A Review of the Department of Energy Classification: Policy and Practice (1995)

A narrowing of the information subject to classification is not itself sufficient to achieve a policy of openness. The Department of Energy (DOE) must find a means for reviewing documents to determine whether they contain only unclassified information and, if so, for releasing them to an often skeptical public. This chapter reviews the process of handling classified documents, including declassification, and suggests a number of both general and specific improvements. Newly generated documents, which are subject to new procedures, policies, and guidelines, are discussed separately.

The changes in policy discussed in Chapters 3 and 4 should serve to minimize overclassification of newly generated documents and to establish a framework for declassification of old documents. But changes in policy must lead to processes for screening and releasing old documents as well.

The Committee encourages efforts by DOE to declassify and publicly release many of the documents that have accumulated over the past 50 years. This action will benefit many groups: scientists who can apply the basic data in their own work; engineers working on problems that may have been solved in connection with classified work; historians interested in details of the past confrontation between the two former superpowers and in U.S. nuclear science and technology programs; DOE site employees tasked with characterizing buried wastes and other contaminants; and members of the general public concerned about the effects of nuclear materials and processes on society, health, safety, and the environment.

Even after information has been declassified, the Department still faces a significant task in making documents containing that information available. The constant challenge for the Department is to identify and then locate the documents of interest to the inquirer, determine whether they are commingled with classified information, and—if not—make them available.

The first aspect of the challenge is simply the huge quantity of classified documents. As noted earlier, according to DOE's current estimate, there are some 280 million classified pages. This estimate is larger by a factor of 10 than the estimate made in late 1993, and it would not be surprising if the estimate continues to increase. Moreover, no master index of historical documents (not even a simple title index) exists, making it difficult for researchers to know what information—classified or not—might exist and where it is located. The Department is currently conducting an inventory of all records, classified and unclassified. This will be at the series, not the individual document, level. (It will include title, dates, general information content, and the highest classification level of documents in series.)

Locating a relevant document is only the first step in its release. As discussed in earlier chapters, a classified document that contains information that has been declassified may contain other information that is still classified, or might be commingled with other still-classified documents in files, boxes, or storage areas. The Department must review the document in question to determine whether it contains any information that is still classified. Under the present system, this is a labor-intensive process that appears ill suited to any effort to declassify documents on a large scale.

Some 200 individuals in the Department currently have declassification authority. As part of the Secretary's Openness Initiative, the staff was increased from 130 people, with a goal of more than 300 people by the beginning of fiscal year 1996.¹ These individuals work page by page, line by line to determine whether a document is properly classified. A decision to declassify a document is customarily reviewed

and verified by a second individual before its classified marking is removed.²

The Committee was told that an experienced declassifier can review approximately 200 pages a day, and we assume that the subsequent reviewer can work twice as fast. If we assume that a declassifier works approximately 240 days a year, reviewing the estimated existing inventory of about 280 million pages of classified documents, at 200 pages per person-day, would require almost 9,000 person-years of effort.³ The cost of this review, on the assumption of direct and indirect costs per employee of $100,000/year, would be $900 million. This is a large enough sum to warrant careful attention to setting priorities for declassification and serious efforts to develop more cost-effective document review methods. But DOE's estimates of its inventory are uncertain, so these and other cost-and-time estimates must be treated as uncertain as well. Despite the uncertainty of the estimates, there is widespread agreement that quicker, more cost-effective methods must be found.

At present, DOE is only beginning to assess the magnitude of the declassification task it faces. That assessment is an essential first step in formulating plans to address the problem. As part of the Openness Initiative, in June 1994 the Under Secretary of Energy requested heads of DOE headquarters elements, field office managers, and contractors to prepare plans for a systematic review of classified records for declassification as part of a Department-wide systematic declassification review program, including proposed schedules and budgets.⁴ It is appropriate to defer commitment to any firm schedule for declassification

review of the relevant collection of classified documents until this information is available. There is still too much uncertainty about the number of classified documents, the cost and effectiveness of alternative declassification methods, the relative urgency of review of different categories of documents, the cost savings achievable by declassification,⁵ and the available resources to warrant any commitment to a schedule at this time. The Committee commends the Department's decision to proceed promptly to obtain the needed information. Although the Committee is not in a position to assess the relative priorities for DOE funds, the declassification effort is important and should be accomplished.

