts were enacted. Perhaps it also reflects a major lack of agreement between government agencies in what constitutes damage, serious damage, and exceptionally grave damage to the national security (see Chapter 7 for a discussion of classification levels).

The de minimis damage to national security could be set equal to estimated minimum costs of classification (minimum benefits of not classifying information). The costs to be included in this de minimis damage would generally be only the direct costs of classification, not the indirect costs (see Chapter 5 for a discussion of those direct and indirect costs). Those direct costs would include record-keeping and storage costs for the classified documents that would be generated by such a classification decision, the eventual declassification (decommissioning) costs, and other costs that are inescapable when information is classified. Costs might reasonably be estimated for parameters such as level of classification, number of copies distributed, extent of distribution of copies, number and level of personnel clearances required, etc. Estimates could also be made for other classification costs, such as costs of classified procurements. The sum of those costs would then be established as the de minimis cost of classifying an item of information.

When following the de minimis damage policy, a classifier would first make a rough estimate of the expected damage to national security caused by disclosure of the information. That information would not be further considered for classification unless that damage estimate exceeded the de minimis cost. That is, a classifier would not go through the complete decision process for classifying information, as described in this and preceding chapters, unless the de minimis damage was expected to be exceeded when the information was disclosed.

Classification of Basic Scientific Research

EO 12356 states that "basic scientific research information not clearly related to the national security may not be classified." 47 DoD has provided somewhat more extensive guidance in this area:

Ordinarily, except for information that meets the definition of Restricted Data, basic scientific research or its results shall not be classified. However, classification would be appropriate if the information concerns an unusually significant scientific breakthrough and there is sound reason to believe that it is not known or within the state-of-the-art of other nations, and it supplies the United States with an advantage directly related to national security.48

Earlier DoD classification guidance contained additional information about classifying basic research:

Classification also would be appropriate if the potential military or industrial application of the information, although not specifically visualized, would afford the United States a significant military or technological advantage in terms of technological lead time.49
It is DOE policy to conduct as much atomic energy research and development work as possible on an unclassified basis "so as to promote the free interchange of ideas which is essential to scientific and industrial progress." 50 Whenever possible, DOE atomic energy-related research in basic sciences such as chemistry, physics, and the properties of materials is carried out on an unclassified basis. However, in certain areas such as isotope separation, nuclear weapons, and naval nuclear propulsion systems, some basic research is classified.

Sometimes, it has been proposed that a distinction between basic and applied research be used as a basis for classification (e.g., classify the applied research and generally leave the basic research unclassified). However, it is frequently difficult to differentiate between basic and applied research.*

Classification guides occasionally specify that only key or breakthrough scientific or technical information should be classified. "Key information" has been defined as follows:

This definition would be more useful if the last part of (a), "or are difficult or time consuming to duplicate," was eliminated. A "difficult or time consuming to duplicate" criterion is out of place with respect to breakthroughs, which usually originate from creative insights or prepared-for serendipity (breakthroughs are "innovative, not obvious, unexpected"). Breakthroughs are not usually directly connected to expenditures of massive amounts of ordinary scientific or technical effort. Scientific or technological breakthroughs are rare. Examples include discovery of nuclear fission, the transistor, and the maser—discoveries for which Nobel prizes are generally awarded.52

To go from breakthroughs to hardware that is usable for national security purposes usually requires 5 to 10 years or more and hundreds of smaller technological steps.52 About 6 years were required to develop the first atomic bomb, even though it was developed under circumstances where the military implications of the discovery of nuclear fission were immediately recognized by scientists and where three of those years involved a massive, well- directed and well-funded U.S. effort (the Manhattan Project). Therefore, generally there is little reason to classify most breakthroughs in pure research.

