In 1992, the Department of Energy performed what may have been the most thoughtful and self-critical assessment of classification policy that any government agency has ever carried out. It is now available online.
“This study represents the first fundamental review of classification policy for nuclear weapons and nuclear weapon-related information since the Atomic Energy Act became law [in 1946],” wrote George L. McFadden, then-director of the DOE Office of Security Affairs, in a transmittal letter (pdf). It laid the foundation for the subsequent revision of specific classification practices in the 1995 Fundamental Classification Policy Review and other reforms.
The study asked basic questions — What is the purpose of classification (specifically, of nuclear weapons information)? What is wrong with the status quo? How can it be improved? — and then it considered various answers to these questions. Many of the questions, and a few of the answers, are still valid today. And the study as a whole remains impressive as a model for taking a “fresh look” at classification activity, especially at a time when the National Security Advisor is gathering recommendations for “a more fundamental transformation of the security classification system.”
The 1992 DOE study predated the world wide web, and as far as I know it has not previously been published online. A copy is now posted on the Federation of American Scientists web site. See “Classification Policy Study,” U.S. Department of Energy, July 4, 1992.
AI’s high-stakes outputs directly shape American economic, industrial, and technological outcomes. They must not be accepted entirely on the basis of institutional faith rather than verifiable evidence.
It is difficult to assess from photos whether the Sentinel missile is longer or wider overall than the Minuteman, but the Sentinel clearly has a thicker second stage, which could potentially contribute to the missile’s extended range.
By mandating data centers that are clean, flexible, efficient, and beneficial, Colorado can bring opportunities to communities and enhance its overall infrastructure.
FAS is eager to see implementation of metascience to measure what works, promote novel and effective research programs such as the work of the NSF X-Labs program, and create policy that enables the AI-enabled scientific enterprise to reach its full potential, safely.