DOE should develop better estimates of the direct costs of classification. Creating a classified document imposes a mortgage on the Department to pay for protection of that document and its ultimate review for declassification. At present, DOE has only very aggregated and approximate figures for security costs.⁶ Requiring each budgetary unit in the Department to estimate the net current year and long-term costs resulting from classification and declassification actions in that year would provide some incentive to minimize needless creation of classified documents and to expedite declassification.⁷

Since resources for the declassification effort will be limited, priorities will have to be set to determine which documents or classes of documents should be declassified first. Setting such priorities is primarily a policy or value judgment that DOE should make with substantial input from stakeholders. We agree with the Joint Security Commission that the

declassification review process should be driven by customer demand.⁸A national DOE advisory committee, such as the Information Policy Advisory Committee proposed in this report, could provide advice about declassification of information and documents bearing on national policy debates.

Information about the health, safety, and the environmental effects of DOE nuclear materials production activities are of particular interest to the communities in the neighborhood of DOE sites. Therefore, establishing priorities for declassification of documents containing such information should have strong local and regional input. Appropriate clearances must be provided to the selected reviewers from the public to enable informed advice.⁹

Another important source of demand for declassification of documents is the historical value of such documents in understanding the nuclear era. The Committee notes that the responsibility for preserving records of historical value lies with the Archivist of the United States, who ultimately must decide which records are to be preserved and which may be destroyed. DOE should set priorities for declassifying documents of historical value using a process like the one it has already established with the National Archives and Records Administration and stakeholders to deal with documents transferred to the National Archives.

DOE estimates that there are already on the order of 3 million classified DOE documents in the National Archives.¹⁰ The DOE is

supporting the Archivist in declassifying these documents by providing personnel. According to the DOE Office of Declassification, as of February 1995 five full-time declassifiers were working at the National Archives, with a sixth reviewer awaiting assignment.¹¹ The Department and the National Archives and Records Administration co-hosted a stakeholders meeting in January 1994 to discuss priorities for declassification review with historians, archivists, and researchers. The priority list that the Archives supplied was approved by the stakeholders and is being followed.¹² This process should be continued and expanded. Priorities for declassification of documents of potential historical interest that have not yet been turned over to the National Archives could also be addressed through this mechanism or through the stakeholder advisory bodies at the site or national level, as appropriate.

Before launching into a large-scale document review (as opposed to some of the specific steps discussed in this and the next section), DOE should complete the fundamental review of classification policy and revise the guidelines accordingly. Only in this way can DOE avoid having to repeat the review under revised guidelines. In the meantime, DOE should proceed with demand-driven reviews. Once new DOE policy is established (with appropriate stakeholder involvement), a more regular process that gives the public a significant voice in setting priorities should be created.

As a first step to making the current process work better, DOE needs to expand the base of trained reviewers. Reviewers could be drawn from several sources. Help might be provided by employees and contractors who are currently examining documents for purposes other than declassification—for example, reviewing documents for dose reconstruction and for waste-characterization studies—and who might be asked to review them simultaneously for sensitive items. Other sources of knowledgeable reviewers include the reservoir of retired people from all parts of the DOE operations and personnel at the weapons labs whose jobs are being changed or eliminated. Such reviewers would need only a minimum of additional training to be effective in the screening effort.¹³

The plans for systematic declassification review should include planning for production of a record index (with at least a listing of unclassified title, author, date, and document number). Indeed, the early generation of an index could facilitate the overall declassification effort. Stakeholder inputs to priority setting would be greatly facilitated if a simple index of the titles of classified documents were available. Such indexes have been produced at several sites as part of systematic efforts to estimate the radiation doses that have been received by workers and the public over time at those sites. The Committee has been told that those indexes have been very helpful.¹⁴

Once such indexes are developed, they should be made publicly available so that actual user requests can help set declassification priorities. DOE has taken an important step in this direction by making available through INTERNET a publicly accessible computer database, called OPENNET,¹⁵ that provides bibliographic and locator information on

declassified documents.¹⁶ The Committee commends this initiative and recommends that any record index of documents not yet declassified also be included in OPENNET. DOE should also consider making any such index available on CD-ROM, accompanied by a text search program to facilitate access, and in DOE information centers and major public libraries.