A 1982 report, Scientific Communication and National Security (also known as the Corson Report), recommended some guidelines with respect to classification of scientific research in universities (generally considered as basic research). Those guidelines may also be applicable to scientific research in other institutions. That report recommended that information from scientific research carried out at universities be controlled only if it involved a technology that met all of the following criteria:

A classification standard with respect to basic scientific research might be the following:

Only rarely would information from basic scientific research also be key information clearly related to the national security.

Other factors that might lead to classification of basic research include whether the research indicates the direction in which programs are heading, whether it would contribute to feasibility assessments of weapons systems, or whether it reveals specific materials essential to a classified program.


1. Executive Order 12356, Fed. Reg., 47, 14874 (Apr. 6, 1982), Preamble, §1.3(b), §6.1(c). Hereafter cited as "EO 12356."

2. A. S. Quist, "Some Adverse Impacts of Classification," Chap. 6 in Security Classification of Information, Volume 1. Introduction, History, and Adverse Impacts, KCG-1077/V1, Martin Marietta Energy Systems, Inc., Oak Ridge, Tenn., 1989, pp. 85–102. See also the previous chapter of this document.

3. U.S. Department of Defense, Security Classification of Official Information, Instruction 5210.47, §VI.E (Dec. 31, 1964). See also J. P. Gillis, "Panel—Classified Research in the University Environment," J. Natl. Class. Mgmt. Soc., 4(2), 80–85 (1968), p. 83.

4. F. Thomas, "Keynote Address—Classification and Technological Breakthroughs," J. Natl. Class. Mgmt. Soc., 2, 11–19 (1966), p. 19. Hereafter cited as "Thomas."

5. U.S. Department of Defense, Report of the Defense Science Board Task Force on Secrecy, F. Seitz, Chmn., Office of the Director of Defense Research and Engineering, July 1, 1970, p. 1.

6. D. O. Cooke, U.S. Department of Defense, in U.S. Government Information Policies and Practices—The Pentagon Papers (Part 2), Hearings Before a Subcommittee of the Committee on Government Operations, House of Representatives, 92nd Congress, 1st Session, June 28 and 29, 1971, U.S. Govt. Printing Office, 1971, p. 603.

7. G. MacLain, "Updating Security Classification Management and Its Support by DCAS Organization," J. Natl. Class. Mgmt. Soc., 7, 9–14 (1971), pp. 10–11.

8. See also R. E. Coffin, "Keynote Address," J. Natl. Class. Mgmt. Soc., 7, 5–9 (1971), p. 9; L. C. Myers, "Panel on Implementation of Executive Order 11652," J. Natl. Class. Mgmt. Soc., 8, 33–34 (1972), p. 34.

9. U.S. Department of Defense, Department of Defense Handbook for Writing Security Classification Guidance, DoD 5200.1-H, March 1986, §3-3, p. 3-2.

10. U.S. Department of Defense, Information Security Program Regulation, DoD 5200.1-R, Chap. II, §2-200, June 1986. Hereafter cited as "DoD 5200.1-R."

11. Executive Order 12065, Fed. Reg., 43, 28949 (July 2, 1978), §3-303.

12. 5 U.S.C. §552, as amended.

13. U.S. Congress, House of Representatives, Committee on Government Operations, Security Classification Policy and Executive Order 12356, 29th Report by the Committee on Government Operations, House Report 97-731, 97th Cong., 2nd Sess., 1982, pp. 25–26.

14. S. Garfinkel, "An Information Security Oversight Office Overview of Executive Order 12356 and Its Implementing Directive," J. Natl. Class. Mgmt. Soc., 18, 17–23 (1982), p. 21. Hereafter cited as "Garfinkel."

15. A. Adler, "ISOO Panel—Why Do We Have So Much Classified Information? How Can We Protect It?," J. Natl. Class. Mgmt. Soc., 22, 43–44 (1986).

