Ann Levin, CACI: “Self-Declassifying Documents: A System for Letting the Data Identify When It is Ready for Declassification”


The creation of the National Declassification Center (NDC) by President Obama in December 2009 specifies the centralization and streamlining of all declassification processes with the objective of shortening the time to declassify a document. This enormous task mandates the use of the newest technology to assist with streamlining processes as well as creating consistent, more accurate processes. The Executive Order also specifies that declassification instructions accompany all document classifications. This leads the way to a major new form of support and automation for consistent declassification: by embedding or formally associating declassification information in the document at the time of creation and classification, a system can treat these documents as essentially self-declassifying, rather than requiring intensive manual labor to identify their suitability for declassification and redactions required for release.

The Problem

When classification is first applied to a document or portion, the classification authority must determine the elements of content that cause the classification and the reason that this content must be classified. In addition, according to Executive Order 13526, the classification authority must establish a date or event for the declassification of the material, at the occurrence of which the material automatically becomes declassified. Making full practical use of this information requires a new approach to classification and declassification. Without this, the fact that the material becomes declassified as a matter of policy may have no direct effect in practice; a reviewer must still read the material, consider its original classification, determine the applicability of declassification, and apply the change. This creates essentially duplicate work, and it limits the value of identifying the declassification criteria at the time of the original classification. To take full advantage of the available declassification criteria therefore requires a new approach, with automation that follows the content through its classification life cycle.


We envision a system that enables document self-declassification while maintaining security safeguards and the extent of manual verification and additional review required by policy. The self-declassification (SDC) system should consist of a set of software and networked components to track and apply declassification instructions. Any person who creates a document, or edits one if allowed by policy, must enable the association of applicable classification information. The classification authority must specify not only the classification of each portion but also the conditions for declassification of any classified portion. The system should then aid declassification in two ways:

  • Based on its own recognition, the system should identify when documents become releasable, or when certain document portions as no longer require redaction before release. For example, the system can recognize on its own that the date for declassification has arrived, and can automatically identify this document for declassification, either applying the automatic declassification directly or placing the document in a queue for any final review that a specialized policy may require. The system may also include a capability to track events that are entered by authorized individuals, in order to automatically apply event-driven declassification; this is likely to apply primarily to events that may affect the classification of large numbers of documents.
  • The system should further present documents or redactions for potential release to a reviewer, with a specification of the conditions for release. This allows the reviewer to determine directly whether those conditions have been met, without requiring the reviewer to engage in a labor-intensive process of considering all content in the document for potential sensitivity. This approach is suitable for conditions the system cannot verify directly. Once the reviewer has identified that release criteria are met, the system should handle the document as above, thus relieving the reviewer of the requirement to read the entire document and re-apply the same reasoning that was applied when the document or portion was originally classified.

The SDC system should thus support classification awareness throughout the life cycle of a document.

  • At creation and editing time, classification and declassification information should be associated with the document and its portions, systematically and securely, in a form designed for use with automated systems. Documents for which this information is complete and unchanging should be considered “closed” and unavailable for editing, a condition that the system should enforce.
  • Declassification can be triggered either by automated recognition that the criteria are satisfied or by external system or human request.
  • All declassification should be performed to the maximum extent possible by automated application of previously defined rules.
  • All information about the classification decision and the rules applied in this decision should be made directly available to the individuals who perform any required manual review.


A self-declassification system as described will streamline the declassification process by allowing subject matter experts to make decisions about classification only once. By maintaining full classification and declassification information tightly linked to the document and its portions throughout the content life cycle, the system will take responsibility for maintaining a consistent application of the decision made by the original classifier. In essence, from the point of view of an information consumer or declassification reviewer, each document will perform its own declassification, subject only to the entry of information not directly available to the system and to any specialized manual review that is required for verification and policy.

3 thoughts on “Ann Levin, CACI: “Self-Declassifying Documents: A System for Letting the Data Identify When It is Ready for Declassification”

  1. This is perhaps the Holy Grail of document and content management. It is all technologically feasible but practically impossible in the next 10 years. To get to this state both the Legislative and Executive branches must decide that this level of content understanding and management is a national priority. Failing that. this will remain no more than a pipe dream and the magnitude of the problem with records management, classification management and declassification will continue to grow exponentially each decade. Within the next two decades we will have so many billions of pages of data that we will never be able to identify what we know or manage in any way what we have.

  2. Ann, thank you for submitting this proposal and thanks also to your colleague Kristen for speaking to us at our forum.

    Could you expand on what types of testing and prototyping of the SDC system are taking place right now? Have there been any pilot projects established to test its accuracy against that of an experienced human reviewer? From a risk management perspective, it seems the effectiveness of the system would need to be assured even before cost-effectiveness could be considered.

  3. Mr. Faga, thank you for hosting a forum that allows us to share our ideas.
    At this time we have not begun any testing or prototyping of this system. We put together the concept to see if there was any interest from the community and elicit any thoughts or ideas to begin moving forward to the prototype stage.

Comments are closed.