DOE should take steps to ensure that significant documents are not destroyed before they can be reviewed for declassification. Where there is concern that classification may be misused to hide inappropriate or even illegal actions, there will also be concern that important documents revealing such actions could be destroyed before they can be made available for public scrutiny.¹⁷ As noted in Chapter 2, the Federal Records Act (FRA) forbids the arbitrary destruction of federal records, and an agency must secure the approval of the Archivist before disposing of any records. Although the Committee has no information to suggest that documents are being inappropriately destroyed, DOE should ensure that employees and contractors are reminded of their legal obligations under the FRA not to destroy records except as provided in the Act.

Finally, the Department should assure that declassification and classification decisions are made in a uniform and consistent fashion for both existing and new documents. Anecdotal evidence suggests that different declassification officials applying the same guidance may reach startlingly different conclusions as to the classified content of a document.¹⁸ The reliability of the entire system is suspect if significant variances are a frequent occurrence. If there is a problem, the solution would seem to rest in improved guidance for declassification officials or perhaps in improved training in applying that guidance. This could be a particularly important step if the entire set of guidelines is revised as a result of the fundamental policy review.

Looking ahead to hopes for better document management systems, any computerized system created by the Department should be designed to facilitate declassification of documents and public access to unclassified and declassified documents. The Committee is not aware of the Department's plans in this regard, but facilitating declassification and public access to information would be a logical component in the design of any such system.¹⁹

The current process of line-by-line review for large-scale declassification review entails substantial costs and delays caused by limitations in qualified personnel and available funding.²⁰DOE should develop and evaluate faster and more cost-effective declassification methods. Any expedited process will entail an additional risk of errors—that is, the unintended release of classified information. Given the concern that it is more damaging to national interests to release classified information inadvertently than to fail to release declassified information, it is not surprising that there are objections to proposals for bulk declassification in which large numbers of documents meeting certain criteria, such as age or inclusion in a particular set of files, would be

declassified without individual review.²¹ The Committee agrees that such approaches are not promising.

Nonetheless, DOE should investigate screening methods to identify documents that could be subjected to a lower level of review for declassification. Although bulk declassification may generally entail too much risk of inadvertent release of sensitive information, DOE should investigate intermediate approaches in which large quantities of documents are first screened to segregate them into categories according to the likelihood that they contain classified information. This initial screening could be conducted by suitably trained individuals or perhaps by machine using the artificial intelligence approach discussed below. Documents in the categories least likely to contain such information could then be subjected to a less rigorous declassification review, such as by one person instead of two. They could also be given priority for review, since there would be a greater likelihood that such review would lead to declassification of the documents in the lowest risk category. The Committee understands that the Hanford site is currently conducting a large-scale initial screening of this sort (followed by normal declassification reviews) with substantial success.

DOE should experiment with artificial intelligence (AI) as a screening tool to identify documents most likely to contain classified Restricted Data. Some explorations of AI applications are already under way.²² The hope is that use of AI, combined with optical scanning and optical character recognition (OCR) techniques, could permit machine identification of classified information in text or even in stored images and hence greatly reduce the amount of labor involved in the document review. The Committee encourages this effort, but cautions against over-optimistic

expectations about savings in time and costs. An AI system might indeed reduce the amount of labor involved in screening documents, but it adds a need for labor to scan documents into the system, edit and correct the initial OCR conversion of scanned text images to standardized computer format, and add necessary document headers. Some records are handwritten and therefore difficult to handle by OCR. It is not clear that the net cost and time required for declassification review alone would be greatly reduced. The most promising application of AI might be as a means of identifying documents with a high probability of containing restricted data (RD), so they can go to the bottom of the priority list for declassification review. This would avoid questions about reliability of AI as a means of ensuring that documents do not contain RD.

If AI proves to be a useful method for screening in declassification reviews, its implementation should be integrated with any broader document management system developed by DOE in order to get double value from the cost of entering documents into computer form for the AI system. Once documents are scanned into an information system and the text images are converted to machine-readable form with OCR software, it is easy to add a bibliographic header and to index the text to allow searches that identify documents of interest to a user. Incorporating the AI declassification review system into a broader document management system would ease user access to declassified materials and would facilitate the public release of documents at the time such documents are declassified.²³

As part of its investigation of AI, the Department is seeking to enter into cooperative research and development agreements with the private sector. The Department should not limit this effort to AI, but should open the process up to proposals for innovative methods for quicker and more cost-effective document review, whether or not they involve cutting-edge technology.