16. EO 12356, §5.1(b).

17. Atomic Energy Act of 1946, 60 Stat. 755, 766 (Aug. 1, 1946), §10(b)(1).

18. Atomic Energy Act of 1946, 60 Stat. 755, 756, 766 (Aug. 1, 1946), §2(a), §10(a).

19. U.S. Atomic Energy Commission, Seventh Semiannual Report, January 1950, p. 169.

20. U.S. Atomic Energy Commission, Nineteenth Semiannual Report, January 1956, p. 99.

21. U.S. Department of Energy, Washington, D.C., internal document, 1987.

22. E. W. Cleary, ed., McCormick's Handbook of the Law of Evidence, 2nd Ed., West Publishing Co., St. Paul, Minn., 1972, §339. Hereafter cited as McCormick's Law of Evidence.

23. F. James, Jr., Civil Procedure, Little, Brown and Company, Boston, 1965, §7.6. Hereafter cited as "James."

24. James, p. 250, citing Morgan, Some Problems of Proof Under the Anglo-American System of Litigation, 1956, p. 83.

25. McCormick's Law of Evidence, §340.

26. M. F. Jager, Trade Secret Law, Clark Boardman Co., Ltd, New York, 1988, p. 5-10. Hereafter cited as "Jager."

27. Jager, p. 5-10, citing Monsanto Chemical Co. v. Miller, 118 U.S.P.Q. 74, 78 (D. Utah, 1958).

28. Jager, pp. 5-10, 5-11.

29. Jager, p. 5-12.

30. McCormick's Law of Evidence, §341.

31. McCormick's Law of Evidence, §338.

32. McCormick's Law of Evidence, §355.

33. United States v. Reynolds, 345 U.S. 1, 10 (1953).

34. R. C. Tolman, R. F. Bacher, A. H. Compton, E. O. Lawrence, J. R. Oppenheimer, F. H. Spedding, and H. C. Urey, Report of Committee on Declassification, Memorandum to Maj. Gen. L. R. Groves, Nov. 17, 1945, p. 4.

35. U.S. Atomic Energy Commission, Washington, D.C., internal document, 1974.

36. U.S. Atomic Energy Commission, Washington, D.C., internal document, 1974.

37. U.S. Atomic Energy Commission, Washington, D.C., internal document, 1962.

38. W. D. Rowe, An Anatomy of Risk, John Wiley & Sons, New York, §9.4, 1977.

39. McCormick's Law of Evidence, §337.

40. Hayes-Albion v. Kuberski, 364 N.W.2d 609, 616 (1984).

41. Jager, p. 5-61.

42. S. J. Davidson and R. L. DeMay, "Application of Trade Secret Law to New Technology— Unwinding the Tangled Web," Wm. Mitchell L. Rev., 12, 579–621 (l986), p. 607.

43. C. L. Comar, "Risk: A Pragmatic De Minimis Approach," Science, 203(4378) (Jan. 26, 1979), p. 319.

44. J. Fiksel, "De Minimis Risk—From Concept to Practice," Chapter 1 in De Minimis Risk, C. Whipple, ed., Plenum Press, New York, 1987, pp. 3–7. See also other chapters in that book.

45. 42 U.S.C. §2162(a).

46. 42 U.S.C. §2162(c).

47. EO 12356, §1.6(b).

48. DoD 5200.1-R, Chap. II, §2-205.

49. U.S. Department of Defense, Security Classification of Official Information, Instruction 5210.47, Sect. VI.G, Dec. 31, 1964.

50. U.S. Department of Energy, "Guide to the Declassified Areas of Nuclear Energy Research," G- DAR-1, §1.2, August 1984.

51. U.S. Department of Energy classification guide, 1988.

52. Thomas, p. 14.

53. D. R. Corson, Chairman, Panel on Scientific Communication and National Security, Scientific Communication and National Security, National Academy of Sciences, National Academy of Engineering, Institute of Medicine, National Academy Press, Washington, D.C., 1982, p. 49.

On to Chapter Seven.

FAS | Government Secrecy | Library || Index | Search | Join FAS