U.S. Intelligence Agencies Rethink Classification Policy
U.S. intelligence agencies have embarked upon a process to develop a uniform classification policy and a single classification guide that could be used by the entire U.S. intelligence community, according to a newly obtained report (pdf) from the Office of the Director of National Intelligence.
The way that intelligence agencies classify information is not only frustrating to outsiders, as it is intended to be, but it has also impeded interagency cooperation and degraded agency performance.
In order to promote improved information sharing and intelligence community integration, the ODNI undertook a review of classification policies as a prelude towards establishing a new Intelligence Community Classification Guide that would replace numerous individual agency classification policy guides.
The initial ODNI review, completed in January 2008, identified fundamental defects in current intelligence classification policy.
“The definitions of ‘national security’ and what constitutes ‘intelligence’ — and thus what must be classified — are unclear,” the review team found.
“Many interpretations exist concerning what constitutes harm or the degree of harm that might result from improper disclosure of the information, often leading to inconsistent or contradictory guidelines from different agencies.”
“There appears to be no common understanding of classification levels among the classification guides reviewed by the team, nor any consistent guidance as to what constitutes ‘damage,’ ‘serious damage,’ or ‘exceptionally grave damage’ to national security… There is wide variance in application of classification levels.”
Among the recommendations presented in the initial review were that original classification authorities should specify clearly the basis for classifying information, e.g. whether the sensitivity derives from the content of the information, or the source of the information, or the method by which it is analyzed, the date or location it was acquired, etc. Current policy requires that the classifier be “able” to describe the basis for classification but not that he or she in fact do so.
A copy of the unreleased ODNI report on classification policy was obtained by Secrecy News. See “Intelligence Community Classification Guidance: Findings and Recommendations Report,” January 2008.
From Secrecy News’ perspective, the initial ODNI review falls short in two respects.
First, it assumes that consistency in classification is intrinsically desirable and should therefore be imposed by a community-wide classification guide. But consistency is at most a secondary virtue. When a classification policy is poorly justified, it is preferable for it to be inconsistently applied, as in the case of intelligence budget secrecy.
Second, the review does not touch upon what is probably the single most necessary change in intelligence classification policy, namely the need to narrow the definition of intelligence sources and methods that require protection. Almost anything can serve as an intelligence source or method, including a subscription to the daily newspaper. But not every intelligence source or method requires or deserves classification or other protection from disclosure.
Americans are paying too much for almost everything, because the United States has long treated its trucking industry as an artifact to be preserved rather than as an opportunity for innovation.
These ideas aim to advance the detailed policy solutions needed to foster public trust and implement fairness in the adoption of AI across diverse domains, from healthcare and government benefits to rural access, education, and worker protections.
The evidence is clear: algorithmic pay-setting is established in app-based work, and payroll/timekeeping failures show how software can produce systemic wage harm at scale
While a few states have taken steps to implement decision-making mechanisms for certain AI systems, too many leaders are simply accepting narratives about AI’s purported public benefit at face value – jumping to the “how” of AI implementation before thoroughly vetting potential systems and deciding whether they are appropriate to use at all.