In the absence of data about the effectiveness of such approaches, or even about the effectiveness of the current process, it is not possible to evaluate either the cost savings of new approaches to declassification review or the increase in risk of disclosing classified information. Before committing to release large quantities of documents using such methods, DOE should conduct experiments in which various methods are applied to a significant quantity of documents. The error rate should be assessed by the current two-person technique to review the documents that the new method indicates are suitable for declassification. DOE should then carefully reassess the consequences of inadvertent release of classified information of the type that might accidentally be disclosed.

As noted earlier, the Committee has been told that DOE is losing ground on the sheer volume of classified documents: new classified records are being generated faster than others are being declassified. If DOE hopes to achieve a significant reduction in the number of classified documents it must manage and to maintain the inventory at a reduced size, it must take steps to minimize the generation of new classified documents and make those that are created as easy as possible to review for declassification later.

At present DOE does not appear to have a consistent policy to achieve those objectives. The Secretary's directive concerning classification of information related to environment, safety, and health (ES&H) does direct that classification or other dissemination restrictions be used only if they are essential,²⁴ but such a policy has not been applied generally to all documents. This directive, as well as the UCNI

guidelines,²⁵ also encourages segregating classified or otherwise restricted material into an appendix or separable attachment; but again, this does not apply generally to documents containing RD. Exec. Order No. 12,958 requires the portions of documents containing national security information (NSI) to be clearly marked ("portion marking"), but DOE has obtained a waiver of that requirement for DOE documents containing NSI (except for documents that are originally, rather than derivatively, classified, and that contain no RD) and does not use portion marking on RD documents.²⁶

To minimize the needless generation of classified documents and to facilitate declassification, DOE should put in place a number of specific procedures:

1. Classified or otherwise controlled information should be included in documents only if absolutely necessary. The general rule should be to avoid the future costs inherent in classification wherever possible. The Department could take the language from the Secretary's directive concerning documents dealing with ES&H information and apply it broadly to all documents.

2. The classifier of new documents should be required to identify the paragraphs of the DOE classification guide requiring the classification action. The DOE's classification process should require that all newly generated documents be marked in a way that not only facilitates declassification, but also increases accountability and discourages needless or automatic classification. The proposed measure would enhance accountability for the classifying official²⁷ and facilitate declassification as

classification guides change, and would be less onerous than requiring a detailed justification for the classification.

3. Portion marking should be required. As noted above, paragraph-by-paragraph marking of documents containing NSI to indicate the classified portions is practiced in DOD, but not DOE. Such marking for new documents should greatly decrease the burden of declassification of documents once a decision has been made to declassify an area of information.

4. Segregating the classified portions should be encouraged. Classified addenda or tear sheets to separate classified from unclassified information could be used whenever only a small portion of an otherwise unclassified document contains classified information.²⁸ This would facilitate review and release of the unclassified portions even if the classified portion cannot be released.

5. Where segregation is not practical, unclassified versions of significant documents of widespread public interest should be prepared. The Secretary's policy memorandum on ES&H information directs that preparation of an unclassified version be considered if it would allow coherent communication to the public of significant information in these areas. This policy should be extended to other types of documents.

6. Documents should be coded and indexed so they can be easily tracked, identified, and reviewed for declassification when guides change. The Department should take advantage of modern computer technology to facilitate the handling of newly created classified documents. Among the opportunities might be the inclusion of coded information with the electronic form of each document that

preserves a paragraph-by-paragraph justification of the reason for classification by cross reference to the appropriate Classification Guide. Such a system might allow the future rapid declassification of a document (or sections of a document) as the Declassification Guide is modified. Moreover, computer capabilities might be applied to build indices so that a given document or page, which might be found in many files, need be reviewed for classification or declassification only once.

7. Strict guidance for use of derivative classification should be provided. Derivative classification occurs when new documents make reference to material in classified documents and must be themselves classified. But, if material in the source documents is no longer classified, then material in the derivative documents should not be classified. The policy should encourage the generators of derivative documents to verify that the content is still classified. The operative guideline should be; When in doubt about classification, check.

8. Each document, when classified, should carry with it a schedule for declassification review. Guidelines provided for duration of classification of NSI under Exec. Order No. 12,958 would be useful in this endeavor.

********

Taken together, this combination of broad policy changes, investment in better processes, and specific adjustments in procedures should go a long way to address the currently daunting problems of managing DOE's classified document holdings.